You are on page 1of 166

www.it-ebooks.

info

InformationDashboardDesign ByStephenFew ............................................... Publisher:O'Reilly PubDate:January2006 ISBN:0596100167 Pages:223

www.it-ebooks.info

Copyright
Copyright2006StephenFewAllrightsreserved. PrintedinItaly. PublishedbyO'ReillyMedia,Inc.1005GravensteinHighwayNorthSebastopol,CA95472 O'Reillybooksmaybepurchasedforeducational,business,orsalespromotionaluse.Onlineeditionsare alsoavailableformosttitles(safari.oreilly.com).Formoreinformation,contactourcorporate/institutional salesdepartment:8009989938orcorporate@oreilly.com. Editor ProductionEditor ArtDirector CoverDesigner InteriorDesigners ProductionServices PrintHistory January2006: TheO'ReillylogoisaregisteredtrademarkofO'ReillyMedia,Inc.InformationDashboardDesignand relatedtradedressaretrademarksofO'ReillyMedia,Inc. Manyofthedesignationsusedbymanufacturersandsellerstodistinguishtheirproductsareclaimedas trademarks.Wherethosedesignationsappearinthisbook,andO'ReillyMedia,Inc.wasawareofa trademarkclaim,thedesignationshavebeenprintedincapsorinitialcaps. Whileeveryprecautionhasbeentakeninthepreparationofthisbook,thepublisherandauthorassumeno responsibilityforerrorsoromissions,orfordamagesresultingfromtheuseoftheinformationcontained herein. 0596100167 ColleenWheeler Genevieved'Entremont MikeKohnke StephenFew MikeKohnke,TerriDriscoll SpecializedComposition,Inc. FirstEdition.

www.it-ebooks.info

Tomyparents,BobandJoyceFew,whoseprideinmyjourneyhoweverstrangethatjourneymusthave sometimesseemedinstilleddeepdownintomybonestheresolvetokeepplacingonefootinfrontofthe other.

www.it-ebooks.info

AbouttheAuthor
StephenFewhasover20yearsofexperienceasanITinnovator,consultant,andeducator.Today,as PrincipaloftheconsultancyPerceptualEdge,Stephenfocusesondatavisualizationforanalyzingand communicatingquantitativebusinessinformation.Heisworkingtoraiseconsciousnessandtoprovidea treatmentplanthataddressestheneedsofbusinessinthelanguageofbusiness.Hispreviousbook,Show MetheNumbers:DesigningTablesandGraphstoEnlighten,isapowerfulfitnessprogramdesignedto targetthedatapresentationaspectsofthisproblem. Today,fromhisofficeinBerkeley,California,Stephenprovidesconsultingandtrainingservices,speaks frequentlyatconferences,andteachesintheMBAprogramattheUniversityofCaliforniainBerkeley. Moreabouthiscurrentworkcanbefoundatwww.perceptualedge.com.

www.it-ebooks.info

Introduction
Fewphenomenacharacterizeourtimemoreuniquelyandpowerfullythantherapidriseandinfluenceof informationtechnologies.Thesetechnologieshaveunleashedatsunamiofdatathatrollsoverandflattens usinitswake.Tamingthisbeasthasbecomeaprimarygoaloftheinformationindustry.Onetoolthathas emergedfromthiseffortinrecentyearsistheinformationdashboard.Thissinglescreendisplayofthe mostimportantinformationpeopleneedtodoajob,presentedinawaythatallowsthemtomonitor what'sgoingoninaninstant,isapowerfulnewmediumofcommunication.Atleastitcanbe,butonly whenproperlydesigned. Mostinformationdashboardsthatareusedinbusinesstodayfallfarshortoftheirpotential.Therootof theproblemisnottechnologyatleastnotprimarilybutpoorvisualdesign.Toservetheirpurposeandfulfill theirpotential,dashboardsmustdisplayadensearrayofinformationinasmallamountofspaceina mannerthatcommunicatesclearlyandimmediately.Thisrequiresdesignthattapsintoandleveragesthe powerofvisualperceptiontosenseandprocesslargechunksofinformationrapidly.Thiscanbeachieved onlywhenthevisualdesignofdashboardsiscentraltothedevelopmentprocessandisinformedbyasolid understandingofvisualperceptionwhatworks,whatdoesn't,andwhy. Notechnologycandothisforyou.Youmustbringthisexpertisetotheprocess.Takeheartthevisualdesign skillsthatyouneedtodevelopeffectivedashboardscanbelearned,andhelpingyoulearnthemisthesole purposeofthisbook. Iftheinformationisimportant,itdeservestobecommunicatedwell.

www.it-ebooks.info

Acknowledgments
WithoutadoubtIowethegreatestdebtofgratitudetothemanysoftwarevendorswhohavedoneso muchtomakethisbooknecessarybyfailingtoaddressorevencontemplatethevisualdesignneedsof dashboards.Theirkinddisregardforvisualdesignhasgivenmefocus,ignitedmypassion,andguaranteed mylivelihoodforyearstocome. Now,ontothosewhohavecontributedmoredirectlyandpersonallytothiseffort.Asaman,Iwillneverbe abletocreate,shelter,andnourishanemerginglifewithinthisbodyofmine.Inrecentyears,however,I haverecognizedandpursuedtheopportunitytobreathelifeintotheproductsofmyimaginationandpass themontotheworldintheformofbooks.Writingabookisabitlikebearingachild.Workingwitha publishertohelpthechildlearntowalkbeforeventuringintotheworldisalessonintrust.Thefolksat O'ReillyMediahavetaughtmetoentrusttothembeginningwithunspeakableangst,butproceeding throughunfalteringstepstowardeverincreasingcomfortthecollegialcareofthisbelovedchild.Manyat O'Reillyhavecontributedsomuch,buttwoinparticularhavestoodbymysidefromthebeginningwith soothingvoicesofconfidenceandcalm.Myeditor,ColleenWheeler,knewwhentolisteninsilence,when toteasemeoutofmyopia,andwhentogentlyremindmethatIwasinherconsiderateandconsiderable care.Myacquisitionseditor,SteveWeiss,soughtmeoutandwooedmethroughmonthsofthoughtful discussionintotheO'Reillyfold.Hegaveassurancesandhasmadesurethattheywerefulfilled.

www.it-ebooks.info

Sommario
Copyright.......................................................................................................................................................3 AbouttheAuthor ...........................................................................................................................................5 Introduction...................................................................................................................................................6 Acknowledgments.........................................................................................................................................7 Chapter1.ClarifyingtheVision....................................................................................................................... 11 1.1.AllThatGlittersIsNotGold.................................................................................................................. 12 1.2.EvenDashboardsHaveaHistory.......................................................................................................... 14 1.3.DispellingtheConfusion....................................................................................................................... 15 1.3.1.WhatIsaDashboard?........................................................................................................................26 1.4.ATimelyOpportunity........................................................................................................................... 28 Chapter2.VariationsinDashboardUsesandData........................................................................................ 29 2.1.CategorizingDashboards...................................................................................................................... 30 2.1.1.ClassifyingDashboardsbyRole..................................................................................................... 31 2.2.TypicalDashboardData........................................................................................................................ 33 2.2.1.TheCommonThreadinDashboardDiversity................................................................................ 33 Chapter3.ThirteenCommonMistakesinDashboardDesign........................................................................ 38 3.1.ExceedingtheBoundariesofaSingleScreen....................................................................................... 39 3.1.1.FragmentingDataintoSeparateScreens...................................................................................... 40 3.1.2.RequiringScrolling......................................................................................................................... 42 3.2.SupplyingInadequateContextfortheData......................................................................................... 43 3.3.DisplayingExcessiveDetailorPrecision............................................................................................... 45 3.4.ChoosingaDeficientMeasure.............................................................................................................. 46 3.5.ChoosingInappropriateDisplayMedia................................................................................................ 47 3.6.IntroducingMeaninglessVariety.......................................................................................................... 51 3.7.UsingPoorlyDesignedDisplayMedia.................................................................................................. 52 3.8.EncodingQuantitativeDataInaccurately............................................................................................. 56 3.9.ArrangingtheDataPoorly.................................................................................................................... 56 3.10.HighlightingImportantDataIneffectivelyorNotatAll..................................................................... 57 3.11.ClutteringtheDisplaywithUselessDecoration................................................................................. 58 3.12.MisusingorOverusingColor.............................................................................................................. 61 3.13.DesigninganUnattractiveVisualDisplay........................................................................................... 62 Chapter4.TappingintothePowerofVisualPerception................................................................................ 64 4.1.UnderstandingtheLimitsofShortTermMemory............................................................................... 65 4.2.VisuallyEncodingDataforRapidPerception....................................................................................... 67

www.it-ebooks.info

4.2.1.AttributesofColor......................................................................................................................... 69 4.2.2.AttributesofForm......................................................................................................................... 70 4.2.3.AttributesofPosition.................................................................................................................... 71 4.2.4.AttributesofMotion ...................................................................................................................... 71 4.2.5.EncodingQuantitativeVersusCategoricalData............................................................................ 71 4.2.6.LimitstoPerceptualDistinctness.................................................................................................. 73 4.2.7.UsingVividandSubtleColorsAppropriately................................................................................. 74 4.3.GestaltPrinciplesofVisualPerception................................................................................................. 74 4.3.1.ThePrincipleofProximity............................................................................................................. 75 4.3.2.ThePrincipleofSimilarity.............................................................................................................. 75 4.3.3.ThePrincipleofEnclosure............................................................................................................. 76 4.3.4.ThePrincipleofClosure................................................................................................................. 77 4.3.5.ThePrincipleofContinuity............................................................................................................ 78 4.3.6.ThePrincipleofConnection.......................................................................................................... 78 4.4.ApplyingthePrinciplesofVisualPerceptiontoDashboardDesign.....................................................79 Chapter5.EloquenceThroughSimplicity....................................................................................................... 80 5.1.CharacteristicsofaWellDesignedDashboard.................................................................................... 81 5.1.1.CondensingInformationviaSummarizationandException.........................................................82 5.2.KeyGoalsintheVisualDesignProcess................................................................................................ 83 5.2.1.ReducetheNonDataPixels.......................................................................................................... 86 5.2.2.EnhancetheDataPixels................................................................................................................ 94 Chapter6.EffectiveDashboardDisplayMedia............................................................................................. 101 6.1.SelecttheBestDisplayMedium......................................................................................................... 102 6.2.AnIdealLibraryofDashboardDisplayMedia.................................................................................... 106 6.2.1.Graphs.........................................................................................................................................107 6.2.2.Icons .............................................................................................................................................131 6.2.3.Text..............................................................................................................................................133 6.2.4.Images.........................................................................................................................................133 6.2.5.DrawingObjects.......................................................................................................................... 134 6.2.6.Organizers....................................................................................................................................135 6.3.Summary.............................................................................................................................................137 Chapter7.DesigningDashboardsforUsability............................................................................................. 138 7.1.OrganizetheInformationtoSupportItsMeaningandUse...............................................................139 7.1.1.OrganizeGroupsAccordingtoBusinessFunctions,Entities,andUse........................................139 7.1.2.ColocateItemsThatBelongtotheSameGroup........................................................................ 139

www.it-ebooks.info

7.1.3.DelineateGroupsUsingtheLeastVisibleMeans........................................................................ 140 7.1.4.SupportMeaningfulComparisons............................................................................................... 141 7.1.5.DiscourageMeaninglessComparisons........................................................................................ 142 7.2.MaintainConsistencyforQuickandAccurateInterpretation...........................................................143 7.3.MaketheViewingExperienceAestheticallyPleasing........................................................................ 143 7.3.1.ChooseColorsAppropriately....................................................................................................... 144 7.3.2.ChooseHighResolutionforClarity.............................................................................................. 145 7.3.3.ChoosetheRightText .................................................................................................................. 145 7.4.DesignforUseasaLaunchPad.......................................................................................................... 145 7.5.TestYourDesignforUsability............................................................................................................. 146 Chapter8.PuttingItAllTogether.................................................................................................................. 147 8.1.SampleSalesDashboard..................................................................................................................... 148 CritiqueofSalesDashboardExample1................................................................................................. 151 CritiqueofSalesDashboardExample2................................................................................................. 152 CritiqueofSalesDashboardExample3................................................................................................. 153 CritiqueofSalesDashboardExample4................................................................................................. 154 CritiqueofSalesDashboardExample5................................................................................................. 155 CritiqueofSalesDashboardExample6................................................................................................. 156 CritiqueofSalesDashboardExample7................................................................................................. 157 CritiqueofSalesDashboardExample8................................................................................................. 158 8.2.SampleCIODashboard....................................................................................................................... 159 8.3.SampleTelesalesDashboard.............................................................................................................. 161 8.4.SampleMarketingAnalysisDashboard.............................................................................................. 162 8.5.AFinalWord.......................................................................................................................................164 AppendixA.RecommendedReading............................................................................................................ 165 Colophon.......................................................................................................................................................166

www.it-ebooks.info

Chapter1.ClarifyingtheVision
Dashboardsofferauniqueandpowerfulsolutiontoanorganization'sneedforinformation,buttheyusually fallfarshortoftheirpotential.Dashboardsmustbeseeninhistoricalcontexttounderstandandappreciate howandwhythey'vecomeabout,whythey'vebecomesopopular,andwhydespitemanyproblemsthat underminetheirvaluetodaytheyofferbenfitsworthpursuing.Todate,littleseriousattentionhasbeen giventotheirvisualdesign.Thisbookstrivestofillthisgap.However,confusionabounds,demandinga cleardefinitionofdashboardsbeforewecanexplorethevisualdesignprinciplesandpracticesthatmustbe appliediftheyaretoliveuptotheiruniquepromise.

www.it-ebooks.info

Problemswithdashboardstoday Dashboardsinhistoricalcontext Currentconfusionaboutwhatdashboardsare Aworkingdefinitionof"dashboard" Atimelyopportunityfordashboards Aboveallelse,thisisabookaboutcommunication.Itfocusesexclusivelyonaparticularmediumof communicationcalledadashboard.Inthefastpacedworldofinformationtechnology(IT),termsare constantlychanging.Justwhenyouthinkyou'vewrappedyourmindaroundthelatestinnovation,the technologylandscapeshiftsbeneathyouandyoumuststruggletoremainupright.Thisiscertainlytrueof dashboards. Liveyourlifeonthesurfaceoftheseshiftingsands,andyou'llnevergetyourbalance.Lookalittledeeper, however,andyou'lldiscovermorestableground:abedrockofobjectives,principles,andpracticesfor informationhandlingthatremainsrelativelyconstant.Dashboardsareuniqueinseveralexcitinganduseful ways,butdespitethehypesurroundingthem,whattheyareandhowtheyworkasameansofdelivering informationarecloselyrelatedtosomelongfamiliarconceptsandtechnologies.It'stimetocutthrough thehypeandlearnthepracticalskillsthatcanhelpyoutransformdashboardsfromyetanotherfadriding thewavesofthetechnologybuzzintotheeffectivemeanstoenlightenthattheyreallycanbe. Today,everybodywantsadashboard.Likemanynewcomerstothetechnologyscene,dashboardsaresexy. Softwarevendorsworkhardtomaketheirdashboardsshimmywithsexappeal.Theytaunt,"Youdon't wanttobetheonlycompanyinyourneighborhoodwithoutone,doyou?" Theywarn,"Youcannolongerlivewithoutone."Theywhispersweetly,"Stillhaven'tachievedthe expectedreturnoninvestment(ROI)fromyourexpensivedatawarehouse?Juststickadashboardinfront ofitandwatchthemoneypourin."Bestillmyheart. Thosegauges,meters,andtrafficlightsaresodamnflashy!Youcanimaginethatyou'resittingbehindthe wheelofaGermanengineeredsportscar,feelingthewindwhipthroughyourhairasyouteararound curvesontheautobahnathighspeeds,allwithoutleavingyourdesk. Everyonewantsadashboardtoday,butoftenforthewrongreasons.Restassured,however,that somewherebeyondthehypeandsizzlelivesauniqueandeffectivesolutiontofamiliarbusinessproblems thatarerootedinaveryrealneedforinformation.That'sthedashboardthatdeservestoliveonyour screen.

1.1.AllThatGlittersIsNotGold
Dashboardscanprovideauniqueandpowerfulmeanstopresentinformation,buttheyrarelyliveupto theirpotential.Mostdashboardsfailtocommunicateefficientlyandeffectively,notbecauseofinadequate technology(atleastnotprimarily),butbecauseofpoorlydesignedimplementations.Nomatterhowgreat thetechnology,adashboard'ssuccessasamediumofcommunicationisaproductofdesign,aresultofa displaythatspeaksclearlyandimmediately.Dashboardscantapintothetremendouspowerofvisual perceptiontocommunicate,butonlyifthosewhoimplementthemunderstandvisualperceptionandapply

www.it-ebooks.info

thatunderstandingthroughdesignprinciplesandpracticesthatarealignedwiththewaypeopleseeand think.Softwarewon'tdothisforyou.It'suptoyou. Unfortunately,mostvendorsthatprovidedashboardsoftwarehavedonelittletoencouragetheeffective useofthismedium.Theyfocustheirmarketingeffortsonflashanddazzlethatsubvertthegoalsofclear communication.Theyfighttowinourinterestbymaximizingsizzle,highlightingflashydisplaymechanisms thatappealtoourdesiretobeentertained.Onceimplemented,however,thesecutedisplayslosetheir sparkinamatterofdaysandbecomejustplainannoying.Aneffectivedashboardistheproductnotofcute gauges,meters,andtrafficlights(Figure11),butratherofinformeddesign:moresciencethanart,more simplicitythandazzle.Itis,aboveallelse,aboutcommunication.

Figure11.Atypicalflashydashboard.Can'tyoujustfeeltheenginerevving?

Thisfailurebysoftwarevendorstofocusonwhatweactuallyneedishardlyuniquetodashboards.Most softwaresuffersfromthesameshortcomingdespiteallthehypeaboutuserfriendliness,itisdifficultto use.Thissadstateissocommon,andhasbeenthecaseforsolong,we'vegrownaccustomedtothepain. Onthoseoccasionswhenthisuglytruthbreechesthesurfaceofourconsciousness,weusuallyblamethe problemonourselvesratherthanthesoftware,framingitintermsof"computerilliteracy."Ifwecouldonly adaptmoretothecomputerandhowitworks,therewouldn'tbeaproblemorsowereason.Inhis insightfulbookentitledTheInmatesAreRunningtheAsylum,masterdesignerAlanCooperwrites: Thesadthingaboutdancingbearware[Cooper'stermforpoorlydesignedsoftware thatisdifficulttouse]isthatmostpeoplearequitesatisfiedwiththelumbering beast.Onlywhentheyseesomerealdancingdotheybegintosuspectthatthereisa worldbeyondursineshuffling.Sofewsoftwarebasedproductshaveexhibitedany

www.it-ebooks.info

realdancingabilitythatmostpeoplearehonestlyunawarethatthingscouldbe betteralotbetter.1 Cooperarguesthatthisfailureisrootedinanapproachtosoftwaredevelopmentthatsimplydoesn'twork. Inagenuineattempttopleasetheircustomers,softwareengineersfocusoncheckingalltheitems,oneby one,offoflistsofrequestedfeatures.Thisapproachmakessensetotechnologyorientedsoftware engineers,butitresultsinlumberingbeasts.Customersareexpertinknowingwhattheyneedto accomplish,butnotinknowinghowsoftwareoughttobedesignedtosupporttheirneeds.Allowing customerstodesignsoftwarethroughfeaturerequestsistheworstformofdisasterbycommittee. Softwarevendorsshouldbringdesignvisionandexpertisetothedevelopmentprocess.Theyoughtto knowthedifferencebetweensuperficialglitzandwhatreallyworks.Butthey'resoexhaustedfromworking ungodlyhourstryingtosqueezemorefeaturesintothenextreleasethatthey'releftwithnotimetodothe researchneededtodiscoverwhatactuallyworks,oreventostepbackandobservehowtheirproductsare reallybeingused(andfailingintheprocess). Thepartofinformationtechnologythatfocusesonreportingandanalysiscurrentlygoesbythename businessintelligence(BI).Todate,BIvendorshaveconcentratedondevelopingtheunderlyingtechnologies thatareusedtogatherdatafromsourcesystems,transformdataintoamoreusableform,storedatain highperformancedatabases,accessdataforuse,andpresentdataintheformofreports.Tremendous progresshasbeenmadeintheseareas,resultinginrobusttechnologiesthatcanhandlehugerepositories ofdata.However,whilewehavemanagedtowarehouseagreatdealofinformation,wehavemadelittle progressinusingthatinformationeffectively.Relativelylittleefforthasbeendedicatedtoengaginghuman intelligence,whichiswhatthisindustry,bydefinition,issupposedtobeabout. AglossaryontheGartnerGroup'swebsitedefinesbusinessintelligenceas"Aninteractiveprocessfor exploringandanalyzingstructured,domainspecificinformationtodiscernbusinesstrendsorpatterns, therebyderivinginsightsanddrawingconclusions" (http://www.gartner.com/6_help/glossary/GlossaryB.jsp).Toprogressinthisworthwhileventure,theBI industrymustshiftitsfocusnowtoanengaginginteractionwithhumanperceptionandintelligence.Todo this,vendorsmustbasetheireffortsonafirmunderstandingofhowpeopleperceiveandthink,building interfaces,visualdisplays,andmethodsofinteractionthatfitseamlesslywithhumanability.

1.2.EvenDashboardsHaveaHistory
Inmanyrespects,"dashboard"issimplyanewnamefortheExecutiveInformationSystems(EISs)first developedinthe1980s.Theseimplementationsremainedexclusivelyintheofficesofexecutivesandnever numberedmorethanafew,soitisunlikelythatyou'veeveractuallyseenone.Isatthroughafewvendor demosbackinthe1980sbutneverdidseeanactualsysteminuse.TheusualpurposeofanEISwasto displayahandfulofkeyfinancialmeasuresthroughasimpleinterfacethat"evenanexecutivecould understand."Thoughlimitedinscope,thegoalwasvisionaryandworthwhile,butaheadofitstime.Back then,beforedatawarehousingandbusinessintelligencehadevolvedthenecessarydatahandling methodologiesandgivenshapetothenecessarytechnologies,thevisionsimplywasn'tpractical;itcouldn't berealizedbecausetherequiredinformationwasincomplete,unreliable,andspreadacrosstoomany disparatesources.Thus,inthesamedecadethattheEISarose,italsowentintohibernation,preservingits visionintheshadowsuntilthetimewasripeThatis,untilnow.
1

TheInmatesAreRunningtheAsylum(Indianapolis,IN:SAMSPublishing,1999),59.

www.it-ebooks.info

Duringthe1990s,datawarehousing,onlineanalyticalprocessing(OLAP),andeventuallybusiness intelligenceworkedaspartnerstotamethewildonslaughtoftheinformationage.Theemphasisduring thoseyearswasoncollecting,correcting,integrating,storing,andaccessinginformationinwaysthat soughttoguaranteeitsaccuracy,timeliness,andusefulness.Fromtheearlydaysofdatawarehousingon intotheearlyyearsofthisnewmillennium,theefforthaslargelyfocusedonthetechnologies,andtoa lesserdegreethemethodologies,neededtomakeinformationavailableanduseful.Thedirectbeneficiaries sofarhavemostlybeenfolkswhoarehighlyproficientintheuseofcomputersandabletousethe availabletoolstonavigatethroughlarge,oftencomplexdatabases. Whatalsoemergedintheearly1990s,butdidn'tbecomepopularuntillateinthatdecade,wasanew approachtomanagementthatinvolvedtheidentificationanduseofkeyperformanceindicators(KPIs), introducedbyRobertS.KaplanandDavidP.NortonastheBalancedScorecard.Theadvancesindata warehousinganditstechnologypartnerssetthestageforthisnewinterestinmanagementthroughtheuse ofmetricsandnotjustfinancialmetricsthatstilldominatesthebusinesslandscapetoday.Business PerformanceManagement(BPM),asitisnowcommonlyknown,hasbecomeaninternational preoccupation.Theinfrastructurebuiltbydatawarehousingandthelike,aswellastheinterestofBPMin metricsthatcanbemonitoredeasily,togethertilledandfertilizedthesoilinwhichthehibernatingseedsof EIStypedisplayswereonceagainabletogrow. Whatreallycausedheadstoturninrecognitionofdashboardsasmuchmorethanyoureverydayfledgling technology,however,wastheEnronscandalin2001.Theaftermathputnewpressureoncorporationsto demonstratetheirabilitytocloselymonitorwhatwasgoingonintheirmidstandtotherebyassure shareholdersthattheywereincontrol.Thisincreasedaccountability,combinedwiththeconcurrent economicdownturn,sentChiefInformationOfficers(CIOs)onamissiontofindanythingthatcouldhelp managersatalllevelsmoreeasilyandefficientlykeepaneyeonperformance.MostBIvendorsthathadn't alreadystartedofferingadashboardproductsoonbegantodoso,sometimesbycleverlychangingthe nameofanexistingproduct,sometimesbyquicklypurchasingtherightstoanexistingproductfroma smallervendor,andsometimesbycobblingtogetherpiecesofproductsthatalreadyexisted.The marketplacesoonofferedavastarrayofdashboardsoftwarefromwhichtochoose.

1.3.DispellingtheConfusion
Likemanyproductsthathitthehightechscenewithasplash,dashboardsareveiledinmarketinghype. VirtuallyeveryvendorintheBIspaceclaimstoselldashboardsoftware,butfewclarifywhatdashboards actuallyare.I'mremindedoftheearlyyearsofdatawarehousing,wheneagertolearnaboutthisnew approachtodatamanagementIaskedmyIBMaccountmanagerhowIBMdefinedtheterm.Hisresponse wasclassicandrefreshinglycandid:"BydatawarehousingweatIBMmeanwhateverthecustomerthinksit means."Irealizethatthiswasn'tIBM'sofficialdefinition,whichI'msureexistedsomewhereintheir literature,butitwasmybluesuitedfriend'swayofsayingthatasasalesperson,itwasusefultoleavethe termvagueandflexible.Aslongasaproductorserviceremainsundefinedorlooselydefined,itiseasyto claimthatyourcompanysellsit. Thoseraresoftwarevendorsthathavetakenthetimetodefinethetermintheirmarketingliteraturestart withthespecificfeaturesoftheirproductsasthecoreofthedefinition,ratherthanagenericdescription. Asaresult,vendordefinitionstendtobeselfvalidatinglistsoftechnologiesandfeatures.Forexample,Dr. GregoryL.Hovis,DirectorofProductDeploymentforSnippetsSoftware,Inc.,asserts:

www.it-ebooks.info

AbletouniversallyconnecttoanyXMLorHTMLdatasource,robustdashboard productsintelligentlygatheranddisplaydata,providingbusinessintelligence withoutinterruptingworkflowAnenterprisedashboardischaracterizedbya collectionofintelligentagents(orgauges),eachperformingfrequentbidirectional communicationwithdatasources.Likeavirtualstaffof24x7analysts,eachagentin thedashboardintelligentlygathers,processesandpresentsdata,generatingalerts andrevisingactionsasconditionschange.1 AnarticleintheJune16,2003editionofComputerworldcitesstatisticsfromastudydonebyAMR Research,Inc.,whichdeclaresthat"morethanhalfofthe135companiesrecentlysurveyedare implementingdashboards."2 Unfortunately,theauthornevertellsuswhatdashboardsare.Heteasesuswithhints,statingthat dashboardsandscorecardsareBItoolsthat"havefoundanewhomeinthecubicles,"havingmovedfrom wheretheyonceresided(exclusivelyinexecutivesuites)underthenameExecutiveInformationSystems. Hegivesexamplesofhowdashboardsarebeingusedandspeaksoftheirbenefits,butleavesittousto piecetogetherasenseofwhattheyare.TheclosesthecomestoadefinitioniswhenhequotesJohn HagertyofAMRResearch,Inc.:"Dashboardsandscorecardsareaboutmeasuring." Whileconductinganextensiveliteraturereviewin2003insearchofagoodworkingdefinition,Ivisited DataWarehousingOnline.comandclickedonthelinkto"ExecutiveDashboard"articles.Inresponse,I receivedthesame18webpagesoflinksthatIfoundwhenIseparatelyclickedonlinksfor"Balanced Scorecard,""DataQualityandIntegration,"and"DataMining."Eitherthelinksweren'tworkingproperly,or thiswebportalforthedatawarehousingindustryatthetimebelievedthatthesetermsallmeantthesame thing.3 Ifinallydecidedtobeginthetaskofdevisingaworkingdefinitionofmyownbyexaminingeveryexampleof adashboardIcouldfindontheWeb,insearchoftheircommoncharacteristics.Youmightfindit interestingtotakeasimilarjourney.Inthenextfewpages,you'llseescreenshotsofanassortmentof dashboards,whichweremostlyfoundonthewebsitesofvendorsthatselldashboardsoftware.Takethe timenowtobrowsethroughtheseexamplesandseeifyoucandiscerncommonthreadsthatmightbe wovenintoausefuldefinition.

GregoryL.Hovis,"StopSearchingforInformationMonitoritwithDashboardTechnology,"DMDirect,February2002. MarkLeon,"DashboardDemocracy,"Computerworld,June16,2003 3 Byincludingtheseexamplesfromthewebsitesofsoftwarevendorsandafewothersources,Idonotmeanto endorseanyofthesedashboardsorthesoftwareproductsusedtocreatethemasexamplesofgooddesign,noras extraordinaryexamplesofpoordesign.TovaryingdegreestheyallexhibitvisualdesignproblemsthatI'lladdressin laterchapters.


2 1

www.it-ebooks.info


Figure12.ThisdashboardfromBusinessObjectsreliesprimarilyongraphicalmeanstodisplayaseriesofperformance measures.alongwithalistofalerts,Noticethatthetitleofthisdashboardis"MyKPIs."Keyperformanceindicatorsand dashboardsappeartobesynonymousinthemindsofmostvendors.Noticethegaugesaswell.We'llseequiteafewofthem.

www.it-ebooks.info

Figure13.ThisdashboardfromOracleCorporationdisplaysacollectionofsalesmeasuresforanalyzingproductperformanceby category.Allofthemeasuresaredisplayedgraphically.We'llfindthatthisemphasisongraphicaldisplaymediaisfairly common.

Figure14.ThisdashboardfromInformaticaCorporationdisplaysmeasuresofrevenuebysaleschannelalongwithalistof reportsthatcanbeviewedseparately.Thepredominanceofgraphicaldisplaymediathatweobservedontheprevious dashboardsappearsonthisoneaswell,notablyintheformofmetersdesignedtolooklikespeedometers.Thelistofreports addsportalfunctionality,enablingthisdashboardtooperateasalaunchpadtocomplementaryinformation.

www.it-ebooks.info


Figure15.ThisdashboardfromPrincipaprovidesanoverviewofacompany'sfinancialperformancecomparedtotargetsforthe monthofMarch,bothintabularformandasaseriesofgauges.Theinformationcanbetailoredbyselectingdifferentmonths andamountsofhistory.Onceagain,weseeastrongexpressionofthedashboardmetaphor,thistimeintheformofgraphical devicesthatweredesignedtolooklikefuelgauges.

www.it-ebooks.info


Figure16.ThisdashboardfromCognos,Inc.displaysatableandfivegraphsoneintheformofaworldmaptocommunicate salesinformation.Despitetheonetable,there'sacontinuedemphasisongraphicalmedia.Noticealsothatathemeregarding thevisualnatureandneedforvisualappealofdashboardsisemergingintheseexamples.

www.it-ebooks.info


Figure17.ThisdashboardfromHyperionSolutionsCorporationdisplaysregionalsalesrevenueinthreeforms:onamap,ina bargraph,andinatable.Datacanbefilteredbymeansofthreesetsofradiobuttonsontheleft.Thesefilteringmechanisms buildrudimentaryanalyticalfunctionalityintothisdashboard.Visualdecorationreinforcesthethemethatdashboards intentionallystriveforvisualappeal.

www.it-ebooks.info


Figure18.ThisdashboardfromCordaTechnologies,Inc.featuresflightloadingmeasuresforanairlineusingfourpanelsof graphs.Hereagainweseeanattentiontothevisualappealofthedisplay.Noticealsointheinstructionsatthetopthatan abilitytointeractwiththegraphshasbeenbuiltintothedashboard,sothatuserscanaccessadditionalinformationinpopups anddrillintogreaterlevelsofdetail.

www.it-ebooks.info


Figure19.ThisdashboardfromVisualMining,Inc.displaysvariousmeasuresofacity'stransitsystemtogivetheexecutivesin chargeaquickoverviewofthesystem'scurrentandhistoricalperformance.Useofthecolorsgreen,yellow,andredtoindicate good,satisfactory,andbadperformance,asyoucanseeonthethreegraphicaldisplaysarrangedhorizontallyacrossthemiddle, iscommonondashboards.

www.it-ebooks.info

Figure110.ThisdashboardfromInfommersion,Inc.givesexecutivesofahotelchainthemeanstoviewmultiplemeasuresof performance,onehotelatatime.Itisnotunusualfordashboardstodividethefullsetofdataintoindividualviews,asthisone doesbyusingthelistboxintheupperleftcornertoenableviewerstoselectanindividualhotelbylocation.Thegreatcarethat weseeinthisexampletorealisticallyreproducethedashboardmetaphor,evendowntothesheenonpolishedmetal,isan effortthatmanyvendorstakequiteseriously.

Figure111.ThisdashboardfromCelequestCorporationintegratesaseriesofrelatedtablesandgraphsthatallowexecutivesto viewseveralaspectsofsalessimultaneously.Itexhibitsanefforttocombinearichsetofrelateddataonthescreentoprovidea comprehensiveoverviewofacompany'ssalesperformance.

www.it-ebooks.info


Figure112.ThisdashboardfromGeneralElectric,calleda"digitalcockpit,"providesatabularsummaryofperformance, complementedbyacolorcodedindicatorlightforeachmeasure'sstatus.Ratherthanadashboarddesignedbyasoftware vendortoexhibititsproduct,thisisanactualworkingdashboardthatwasdesignedbyacompanytoserveitsownbusiness needs.Inthisexample,noeffortwasmadetoliterallyrepresentthedashboard(orcockpit)metaphor.

www.it-ebooks.info

Figure113.ThisdashboardisusedbytheTreasuryBoardofCanadatomonitortheperformanceofaproject.Hereagainwe haveadashboardthatwasdesignedbyanorganizationforitsownuse.Thistime,thedashboardmetaphormakesatoken appearanceintheformofgauges.Thetrafficlightcolorsgreen,yellow,andredherewiththeadditionofblueforthe exceptionallygoodstatusof"aheadofschedule"arealsoused.Unlikesomeoftheexamplesthatwe'veseenthatdisplayed relativelylittleinformation,thisonemakestheattempttoprovidethecomprehensiveoverviewthatwouldbeneededto effectivelymonitorprogressandperformance.

1.3.1.WhatIsaDashboard?
Asyouhavenodoubtdeterminedbyexaminingtheseexamples,there'safairdegreeofdiversityinthe productsthatgobythename"dashboard."Oneofthefewcharacteristicsthatmostvendorsseemtoagree onisthatforsomethingtobecalledadashboarditmustincludegraphicaldisplaymechanismssuchas trafficlightsandavarietyofgaugesandmeters,manysimilartothefuelgaugesandspeedometersfound inautomobiles.ThisclearlyassociatesBIdashboardswiththefamiliarversionsfoundincars,thereby leveragingausefulmetaphorbutthemetaphoralonedoesn'tprovideanadequatedefinition.Aboutthe onlyotherthreadthatiscommontothesedashboardexamplesisthattheyusuallyattempttoprovidean overviewofsomethingthat'scurrentlygoingoninthebusiness. Afteragreatdealofresearchandthought,Icomposedadefinitionofmyownthatcapturestheessenceof whatIbelieveadashboardis(clearlybiasedtowardthecharacteristicsofthismediumthatIfindmost usefulandunique).Toserveuswell,thisdefinitionmustclearlydifferentiatedashboardsfromotherforms ofdatapresentation,anditmustemphasizethosecharacteristicsthateffectivelysupportthegoalof communication.Here'smydefinition,whichoriginallyappearedinIntelligentEnterprisemagazine: Adashboardisavisualdisplayofthemostimportantinformationneededtoachieve oneormoreobjectives;consolidatedandarrangedonasinglescreensothe informationcanbemonitoredataglance.1 Justasthedashboardofacarprovidescriticalinformationneededtooperatethevehicleataglance,aBI dashboardservesasimilarpurpose,whetheryou'reusingittomakestrategicdecisionsforahuge corporation,runthedailyoperationsofateam,orperformtasksthatinvolvenoonebutyourself.The meansisasinglescreendisplay,andthepurposeistoefficientlymonitortheinformationneededto achieveone'sobjectives. Visualdisplayofthemostinformationneededtoachieveoneormoreobjectives whichfitsentirelyonasinglecomputerscreensoitcanbemonitoredataglance Let'sgooverthesalientpoints: Dashboardsarevisualdisplays.Theinformationonadashboardispresentedvisually,usuallyasa combinationoftextandgraphics,butwithanemphasisongraphics.Dashboardsarehighlygraphical,not becauseitiscute,butbecausegraphicalpresentation,handledexpertly,canoftencommunicatewith greaterefficiencyandrichermeaningthantextalone.Howcanyoubestpresenttheinformationsothat humaneyescantakeitinquicklyandhumanbrainscaneasilyextractthecorrectandmostimportant meaningsfromit?Todesigndashboardseffectively,youmustunderstandsomethingaboutvisual perceptionwhatworks,whatdoesn't,andwhy. Dashboardsdisplaytheinformationneededtoachievespecificobjectives.Toachieveevena singleobjectiveoftenrequiresaccesstoacollectionofinformationthatisnototherwiserelated,
1

StephenFew,"DashboardConfusion,"IntelligentEnterprise,March20,2004.

www.it-ebooks.info

oftencomingfromdiversesourcesrelatedtovariousbusinessfunctions.Itisn'taspecifictypeof information,butinformationofwhatevertypethatisneededtodoajob.Itisn'tjustinformation thatisneededbyexecutivesorevenbymanagers;itcanbeinformationthatisneededbyanyone whohasobjectivestomeet.TherequiredinformationcanbeandoftenisasetofKPIs,butnot necessarily,forothertypesofinformationmightalsobeneededtodoone'sjob. Adashboardfitsonasinglecomputerscreen.Theinformationmustfitonasinglescreen,entirely availablewithintheviewer'seyespansoitcanallbeseenatonce,ataglance.Ifyoumustscroll aroundtoseealltheinformation,ithastransgressedtheboundariesofadashboard.Ifyoumust shiftfromscreentoscreentoseeitall,you'vemadeuseofmultipledashboards.Theobjectisto havethemostimportantinformationreadilyandeffortlesslyavailablesoyoucanquicklyabsorb whatyouneedtoknow. Musttheinformationbedisplayedinawebbrowser?Thatmightbethebestmediumformost dashboardstoday,butitisn'ttheonlyacceptablemedium,anditmightnotbethebestmedium10 yearsfromnow.Musttheinformationbeconstantlyrefreshedinrealtime?Onlyiftheobjectives thatitservesrequirerealtimeinformation.Ifyouaremonitoringairtrafficusingadashboard,you mustimmediatelybeinformedwhensomethingiswrong.Ontheotherhand,ifyouaremaking strategicdecisionsabouthowtoboostsales,asnapshotofinformationasoflastnight,orperhaps eventheendoflastmonth,shouldworkfine. Dashboardsareusedtomonitorinformationataglance.Despitethefactthatinformationabout almostanythingcanbeappropriatelydisplayedinadashboard,thereisatleastonecharacteristic thatdescribesalmostalltheinformationfoundindashboards:itisabbreviatedintheformof summariesorexceptions.Thisisbecauseyoucannotmonitorataglanceallthedetailsneededto achieveyourobjectives.Adashboardmustbeabletoquicklypointoutthatsomethingdeserves yourattentionandmightrequireaction.Itneedn'tprovideallthedetailsnecessarytotakeaction, butifitdoesn't,itoughttomakeitaseasyandseamlessaspossibletogettothatinformation. Gettingtheremightinvolveshiftingtoadifferentdisplaybeyondthedashboard,usingnavigational methodssuchasdrillingdown.Thedashboarddoesitsprimaryjobifittellsyouwithnomorethan aglancethatyoushouldact.Itservesyousuperblyifitdirectlyopensthedoortoanyadditional informationthatyouneedtotakethataction.

That'stheessenceofthedashboard.Nowlet'saddtothisdefinitionacouplemoresupportingattributes thathelpdashboardsdotheirjobeffectively: Dashboardshavesmall,concise,clear,andintuitivedisplaymechanisms.Displaymechanismsthat clearlystatetheirmessagewithouttakingupmuchspacearerequired,sothattheentirecollection ofinformationwillfitintothelimitedrealestateofasinglescreen.Ifsomethingthatlookslikea fuelgauge,trafficsignal,orthermometerfitsthisrequirementbestforaparticularpieceof information,that'swhatyoushoulduse,butifsomethingelseworksbetter,youshouldusethat instead.Insistingonsexydisplayssimilartothosefoundinacarwhenothermechanismswould workbetteriscounterproductive. Dashboardsarecustomized.Theinformationonadashboardmustbetailoredspecificallytothe requirementsofagivenperson,group,orfunction;otherwise,itwon'tserveitspurpose.

Adashboardisatypeofdisplay,aformofpresentation,notaspecifictypeofinformationortechnology. Keepthisdistinctionclear,andyouwillbefreedtofocusonwhatreallymatters:designingdashboardsto communicate.

www.it-ebooks.info

1.4.ATimelyOpportunity
Severalcircumstanceshaverecentlycombinedtocreateatimelyopportunityfordashboardstoaddvalue totheworkplace,includingtechnologiessuchashighresolutiongraphics,emphasisonperformance managementandmetrics,andagrowingrecognitionofvisualperceptionasapowerfulchannelfor informationacquisitionandcomprehension.Dashboardsofferauniquesolutiontotheproblemof informationoverloadnotacompletesolutionbyanymeans,butonethathelpsalot.AsDr.Hoviswrotein thatsamearticleinDMDirect: Therealvalueofdashboardproductsliesintheirabilitytoreplacehuntandpeck datagatheringtechniqueswithatireless,adaptable,informationflowmechanism. Dashboardstransformdatarepositoriesintoconsumableinformation.1 Dashboardsaren'tallthatdifferentfromsomeoftheothermeansofpresentinginformation,butwhen properlydesignedthesinglescreendisplayofintegratedandfinelytuneddatacandeliverinsightinan especiallypowerfulway. Dashboardsandvisualizationarecognitivetoolsthatimproveyour"spanofcontrol" overalotofbusinessdata.Thesetoolshelppeoplevisuallyidentifytrends,patterns andanomalies,reasonaboutwhattheyseeandhelpguidethemtowardeffective decisions.Assuch,thesetoolsneedtoleveragepeople'svisualcapabilities.Withthe prevalenceofscorecards,dashboardsandothervisualizationtoolsnowwidely availableforbusinessuserstoreviewtheirdata,theissueofvisualinformation designismoreimportantthanever.2 ThefinalsentimentthatBrathandPetersexpressedinthisexcerptfromtheirarticleunderscoresthe purposeofthisbook.Asdatavisualizationbecomesincreasinglycommonasameansofbusiness communication,itisimperativethatexpertiseindatavisualizationbeacquired.Thisexpertisemustbe groundedinanunderstandingofvisualperception,andofhowthisunderstandingcanbeeffectively appliedtothevisualdisplayofdatawhatworks,whatdoesn't,andwhy.Theseskillsarerarelyfoundinthe businessworld,notbecausetheyaredifficulttolearn,butbecausetheneedtolearnthemisseldom recognized.Thisistrueingeneral,andespeciallywithregardtodashboards.Thechallengeofpresentinga largeassortmentofdataonasinglescreeninawaythatproducesimmediateinsightisbynomeanstrivial. Buckleup;you'reinforafunride.

1 2

GregoryL.Hovis,"StopSearchingforInformationMonitoritwithDashboardTechnology,"DMDirect,February2002 RichardBrathandMichaelPeters,"DashboardDesign:WhyDesignisImportant,"DMDirect,October2004

www.it-ebooks.info

Chapter2.VariationsinDashboardUsesandData
Dashboardscanbeusedtomonitormanytypesofdataandtosupportalmostanysetofobjectivesbusiness deemsimportant.Therearemanywaystocategorizedashboardsintovarioustypes.Thewaythatrelates mostdirectlytoadashboard'svisualdesigninvolvestheroleitplays,whetherstrategic,analytical,or operational.Thedesigncharacteristicsofthedashboardcanbetailoredtoeffectivelysupporttheneedsof eachoftheseroles.Whilecertaindifferencessuchasthesewillaffectdesign,therearealsomany commonalitiesthatspanalldashboardsandinviteastandardsetofdesignpractices.

www.it-ebooks.info

Categorizingdashboards Commonthreadsindashboarddata Nonquantitativedashboarddata Dashboardsareusedtosupportabroadspectrumofinformationneeds,spanningtheentirerangeof businesseffortsthatmightbenefitfromanimmediateoverviewofwhat'sgoingon.Dashboardscanbe tailoredtospecificpurposes,andasingleindividualmightbenefitfrommultipledashboards,each supportingadifferentaspectofthatperson'swork.Thevariousdataandpurposesthatdashboardscanbe usedtosupportareworthdistinguishing,fortheysometimesdemanddifferencesinvisualdesignand functionality.

2.1.CategorizingDashboards
Dashboardscanbecategorizedinseveralways.Nomatterhowlimitedandflawedtheeffort,doingsois usefulbecauseithelpsustoexaminethebenefitsandmanyusesofthemedium.I'moneofthosepeople whoenjoystheprocessofclassifyingthings,breakingthemupintogroups.It'sanintellectualexercisethat forcesmetodigbeneaththesurface.Idon't,however,assignundueworthtoanyonewayofcategorizing something,andIcertainlydon'teverwanttogiveintothearroganceofclaimingthatmineistheonlyway. Taxonomiesascientifictermforsystemsofclassificationarealwaysbasedononeormorevariables(thatis, categoriesconsistingofmultiplepotentialvalues).Forinstance,basedonthevariable"platform,"a dashboardtaxonomycouldconsistofthosethatruninclient/servermodeandthosethatruninweb browsers.Thefollowingtablelistsseveralvariablesthatcanbeusedtostructuredashboardtaxonomies, alongwithpotentialvaluesforeach.Thislistcertainlyisn'tcomprehensive;thesearesimplymyattempts toexpressthevarietyandexplorethepotentialofthedashboardmedium. Table21. Variable Role Values Strategic Analytical Operational Typeofdata Quantitative Nonquantitative Datadomain Sales Finance Marketing Manufacturing

www.it-ebooks.info

HumanResources Typeofmeasures BalancedScorecard(forexample,KPIs) SixSigma Nonperformance Spanofdata Enterprisewide Departmental Individual Updatefrequency Monthly Weekly Daily Hourly Realtimeornearrealtime Interactivity Staticdisplay Interactivedisplay(drilldown,filters,etc.) Mechanismsofdisplay Primarilygraphical Primarilytext Integrationofgraphicsandtext Portalfunctionality Conduittoadditionaldata Noportalfunctionality 2.1.1.ClassifyingDashboardsbyRole Perhapsoneofthemostusefulwaystocategorizeadashboard,andtheonethatI'llfocuson,isbyits rolethetypeofbusinessactivitythatitsupports.Mybreakdownofdashboardsintothreeroles(strategic, analytical,andoperational)iscertainlynottheonlywaytoexpressthetypesofbusinessactivitiesa dashboardcansupport.However,thisistheonlyclassificationthatsignificantlyrelatestodifferencesin visualdesign. 2.1.1.1.Dashboardsforstrategicpurposes Theprimaryuseofdashboardstodayisforstrategicpurposes.Thepopular"executivedashboard,"and mostofthedashboardsthatsupportmanagersatanylevelinanorganization,arestrategicinnature.They providethequickoverviewthatdecisionmakersneedtomonitorthehealthandopportunitiesofthe business.Dashboardsofthistypefocusonhighlevelmeasuresofperformance,includingforecaststolight thepathintothefuture.Althoughthesemeasurescanbenefitfromcontextualinformationtoclarifythe

www.it-ebooks.info

meaning,suchascomparisonstotargetsandbriefhistories,alongwithsimpleevaluatorsofperformance (forexample,goodandbad),toomuchinformationofthistypeortoomanysubtlegradationscandistract fromtheprimaryandimmediategoalsofthestrategicdecisionmaker. Extremelysimpledisplaymechanismsworkbestforthistypeofdashboard.Giventhegoaloflongterm strategicdirection,ratherthanimmediatereactionstofastpacedchanges,thesedashboardsdon'trequire realtimedata;rather,theybenefitfromstaticsnapshotstakenmonthly,weekly,ordaily.Lastly,theyare usuallyunidirectionaldisplaysthatsimplypresentwhatisgoingon.Theyarenotdesignedforthe interactionthatmightbeneededtosupportfurtheranalysis,becausethisisrarelythedirectresponsibility ofthestrategicmanager.You'llbeluckyifyoucangetanexecutivetoviewtheinformationonacomputer screenratherthanapieceofpaper,letalonedealwiththenavigationaldemandsofinteractiveonline analysis. 2.1.1.2.Dashboardsforanalyticalpurposes Dashboardsthatsupportdataanalysisrequireadifferentdesignapproach.Inthesecasestheinformation oftendemandsgreatercontext,suchasrichcomparisons,moreextensivehistory,andsubtlerperformance evaluators.Likestrategicdashboards,analyticaldashboardsalsobenefitfromstaticsnapshotsofdatathat arenotconstantlychangingfromonemomenttothenext.However,moresophisticateddisplaymediaare oftenusefulfortheanalystwhomustexaminecomplexdataandrelationshipsandiswillingtoinvestthe timeneededtolearnhowtheywork.Analyticaldashboardsshouldsupportinteractionswiththedata,such asdrillingdownintotheunderlyingdetails,toenabletheexplorationneededtomakesenseofitthatis,not justtoseewhatisgoingonbuttoexaminethecauses.Forexample,itisn'tenoughtoseethatsalesare decreasing;whenyourpurposeisanalysis,youmustbemadeawareofsuchpatternssothatyoucanthen explorethemtodiscoverwhatiscausingthedecreaseandhowitmightbecorrected.Thedashboarditself, asamonitoringdevicethattellstheanalystwhattoinvestigate,neednotsupportallthesubsequent interactionsdirectly,butitshouldlinkasseamlesslyaspossibletothemeanstoanalyzethedata. 2.1.1.3.Dashboardsforoperationalpurposes Whendashboardsareusedtomonitoroperations,theymustbedesigneddifferentlyfromthosethat supportstrategicdecisionmakingordataanalysis.Thecharacteristicofoperationsthatuniquelyinfluences thedesignofdashboardsmostistheirdynamicandimmediatenature.Whenyoumonitoroperations,you mustmaintainawarenessofactivitiesandeventsthatareconstantlychangingandmightrequireattention andresponseatamoment'snotice.Iftheroboticarmonthemanufacturingassemblylinethatattaches thecardoortothechassisrunsoutofbolts,youcan'twaituntilthenextdaytobecomeawareofthe problemandtakeaction.Likewise,iftrafficonyourwebsitesuddenlydropstohalfitsnormallevel,you wanttobenotifiedimmediately. Aswithstrategicdashboards,thedisplaymediaonoperationaldashboardsmustbeverysimple.Inthe stressfuleventofanemergencythatrequiresanimmediateresponse,themeaningofthesituationandthe appropriateresponsesmustbeextremelyclearandsimple,ormistakeswillbemade.Incontrastto strategicdashboards,operationaldashboardsmusthavethemeanstograbyourattentionimmediatelyif anoperationfallsoutsidetheacceptablethresholdofperformance.Also,theinformationthatappearson operationaldashboardsisoftenmorespecific,providingadeeperlevelofdetail.Ifacriticalshipmentisat riskofmissingitsdeadline,ahighlevelstatisticwon'tdo;youneedtoknowtheordernumber,who's handlingit,andwhereitisinthewarehouse.Detailslikethesemightappearautomaticallyonan operationaldashboard,ortheymightbeaccessedbydrillingdownonorhoveringthemouseoverhigher leveldata,sointeractivityisoftenuseful.

www.it-ebooks.info

Thewaysthatdashboarddesignmusttakedifferentformsinresponsetodifferentrolesareclearlyworth yourattention.We'llexaminesomeofthesedifferencesinmoredetailinChapter8,PuttingItAllTogether, whenwereviewseveralexamplesofwhatworksandwhatdoesn'tforvariouspurposes.

2.2.TypicalDashboardData
Dashboardsareusefulforallkindsofwork.Whetheryou'reameteorologistmonitoringtheweather,an intelligenceanalystmonitoringpotentialterroristchatter,aCEOmonitoringthehealthandopportunitiesof amultibilliondollarcorporation,orafinancialanalystmonitoringthestockmarket,awelldesigned dashboardcouldserveyouwell. 2.2.1.TheCommonThreadinDashboardDiversity Despitethesediverseapplications,inalmostallcasesdashboardsprimarilydisplayquantitativemeasures ofwhat'scurrentlygoingon.Thistypeofdataiscommonacrossalmostalldashboardsbecausetheyare usedtomonitorthecriticalinformationneededtodoajobormeetoneormoreparticularobjectives,and most(butnotall,aswe'llseelater)oftheinformationthatdoesthisbestisquantitative. Thefollowingtablelistsseveralmeasuresof"what'scurrentlygoingon"thataretypicalinbusiness. Table22. Category Sales Measures Bookings Billings Salespipeline(anticipatedsales) Numberoforders Orderamounts Sellingprices Marketing Marketshare Campaignsuccess Customerdemographics Finance Revenues Expenses Profits TechnicalSupport Numberofsupportcalls Resolvedcases Customersatisfaction

www.it-ebooks.info

Calldurations Fulfillment Numberofdaystoship Backlog Inventorylevels Manufacturing Numberofunitsmanufactured Manufacturingtimes Numberofdefects HumanResources Employeesatisfaction Employeeturnover Countofopenpositions Countoflateperformancereviews InformationTechnology Networkdowntime Systemusage Fixedapplicationbugs WebServices Numberofvisitors Numberofpagehits Visitdurations Thesemeasuresareoftenexpressedinsummaryform,mostoftenastotals,slightlylessoftenasaverages (suchasaveragesellingprice),occasionallyasmeasuresofdistribution(suchasastandarddeviation),and rarerstillasmeasuresofcorrelation(suchasalinearcorrelationcoefficient).Summaryexpressionsof quantitativedataareparticularlyusefulindashboards,whereitisnecessarytomonitoranarrayof businessphenomenaataglance.Obviously,thelimitedrealestateofasinglescreenrequiresconcise communication. 2.2.1.1.Variationsintiming Measuresofwhat'scurrentlygoingoncanbeexpressedinavarietyoftimeframes.Afewtypicalexamples include:

Thisyeartodate Thisweektodate Thisquartertodate Yesterday

www.it-ebooks.info

Thismonthtodate Todaysofar

Theappropriatetimeframeisdeterminedbythenatureoftheobjectivesthatthedashboardsupports. 2.2.1.2.Enrichmentthroughcomparison Thesemeasurescanbedisplayedbythemselves,butitisusuallyhelpfultocomparethemtooneormore relatedmeasurestoprovidecontextandtherebyenrichtheirmeaning.Hereareperhapsthemosttypical comparativemeasures,andanexampleofeach. Table23. Comparativemeasure Thesamemeasureatthesamepointintimeinthepast Example Thesamedaylastyear

Thesamemeasureatsomeotherpointintimeinthepast Theendoflastyear Thecurrenttargetforthemeasure Relationshiptoafuturetarget Apriorpredictionofthemeasure Relationshiptoafuturepredictionofthemeasure Somemeasureofthenormforthismeasure Abudgetedamountforthecurrentperiod Percentageofthisyear'sbudgetsofar Forecastofwhereweexpectedtobetoday Percentageofthisquarter'sforecast Average,normalrange,orabenchmark,such asthenumberofdaysitnormallytakesto shipanorder

Anextrapolationofthecurrenttomeasureintheformof Projectionoutintothefuture,suchasthe aprobablefuture,eitherataspecificpointinthefuture comingyearend orasatimeseries Someoneelse'sversionsofthesamemeasure Aseparatebutrelatedmeasure Thesecomparisonsareoftenexpressedgraphicallytoclearlycommunicatethedifferencesbetweenthe values,whichmightnotleapoutasdramaticallythroughtheuseoftextalone.However,textaloneisoften adequate.Forexample,whenonlythecomparisonitselfisrequiredandtheindividualmeasures(aprimary measureandacomparativemeasure)aren'tnecessary,asinglenumberexpressedasapercentagecanbe used(suchas119%ofbudgetor7%ofwherewewerethistimelastyear). Measuresofwhat'scurrentlygoingonmaybedisplayedeitherasasinglemeasure,asasinglemeasure combinedwithoneormoreindividualcomparativemeasures,orasoneofthefollowing: Acompetitor'smeasure,suchasrevenues Ordercountcomparedtoorderrevenue

www.it-ebooks.info

Multipleinstancesofameasure,eachrepresentingacategoricalsubdivisionofthemeasure(for example,salessubdividedintoregionsoracountoforderssubdividedintonumericrangesinthe formofafrequencydistribution) Temporalinstancesofameasure(thatis,atimeseries,suchasmonthlyinstancesofthemeasure)

Timeseriesinparticularproviderichcontextforunderstandingwhat'sreallygoingonandhowwellit's going. 2.2.1.3.Enrichmentthroughevaluation Becausewithadashboardagreatdealofdatamustbeevaluatedquickly,italsoisquiteusefultoexplicitly declarewhethersomethingisgoodorbad.Suchevaluativeinformationisoftenencodedasspecialvisual objects(forexample,atrafficlight)orasvisualattributes(forexample,bydisplayingthemeasureinbright redtoindicateaseriouscondition).Whendesignedproperly,simplevisualindicatorscanclearlyalertusers tothestateofparticularmeasureswithoutalteringtheoveralldesignofthedashboard.Evaluative indicatorsneednotbelimitedtobinarydistinctionsbetweengoodandbad,butiftheyexceedthelimitof morethanafewdistinctstates(forexample,verybad,bad,acceptable,good,andverygood),theyrunthe riskofbecomingtoocomplexforefficientperception. 2.2.2.NonQuantitativeDashboardData ManypeoplethinkofdashboardsandKPIsasnearlysynonymous.Itiscertainlytruethatdashboardsarea powerfulmediumforpresentingKPIs,butnotallquantitativeinformationthatmightbeusefulona dashboardbelongstothelistofdefinedKPIs.Infact,notallinformationthatisusefulondashboardsis evenquantitativethecriticalinformationneededtodoajobcannotalwaysbeexpressednumerically. Althoughmostinformationthattypicallyfindsitswayontoadashboardisquantitative,sometypesofnon quantitativedata,suchassimplelists,arefairlycommonaswell.Hereareafewexamples:

Top10customers Issuesthatneedtobeinvestigated Tasksthatneedtobecompleted Peoplewhoneedtobecontacted

Anothertypeofnonquantitativedataoccasionallyfoundondashboardsrelatestoschedules,including tasks,duedates,thepeopleresponsible,andsoon.Thisiscommonwhenthejobthatthedashboard supportsinvolvesthemanagementofprojectsorprocesses. Ararertypeinvolvesthedisplayofentitiesandtheirrelationships.Entitiescanbestepsorstagesina process,peopleororganizationsthatinteractwithoneanother,oreventsthataffectoneanother,toname afewcommonexamples.Thistypeofdisplayusuallyencodesentitiesascirclesorrectanglesand relationshipsaslines,oftenwitharrowsatoneorbothendstoindicatedirectionorinfluence.Itisoften usefultointegratequantitativeinformationthatisassociatedwiththeentitiesandrelationships,suchas theamountoftimethatpassedbetweeneventsinaprocess(forexample,byassociatinganumberwith thelinethatlinkstheeventsorbyhavingthelengthofthelineitselfencodetheduration)orthesizesof businessentities(perhapsexpressedinrevenuesornumberofemployees).

www.it-ebooks.info

Nowthatyouknowabitabouthowandwhydashboardsareused,it'stimetotakeacloserlookatsome designprinciples.Inthenextchapter,we'lldelveintosomeofthemistakesthatarecommonlymadein dashboarddesign.

www.it-ebooks.info

Chapter3.ThirteenCommonMistakesinDashboardDesign
Preoccupationwithsuperficialandfunctionallydistractingvisualcharacteristicsofdashboardshasledtoa rashofvisualdesignproblemsthatunderminetheirusefulness.Thirteenvisualdesignproblemsare frequentlyfoundindashboards,includingintheexamplesfeaturedasexemplarybysoftwarevendors.

www.it-ebooks.info

Exceedingtheboundariesofasinglescreen Supplyinginadequatecontextforthedata Displayingexcessivedetailorprecision Choosingadeficientmeasure Choosinginappropriatedisplaymedia Introducingmeaninglessvariety Usingpoorlydesigneddisplaymedia Encodingquantitativedatainaccurately Arrangingthedatapoorly Highlightingimportantdataineffectivelyornotatall Clutteringthedisplaywithuselessdecoration Misusingoroverusingcolor Designinganunattractivevisualdisplay Thefundamentalchallengeofdashboarddesignistheneedtosqueezeagreatdealofinformationintoa smallamountofspace,resultinginadisplaythatiseasilyandimmediatelyunderstandable.Ifthisdoesn't soundchallenging,eitheryouareanexpertdesignerwithextensivedashboardexperience,oryouare baskingintheglowofnaivet.Attemptthetask,andyouwillfindthatdashboardsposeauniquedata visualizationchallenge.Anddon'tassumethatyoucanlooktoyoursoftwarevendorforhelpiftheyhave thenecessarydesigntalent,they'redoingagreatjobofhidingit. Sadly,itiseasytofindmanyexamplesofthemistakesyoushouldavoidbylookingnofurtherthantheweb sitesofthesoftwarevendorsthemselves.Let'susesomeoftheseexamplestoexaminedesignthatdoesn't workandlearnwhyitdoesn't. Note:Inalmosteverycase,I'vechosentouseactualexamplesfromvendorwebsitestoillustrate dashboarddesignmistakes.Indoingso,Iamnotsayingthatthesoftwarethatproducedtheexampleis badI'mnotcommentingonthequalityofthesoftwareonewayoranother.WhatIamsayingisthatthe designpracticeisbad.Thisresultsprimarilyfromvendors'lackofexpertiseinorinattentiontovisual design.Thesevendorsshouldknowbetter,butthey'vechosentofocustheirenergiesonotheraspectsof theirproducts,oftenhighlightingglitzyvisualfeaturesthatactuallyundermineeffectivecommunication.I hopethatseeingtheirworkusedtoillustratepoordashboarddesignwillserveasawakeupcalltostart payingattentiontothefeaturesthatreallymatter.

3.1.ExceedingtheBoundariesofaSingleScreen
Myinsistencethatadashboardshouldconfineitsdisplaytoasinglescreen,withnoneedforscrollingor switchingbetweenmultiplescreens,mightseemarbitraryandabitfinicky,butitisbasedonsolidand practicalrationale.Afterstudyingdatavisualizationforawhile,includingvisualperception,onediscovers thatsomethingpowerfulhappenswhenthingsareseentogether,allwithineyespan.Likewise,something

www.it-ebooks.info

criticalislostwhenyoulosesightofsomedatabyscrollingorswitchingtoanotherscreentoseeother data.Partoftheproblemisthatwecanholdonlyafewchunksofinformationatatimeinshortterm memory.Relyingonthemind'seyetorememberinformationthatisnolongervisibleisarockyventure. Oneofthegreatbenefitsofadashboardasamediumofcommunicationisthesimultaneityofvisionthatit offers:theabilitytoseeeverythingthatyouneedatonce.Thisenablescomparisonsthatleadto insightsthose"Aha!"experiencesthatmightnotoccurinanyotherway.Clearly,exceedingtheboundaries ofasinglescreennegatesthisbenefit.Let'sexaminethetwoversionsofthisproblemfragmentingdatainto separatescreensandrequiringscrollingindependently. 3.1.1.FragmentingDataintoSeparateScreens Informationthatappearsondashboardsisoftenfragmentedinoneoftwoways:

Separatedintodiscretescreenstowhichonemustnavigate Separatedintodifferentinstancesofasinglescreenthatareaccessedthroughsomeformof interaction

Enablinguserstonavigatetodiscretescreensordifferentinstancesofasinglescreentoaccessadditional informationisnot,ingeneral,abadpractice.Allowingnavigationtofurtherdetailortoadifferentsetof informationthatachievesitspurposebestbystandingalonecanbeapowerfuldashboardfeature. However,whenalltheinformationshouldbeseenatthesametimetogainthedesiredinsights,that fragmentationunderminestheuniqueadvantagesofadashboard.Fragmentingdatathatshouldbeseen togetherisamistake. Let'slookatanexample.ThedashboardinFigure31fragmentsthedatathatexecutivesneedinto10 separatedashboards.Thiswouldbefineiftheexecutiveswouldn'tbenefitfromseeingthesevarious measurestogether,butthatishardlythecase.

Figure31.Thisdashboardfragmentsthedatainawaythatunderminestheviewer'sabilitytoseemeaningfulrelationships.

www.it-ebooks.info

Inthisexample,abankingexecutiveisforcedtoexaminetheperformanceofthefollowingaspectsofthe businessseparately: Highlights Deposits Pastdueloans Profitability Growth Loans Risk Depositmix Channels Marketshare

Eachofthesescreenspresentsaseparate,highlevelsnapshotofasinglesetofmeasuresthatoughttobe integratedintoasinglescreen.Despitewhatyoumightassumeabouttheavailablescreenlabeled "Highlights,"itdoesnotprovideaconsolidatedvisualoverviewofthedatabutconsistsprimarilyofatext tablethatcontainsseveralofthemeasures.Abankingexecutiveneedstoseethesemeasurestogetherina waythatenablescomparisonstounderstandhowtheyrelatetoandinfluenceoneanother. Splittingthebigpictureintoaseriesofseparatesmallpicturesisamistakewheneverseeingthebigpicture isworthwhile. Asimilarexample,fromthesamesoftwarevendor,isshowninFigure32.Thistimethepictureofdaily saleshasbeensplitintoaseparatedashboardforeachof20products.Iftheintentionistoservetheneeds ofproductmanagerswhoareeachexclusivelyinterestedinasingleproductandneverwanttocompare salesofthatproducttoothers,thisdesigndoesn'tfragmentthedatainaharmfulway.If,however,any benefitcanbegainedbyviewingthesalesofmultipleproductstogether,whichisalmostsurelythecase, thisdesignfails.

www.it-ebooks.info

Figure32.Thisdashboardrequiresviewerstoclickonadesiredproductandviewinformationforonlyoneproductatatime.

3.1.2.RequiringScrolling ThedashboardinFigure33illustratestheproblemthat'screatedwhenscrollingisrequiredtoseeallthe data.Notonlyareweleftwonderingwhatliesbelowthebottomofthescreeninthedashboardasa whole,butwe'realsogivenimmediatevisualaccessonlytothefirstofmanymetricsthatappearinthe scrollableboxatthetopright,beginningwith"No.Transactions."We'dbebetteroffreadingaprinted reportextendingacrossmultiplepages,becauseatleastthenwecouldlayoutallofthepagesatoncefor simultaneousviewing.Peoplecommonlyassumethatanythingthatliesbeyondtheirimmediatefieldof visionandrequiresscrollingtoseeisoflessimportancethanwhat'simmediatelyvisible.Manyviewers won'tbothertolookatwhatliesoffthescreen,andthosewhotakethetimewilllikelyresenttheeffort.

www.it-ebooks.info

Figure33.Thisdashboarddemonstratestheeffectivenessthatissacrificedwhenscrollingisrequiredtoseealltheinformation.

3.2.SupplyingInadequateContextfortheData
Measuresofwhat'scurrentlygoingoninthebusinessrarelydowellasasoloact;theyneedagood supportingcasttosucceed.Forexample,tostatethatquartertodatesalestotal$736,502withoutany contextmeanslittle.Comparedtowhat?Isthisgoodorbad?Howgoodorbad?Areweontrack?Arewe doingbetterthanwehaveinthepast,orworsethanwe'veforecasted?Supplyingtherightcontextforkey measuresmakesthedifferencebetweennumbersthatjustsitthereonthescreenandthosethatenlighten andinspireaction. ThegaugesinFigure34couldeasilyhaveincorporatedusefulcontext,buttheyfallshortoftheirpotential. Forinstance,thecentergaugetellsusonlythat7,822unitshavesoldthisyeartodate,andthatthis numberisgood(indicatedbythegreenarrow).Aquantitativescaleonagraph,suchastheradialscalesof tickmarksonthesegauges,ismeanttoprovideanapproximationofthemeasure,butitcanonlydosoif thescaleislabeledwithnumbers,whichthesegaugeslack.Ifthenumbershadbeenpresent,thepositions ofthearrowsmighthavebeenmeaningful,butherethepresenceofthetickmarksalongaradialaxis suggestsusefulinformationthathasn'tactuallybeenincluded.

Figure34.Thesedashboardgaugesfailtoprovideadequatecontexttomakethemeasuresmeaningful.

www.it-ebooks.info

Thesegaugesuseupagreatdealofspacetotellusnothingwhatsoever.Thesameinformationcouldhave beencommunicatedsimplyastextinmuchlessspace,withoutanylossofmeaning: Table31. YTDUnits OctoberUnits ReturnsRate Anotherfailureofthesegaugesisthattheyteaseusbycoloringthearrowstoindicategoodorbad performance,withouttellingushowgoodorbaditis.Theycouldeasilyhavedonethisbylabelingthe quantitativescalesandvisuallyencodingsectionsalongthescalesasgoodorbad,ratherthanjustencoding thearrowsinthismanner.Hadthisbeendone,wewouldbeabletoseeataglancehowgoodorbada measureisbyhowfarthearrowpointsintothegoodorbadranges. ThegaugethatappearsinFigure35doesabetterjobofincorporatingcontextintheformofmeaningful comparisons.Here,thepotentialofthegraphicaldisplayismorefullyrealized.Thegaugemeasuresthe averagedurationofphonecallsandispartofalargerdashboardofcallcenterdata. Supplyingcontextformeasuresneednotalwaysinvolveachoiceofthesinglebestcomparisonrather, severalcontextsmaybegiven.Forinstance,quartertodatesalesof$736,502mightbenefitfrom comparisonstothebudgettargetof$1,000,000;salesonthisdaylastyearof$856,923;andatimeseries ofsalesfiguresforthelastsixquarters.Suchadisplaywouldprovidemuchricherinsightthanasimple displayofthecurrentsalesfigure,withorwithoutanindicationofwhetherit's"good"or"bad."Youmust becareful,however,whenincorporatingrichcontextsuchasthistodosoinawaythatdoesn'tforcethe viewertogetboggeddowninreadingthedetailstogetthebasicmessage.Itisusefultoprovideavisually prominentdisplayoftheprimaryinformationandtosubduethesupportingcontextsomewhat,sothatit doesn'tgetinthewaywhenthedashboardisbeingquicklyscannedforkeypoints. 7,822 869 0.26%

Figure35.Thisdashboardgauge(foundinapaperentitled"MakingDashboardsActionable,"writtenbyLaurieM.Orlovand publishedinDecember2003byForresterResearch,Inc.)doesabetterjobthanthoseinFigure34ofusingagaugeeffectively.

www.it-ebooks.info

Theamountofcontextthatoughttobeincorporatedtoenrichthemeasuresonadashboarddependson itspurposeandtheneedsofitsviewers.Moreisnotalwaysbetter,butwhenmoreprovidesrealvalue,it oughttobeincludedinawaythatsupportsbothaquickoverviewwithoutdistractionaswellascontextual informationforricherunderstanding.1

3.3.DisplayingExcessiveDetailorPrecision
Dashboardsalmostalwaysrequirefairlyhighlevelinformationtosupporttheviewer'sneedforaquick overview.Toomuchdetail,ormeasuresthatareexpressedtooprecisely(forexample,$3,848,305.93 ratherthan$3,848,305,orperhapseven$3.8M),justslowviewersdownwithoutprovidingthemany benefit.Inaway,thisproblemistheoppositeextremeoftheoneweexaminedintheprevioussectiontoo muchinformationratherthantoolittle. ThedashboardinFigure36illustratesthistypeofexcess.ExaminethetwosectionsthatI'veenclosedin redrectangles.Thelowerrightsectiondisplaysfrom4to10decimaldigitsforeachmeasure,whichmight beusefulinsomecontexts,butdoubtfullyinadashboard.Thehighlightedsectionabovedisplaystime downtothelevelofseconds,whichalsoseemslikeoverkillinthiscontext.Withadashboard,every unnecessarypieceofinformationresultsintimewastedtryingtofilteroutwhat'simportant,whichis intolerablewhentimeisoftheessence.

Figure36.Thisdashboardshowsunnecessarydetail,suchastimesexpressedtothesecondandmeasuresexpressedto10 decimalplaces.

Ibelievethatthecircularshapeusedbygaugeslikethisonewastesvaluablespaceonadashboard,asI'llexplainin Chapter6,EffectiveDashboardDisplayMedia.Nevertheless,Icommendthisgaugefordisplayingricherinformation thanmost.


1

www.it-ebooks.info

3.4.ChoosingaDeficientMeasure
Forameasuretobemeaningful,wemustknowwhatisbeingmeasuredandtheunitsinwhichthemeasure isbeingexpressed.Ameasureisdeficientifitisn'ttheonethatmostclearlyandefficientlycommunicates themeaningthatthedashboardviewershoulddiscern.Itcanbeaccurate,yetnotthebestchoiceforthe intendedmessage.Forexample,ifthedashboardvieweronlyneedstoknowtowhatdegreeactual revenuediffersfrombudgetedrevenue,itwouldbemoredirecttosimplyexpressthevarianceas9%(and perhapsdisplaythevarianceof$8,066aswell)ratherthandisplayingtheactualrevenueamountof$76,934 andthebudgetedrevenueamountof$85,000andleavingittotheviewertocalculatethedifference.In thiscase,apercentageclearlyfocusesattentiononthevarianceinamannerthatisdirectlyintelligible. Figure37illustratesthispoint.Whilethisgraphdisplaysactualandbudgetedrevenuesseparately,its purposeistocommunicatethevarianceofactualrevenuesfromthebudget.

Figure37.Thisgraphillustratestheuseofmeasuresthatfailtodirectlyexpresstheintendedmessage.

Thevariance,however,couldhavebeendisplayedmorevividlybyencodingbudgetedrevenueasa referencelineof0%andthevarianceasalinethatmeandersaboveandbelowbudget(expressedinunits ofpositiveandnegativepercentages,asshownonthenextpageinFigure38).Thepointhereistoalways thinkcarefullyaboutthemessagethatmostdirectlysupportstheviewer'sneeds,andthenselectthe measurethatmostdirectlysupportsthatmessage.

www.it-ebooks.info

Figure38.Thisgraphisdesignedtoemphasizedeviationfromatarget,whichitaccomplishesinpartbyexpressingthe differencebetweenbudgetedandactualrevenuesusingpercentages.

3.5.ChoosingInappropriateDisplayMedia
Choosinginappropriatedisplaymediaisoneofthemostcommondesignmistakesmade,notjustin dashboards,butinallformsofquantitativedatapresentation.Forinstance,usingagraphwhenatableof numberswouldworkbetter,andviceversa,isafrequentmistake.Allowmetoillustrateusingseveral examplesbeginningwiththepiechartinFigure39.

Figure39.Thischartillustratesacommonproblemwithpiecharts.

Thispiechartispartofadashboardthatdisplaysbreastcancerstatistics.Lookatitforamomentandseeif anythingseemsodd. Piechartsaredesignedspecificallytopresentpartsofawhole,andthewholeshouldalwaysaddupto 100%.Here,theslicelabeled"Breast13.30%"lookslikeitrepresentsaround40%ofthepieafarcryfrom 13.3%.Despitethemeaningthatapiechartsuggests,theseslicesarenotpartsofawhole;theyrepresent theprobabilitythatawomanwilldevelopaparticularformofcancer(breast,lung,colon,andsixtypesthat aren'tlabeled).Thismisuseofapiechartinvitesconfusion. Thetruthis,Ineverrecommendtheuseofpiecharts.Theonlythingtheyhavegoingforthemisthefact thateverybodyimmediatelyknowswhentheyseeapiechartthattheyareseeingpartsofawhole(or

www.it-ebooks.info

oughttobe).Beyondthat,piechartsdon'tdisplayquantitativedataveryeffectively.Asyou'llseein Chapter4,TappingintothePowerofVisualPerception,humanscan'tcomparetwodimensionalareasor anglesveryaccuratelyandthesearethetwomeansthatpiechartsusetoencodequantitativedata.Bar graphsareamuchbetterwaytodisplaythisinformation.1 ThepiechartinFigure310showsthatevenwhencorrectlyusedtopresentpartsofawhole,thesegraphs don'tworkverywell.Withoutthevaluelabels,youwouldonlybeabletodiscernthatopportunitiesrated as"Fair"representthelargestgroup,thoseratedas"FieldSales:2VeryHigh"representaminisculegroup, andtheotherratingsgroupsareroughlyequalinsize.

Figure310.Thisexampleshowsthatevenwhentheyareusedcorrectlytopresentpartsofawhole,piechartsaredifficultto interpretaccurately.

Figure311displaysthesamedataasFigure310,thistimeusingahorizontalbargraphthatcanbe interpretedmuchmoreefficientlyandaccurately.

Figure311.Thishorizontalbargraphdoesamuchbetterjobofdisplayingparttowholedatathantheprecedingpiecharts.

RefertomybookShowMetheNumbers:DesigningTablesandGraphstoEnlighten(Oakland,CA:AnalyticsPress, 2004)forathoroughtreatmentofthetypesofgraphsthatworkbestforthemostcommonquantitativemessages communicatedinbusiness.


1

www.it-ebooks.info

Othertypesofgraphscanbeequallyineffective.Forexample,thegraphinFigure312showslittleregard fortheviewer'stimeandnounderstandingofvisualperception.Thisgraphcomparesrevenuetooperating costsacrossfivemonths,usingthesizeofoverlappingcircles(sometimescalledbubbles)toencodethe quantities.Justaswiththeslicesofapie,usingcirclestoencodequantityreliesontheviewer'sabilityto comparetwodimensionalareas,whichwesimplycannotaccuratelydo.Takethevaluesforthemonthof Februaryasanexample.Assumingthatoperatingcostsequal$10,000,whatistherevenuevalue?

Figure312.Thisgraphusesthetwodimensionalareaofcirclestoencodetheirvalues,whichneedlesslyobscuresthedata.

Ournaturaltendencyistocomparethesizesofthetwocirclesusingasingledimensionlengthor widthequaltothediameterofeach,whichsuggeststhatrevenueisaboutthreetimesthatofoperating costs,orabout$30,000.Thisconclusioniswrong,however,toahugedegree.Thetwodimensionalareaof therevenuecircleisactuallyaboutninetimesbiggerthanthatoftheoperatingcostscircle,resultingina valueof$90,000.Oops!Notevenclose. NowcompareoperatingcostsforthemonthsofFebruaryandMay.ItappearsthatcostsinMayaregreater thanthoseinFebruary,right?Infact,theinteriorcirclesarethesamesizemeasurethemandsee.The revenuebubbleinMayissmallerthantheoneinFebruary,whichmakestheenclosedoperatingcosts bubbleinMayseembigger,butthisisanopticalillusion.Asyoucansee,theuseofabubblechartforthis financialdatawasapoorchoice.AsimplebargraphliketheoneinFigure313worksmuchbetter.

Figure313.Thisbargraphdoesagoodjobofdisplayingatimeseriesofactualversusbudgetedrevenuevalues.

ActualversusbudgetedrevenueisalsothesubjectofFigure314,butthistimeit'ssubdividedinto geographicalregionsratherthantimeslicesanddisplayedasaradargraph.Thequantitativescaleona radargraphislaidalongeachoftheaxislinesthatextendfromthecentertotheperimeter,likeradiuslines

www.it-ebooks.info

ofacircle.Thesmallestvaluesarethosewiththeshortestdistancebetweenthecenterpointandthe perimeter.

Figure314.Thisradargraphobscuresthestraightforwarddatathatit'stryingtoconvey.

Thelackoflabeledaxesinthisgraphlimitsitsmeaning,butthechoiceofaradargraphtodisplaythis informationinthefirstplaceisanevenmorefundamentalerror.Onceagain,asimplebargraphlikethe oneinFigure315wouldcommunicatethisdatamuchmoreeffectively.Radargraphsarerarely appropriatemediafordisplayingbusinessdata.Theircircularshapeobscuresdatathatwouldbequiteclear inalineardisplaysuchasabargraph.

Figure315.Thisbargrapheffectivelycomparesactualtobudgetedrevenuedata.

ThelastexamplethatI'llusetoillustratemypointaboutchoosinginappropriatemeansofdisplayappears inFigure316.

www.it-ebooks.info

Figure316.ThisdisplayuselesslyencodesquantitativevaluesonamapoftheUnitedStates.

Therearetimeswhenitisveryusefultoarrangedataspatially,suchasintheformofamaporthefloor planofabuilding,butthisisn'toneofthem.Wedon'tderiveanyinsightbylayingoutrevenue informationinthiscase,whetherrevenuesaregood(greenlight),mediocre(yellowlight),orpoor(red light),inthegeographicalregionsSouth(brown),Central(orange),West(tan),andEast(blue)onamap. Ifthegraphicaldisplaywerepresentingmeaningfulgeographicalrelationshipssay,forshipmentsofwine fromCalifornia,toindicatewherespecialtaxesmustbepaidwheneverdeliveriescrossstatelinesperhapsa geographicaldisplaywouldprovidesomeinsight.Withthissimplesetoffourregionswithnoparticular factorsattachedtogeographicallocation,however,theuseofamapsimplytakesupalotofspacetosay nomorethanwefindinthetablethatappearsonthissamedashboard,whichisshowninFigure317.

Figure317.Thistable,fromthesamedashboard,providesamoreappropriatedisplayoftheregionalrevenuedatathatappears inFigure316.

3.6.IntroducingMeaninglessVariety
Themistakeofintroducingmeaninglessvarietyintoadashboarddesigniscloselytiedtotheonewejust examined.I'vefoundthatpeopleoftenhesitatetousethesametypeofdisplaymediummultipletimeson adashboard,outofwhatIassumeisasensethatviewerswillbeboredbythesameness.Varietymightbe thespiceoflife,butifitisintroducedonadashboardforitsownsake,thedisplaysuffers.Youshould alwaysselectthemeansofdisplaythatworksbest,evenifthatresultsinadashboardthatisfilledwith nothingbutmultipleinstancesofthesametypeofgraph.Ifyouaregivingviewerstheinformationthat theydesperatelyneedtodotheirjobs,thedatawon'tborethemjustbecauseit'salldisplayedinthesame way.Theywilldefinitelygetaggravated,however,ifforcedtoworkharderthannecessarytogetthe informationtheyneedduetoarbitraryvarietyinthedisplaymedia.Infact,whereverappropriate,

www.it-ebooks.info

consistencyinthemeansofdisplayallowsviewerstousethesameperceptualstrategyforinterpretingthe data,whichsavestimeandenergy. Figure318illustratesvarietygoneamok.Thisvisualjumblerequiresashiftinperceptualstrategyforeach displayitemonthedashboard,whichmeansextratimeandeffortontheuser'spart.

Figure318.Thisdashboardexhibitsanunnecessaryvarietyofdisplaymedia.

3.7.UsingPoorlyDesignedDisplayMedia
Itisn'tenoughtochoosetherightmediumtodisplaythedataanditsmessageyoualsomustdesignthe componentsofthatmediumtocommunicateclearlyandefficiently,withoutdistraction.Mostgraphsused inbusinesstodayarepoorlydesigned.Thereasonissimple:almostnoonehasbeentrainedinthe fundamentalprinciplesandpracticesofeffectivegraphdesign.Thiscontentisthoroughlycoveredinmy bookShowMetheNumbers:DesigningTablesandGraphstoEnlighten,soIwon'trepeatmyselfhere. Instead,I'llsimplyillustratetheproblemwithafewexamples. Inadditiontothefactthatabargraphwouldhavebeenabetterchoicetodisplaythisdata(thedivisionof revenuebetweensixsales),Figure319exhibitsseveraldesignproblems.Lookatitforamomentandseeif youcanidentifyaspectsofitsdesignthatinhibitquickandeasyinterpretation.

www.it-ebooks.info

Figure319.Thispiechartillustratesseveraldesignproblems.

HerearetheprimaryproblemsthatIsee: Alegendwasusedtolabelandassignvaluestotheslicesofthepie.Thisforcesoureyestobounceback andforthbetweenthegraphandthelegendtogleanmeaning,whichisawasteoftimeandeffortwhen theslicescouldhavebeenlabeleddirectly. Theorderoftheslicesandthecorrespondinglabelsappearsrandom.Orderingthembysizewouldhave providedusefulinformationthatcouldhavebeenassimilatedinstantly. Thebrightcolorsofthepieslicesproducesensoryoverkill.Brightcolorsoughttobereservedforspecific datathatshouldstandoutfromtherest. ThepiechartinFigure320alsoillustratesaproblemwithcolorchoice.

Figure320.Thispiechartusesofcolorsfortheslicesthataretoomuchaliketobeclearlydistinguished.

Inthiscase,the11colorsthatwerechosenaretoosimilar.Itisdifficulttodeterminewhichofthehues alongtheyellowthroughorangetoredspectruminthelegendcorrespondstoeachsliceofthepie.This kindofeyestrainingexerciseisdeadly,especiallyonadashboard. AnotherexampleofanineffectivedisplaymediumisshowninFigure321.Thesemetersareanattemptto betruetothemetaphorofacardashboard.Noticethatthenumberslookjustliketheywouldonan odometer:theylackthecommasnormallyusedtodelineateeverysetofthreedigitstohelpusdistinguish thousandsfrommillions,andsoon.Inamisguidedefforttomakethesemeterslookrealistic,their developersmadethenumbershardertoreadengineersdesignedthesemeters,notbusinesspeople.Notice alsothatnumbersalongthequantitativescalearepositionedinsideratherthanoutsidetheaxis,whichwill causethemtobeobscuredbytheneedlewhenitpointsdirectlytothem,andthatthepositioningofthe

www.it-ebooks.info

textatthebottomofeachmeter(forexample,"4382934AmountSold"onthe"InternetRevenue"meter) obstructstheneedleformeasuresnearthebottomortopofthescale.

Figure321.Thesedashboardmetershavedefinitelytakenthedashboardmetaphortoofar.

Inthelastsection,Ispokeofbargraphsasapreferablealternativetocertainotherdisplaymedia. However,whilebargraphscandoanexcellentjobofdisplayingquantitativedata,theycanbemisusedas well.ExaminethegraphinFigure322,andtakeamomenttolistanyproblemswithitsdesignthatyousee. Writedownyourobservationsbelowbeforereadingon,ifyou'dlike.

Figure322.Thisbargraph,foundonadashboard,exhibitsseveraldesignproblems.

Youmighthavenoticedthatthegridlinesonthegraph(nottomentionthebackgroundpatternofcolored rectangles)donothingbutdistractfromthedata.Gridlinessuchasthese,especiallywhenvisually prominent,makeitmoredifficulttoseetheshapeofthedata.Perhapsyoualsonoticedthatthe3Deffect ofthegraphnotonlyaddednovalue,butalsomadethevaluesencodedbythebarshardertointerpret. Anythingelse?Well,thisgraphillustratesacommonproblemwithcolor.Whyiseachofthebarsadifferent color?Thecolorsaren'tbeingusedtoidentifythebars,aseachonehasalabeltoitsleft.Differencesinthe colorofdataencodingobjectsshouldalwaysbemeaningful;otherwise,theyneedlesslygrabourattention andcauseustosearchformeaningthatisn'tthere. ThedistinctcolorsofthebarsinFigure323do,thankfully,carrymeaning,butherethecolorsare distractinglybrightandthe3Deffectmakesthemhardtoread.

www.it-ebooks.info

Figure323.Thisbargraph,foundonadashboard,waspoorlydesignedinanumberofways.

However,thisisn'ttheproblemthatImostwantyoutonotice.Thepurposeofthegraphistocompare actualtobudgetedrevenuesforeachofthefourregions,butsomethingaboutitsdesignmakesthis difficult.Canyouseetheproblem?Givenitspurpose,thebarsforactualandbudgetedrevenuesforeach regionshouldhavebeenplacednexttooneanother.Astheyare,itisunnecessarilydifficulttocompare them.Simpledesignmistakeslikethiscansignificantlyunderminethesuccessofadashboard. Severaloftheexamplesthatwe'veexaminedhavebeenrenderedin3D,eventhoughthethirddimension ofdepthdoesn'tencodeanymeaning.Evenwhenthethirddimensionisusedtoencodeaseparate variable,however,itstillposesaproblem.ThegraphinFigure324usesthethirddimensionofdepthto representtime(thefourquartersoftheyear2001).Theprobleminthiscaseisn'tthatthethirddimension ismeaningless,butratherthatyoucan'treadeverythingonthechart.Thisiscausedbyocclusion.Adding thedimensionofdepthcausessomeofthebarstobehiddenbehindoroccludedbyothers.Forinstance, whatwerefaxrevenuesforQuarter3?Youcan'ttellbecausethebariscompletelyhidden.Whetherthe thirddimensionisusedtoencodedataornot,youshouldalmostalwaysavoid3Dgraphs.Exceptionsto thisrulearerarewhendisplayingtypicalbusinessdata.

Figure324.This3Dbargraphillustratestheproblemofocclusion.

www.it-ebooks.info

3.8.EncodingQuantitativeDataInaccurately
Sometimesgraphicalrepresentationsofquantitativedataaremistakenlydesignedinwaysthatdisplay inaccuratevalues.InFigure325,forinstance,thequantitativescalealongtheverticalaxiswasimproperly setforagraphthatencodesdataintheformofbars.Thelengthofabarrepresentsitsquantitativevalue. ThebarsinthisgraphthatrepresentrevenueandcostsforthemonthofJanuarysuggestthatrevenuewas aboutfourtimescosts.Anexaminationofthescale,however,revealstheerrorofthisnaturalassumption: therevenueisactuallylessthandoublethecosts.Theproblemisthatthevaluesbeginat$500,000rather than$0,astheyalwaysshouldinabargraph.

Figure325.Thisbargraphencodesthequantitativevaluesasbarsinaccurately,byfailingtobeginthescaleatzero.

3.9.ArrangingtheDataPoorly
Dashboardsoftenneedtopresentalargeamountofinformationinalimitedamountofspace.Ifthe informationisn'torganizedwell,withappropriateplacementofinformationbasedonimportanceand desiredviewingsequence,alongwithavisualdesignthatsegregatesdataintomeaningfulgroupswithout fragmentingitintoaconfusinglabyrinth,theresultisaclutteredmess.Mostexamplesofdashboards foundontheWebarecomposedofasmallamountofdatatoavoidtheneedforskilledvisualdesign,but theystilloftenmanagetolookclutteredandthrowntogether.Thegoalisnotsimplytomakethe dashboardlookgood,buttoarrangethedatainamannerthatfitsthewayit'sused.Themostimportant dataoughttobeprominent.Datathatrequireimmediateattentionoughttostandout.Datathatshouldbe comparedoughttobearrangedandvisuallydesignedtoencouragecomparisons. ThedashboardinFigure326illustratessomeoftheproblemsoftenassociatedwithpoorarrangementof data.Noticefirstofallthatthemostprominentpositiononthisdashboardthetopleftisusedtodisplaythe vendor'slogoandnavigationalcontrols.Whatawasteofprimerealestate!Asyouscandownthescreen, thenextinformationthatyouseeisagaugethatpresentstheaverageordersize.It'spossiblethataverage ordersizemightbesomeone'sprimaryinterest,butit'sunlikelythat,ofalltheinformationthatappearson thisdashboard,thisisthemostimportant.AsI'lldiscussinChapter5,EloquenceThroughSimplicity,the leastprominentrealestateonthescreenisthelowerrightcorner.However,inthisexamplethelarge amountofspacetakenupbythegraphsthatpresent"ComputersReturnsAcrossModels,"aswellasthe largerfontsizesusedinthissection,tendstodrawattentiontodatathatseemstangentialtotherest.This dashboardlacksanappropriatevisualsequenceandbalancebasedonthenatureandimportanceofthe

www.it-ebooks.info

data.Noticealsothatthebrightredbandsofcoloraboveeachsectionofthedisplay,wherethetitles appearinwhite,arefarmoreeyecatchingthanisnecessarytodeclarethemeaningsoftheindividual displays.Thisvisuallysegmentsthespacetoanunnecessarydegree.Lastly,notethatthesimilarityofthe linegraphsthatdisplayordersizeandprofittrendsinvitesoureyestocomparethem.Thisisprobablya usefulcomparison,butthepositionalseparationandsidebysideratherthanoverunderarrangementof thetwographsmakesclosecomparisondifficult.Asthisexampleillustrates,youcan'tjustthrow informationontothescreenwhereveryoucanmakeitfitandexpectthedashboardtodoitsjob effectively.

Figure326.Thisdashboardexemplifiespoorlyarrangeddata.

3.10.HighlightingImportantDataIneffectivelyorNotatAll
Whenyoulookatadashboard,youreyesshouldimmediatelybedrawntotheinformationthatismost important,evenwhenitdoesnotresideinthemostvisuallyprominentareasofthescreen.InChapter5, EloquenceThroughSimplicity,we'llexamineseveralvisualtechniquesthatcanbeusedtoachievethisend. Fornow,we'lllookatwhathappenswhenthisisn'tdoneatall,orisn'tdonewell. TheproblemwiththedashboardinFigure327isthateverythingisvisuallyprominent,andconsequently nothingstandsout.Thelogoandnavigationcontrols(thebuttonsontheleft)areprominentbothasa resultoftheirplacementonthescreenandtheuseofstrongborders,butthesearen'tdataandtherefore shouldn'tbeemphasized.Thentherearethegraphswherethedatareside:allthedataareequallybold andcolorful,leavinguswithawashofsamenessandnocluewheretofocus.Everythingthatdeserves spaceonadashboardisimportant,butnotequallysotheviewer'seyesshouldalwaysbedirectedtothe mostcrucialinformationfirst.

www.it-ebooks.info

Figure327.Thisdashboardfailstodifferentiatedatabyitsimportance,givingrelativelyequalprominencetoeverythingonthe screen.

3.11.ClutteringtheDisplaywithUselessDecoration
AnothercommonproblemonthedashboardsthatIfindonvendorwebsitesistheabundanceofuseless decoration.Theyeitherhopethatwewillbedrawninbytheartistryorassumethatthedecorative flourishesarenecessarytoentertainus.Iassureyou,however,thatevenpeoplewhoenjoythedecoration uponfirstsightwillgrowwearyofitinafewdays. ThemakersofthedashboardinFigure328didanexceptionaljobofmakingitlooklikeanelectronic controlpanel.Ifthepurposeweretotrainpeopleintheuseofsomerealequipmentbymeansofa simulation,thiswouldbegreat,butthatisn'tthepurposeofadashboard.Thegraphicsdedicatedtothis endarepuredecoration,visualcontentthattheviewermustprocesstogettothedata.

www.it-ebooks.info

Figure328.Thisdashboardistryingtolooklikesomethingthatitisnot,resultinginuselessanddistractingdecoration.

IsuspectthatthedashboardinFigure329lookedtooplaintoitsdesigner,soshedecidedtomakeitlook likeapageinaspiralboundbookcute,butadistractingwasteofspace.

Figure329.Thisdashboardisanotherexampleofuselessdecorationthedesignertriedtomakethedashboardlooklikeapage inaspiralboundnotebook.

www.it-ebooks.info

Likewise,I'dguessthatthedesignerofthedashboardinFigure330aftercreatingamap,abargraph,and atablethatalldisplaythesamedatadecidedthathehadtofilluptheremainingspace,sohewentwild withanexplosionofblueandgraycircles.Blankspaceisbetterthanmeaninglessdecoration.Canyou imagineyourselflookingatthiseveryday?

Figure330.Thisdashboardisavividexampleofdistractingornamentation.

Thelastexample,Figure331,includesseveralelementsofdecorationthatoughttobeeliminated.To beginwith,avisuallyornatelogoandtitleuseupthemostvaluablerealestateacrosstheentiretopofthe dashboard.Ifalogomustbeincludedforbrandingpurposes,makeitsmallandvisuallysubtle,andplaceit somewhereoutoftheway.Thebackgroundcolorsofgoldandbluecertainlydrawoureyestothedata,but theydosoinanunnecessarilyheavyhandedmanner.Also,thecolorgradientsfromdarktolightprovide visualinterestthatsupportsnorealpurposeandisthereforedistracting.Lastly,themapsinthe backgroundofthethreeuppergraphs,thoughvisuallymuted,stilldistractfromthedataitself.

www.it-ebooks.info

Figure331.Thisdashboardexhibitsseveralexamplesofdysfunctionaldecoration.

AsdatavisualizationexpertEdwardTufteobserves: Ineptgraphicsalsoflourishbecausemanygraphicartistsbelievethatstatisticsare boringandtedious.Itthenfollowsthatdecoratedgraphicsmustpepup,animate, andalltoooftenexaggeratewhatevidencethereisinthedataIfthestatisticsare boring,thenyou'vegotthewrongnumbers.1

3.12.MisusingorOverusingColor
We'vealreadyseenseveralexamplesofmisusedoroverusedcolor.TheremainingpointthatIwantto emphasizehereisthatcolorshouldnotbeusedhaphazardly.

EdwardR.Tufte,TheVisualDisplayofQuantitativeInformation(Cheshire,CT:GraphicsPress,1983),80.

www.it-ebooks.info

Colorchoicesshouldbemadethoughtfully,withanunderstandingofhowweperceivecolorandthe significanceofcolordifferences.Somecolorsarehotanddemandourattention,whileothersarecooler andlessvisible.Whenanycolorappearsasacontrastrelativetothenorm,oureyespayattentionandour brainsattempttoassignmeaningtothatcontrast.Whencolorsintwodifferentsectionsofadashboardare thesame,wearetemptedtorelatethemtooneanother.Wemerrilyassumethatwecanusecolorssuch asred,yellow,andgreentoassignimportantmeaningstodata,butindoingsoweexcludethe10%of malesand1%offemaleswhoarecolorblind.InChapter4,TappingintothePowerofVisualPerception, we'lllearnabitaboutcolorandhowitcanbeusedmeaningfullyandpowerfully.

3.13.DesigninganUnattractiveVisualDisplay
Notbeingonetomincewordsforthesakeofpropriety,I'llstatequitedirectlythatsomedashboardsare justplainugly.Whenweseethem,we'reinclinedtoavertoureyeshardlythedesiredreactiontoascreen that'ssupposedtobesupplyinguswithimportantinformation.Youmighthaveassumedfrommyearlier warningagainstunnecessarydecorationthatIhavenoconcernfordashboardaesthetics,butthat'snotthe case.Whenadashboardisunattractiveunpleasanttolookattheviewerisputinaframeofmindthatis notconducivetoitsuse.I'mnotadvocatingthatweaddtouchestomakedashboardspretty,butrather thatweattractivelydisplaythedataitself,withoutaddinganythingthatdistractsfromorobscuresit.(We'll examinetheaestheticsofdashboarddesignabitinChapter7,DesigningDashboardsforUsability.) Figure332onthenextpageisastellarexampleofunattractivedashboarddesign.Itappearsthatthe personwhocreatedthisdashboardattemptedtomakeitlooknice,buthejustdidn'thavethevisualdesign skillsneededtosucceed.Forinstance,inanefforttofillupthespace,somesections(suchasthegraphat thebottomright)weresimplystretched.Also,althoughshadesofgraycanbeusedeffectivelyasthe backgroundcolorofgraphs,thisparticularshadeistoodark.Theimagethatappearsunderthetitle "Manufacturing"isclearlyanattempttoredeemthisdrearydashboardwithasplashofdecoration,butit onlyservestodistractfromthedataandisn'tevenparticularlynicetolookat.Theguidingdesignprinciple ofsimplicityalonewouldhavesavedthisdashboardfromitscurrentagony.

www.it-ebooks.info

Figure332.Thisisanexampleofaratherunattractivedashboard.

Youdon'tneedtobeagraphicartisttodesignanattractivedashboard,butyoudoneedtounderstanda fewbasicprinciplesaboutvisualperception.We'llexaminetheseinthenextchapter.

www.it-ebooks.info

Chapter4.TappingintothePowerofVisualPerception
Visionisbyfarourmostpowerfulsense.Seeingandthinkingareintimatelyconnected.Todisplaydata effectively,wemustunderstandabitaboutvisualperception,gleaningfromtheavailablebodyofscientific researchthosefindingsthatcanbeapplieddirectlytodashboarddesign:whatworks,whatdoesn't,and why.

www.it-ebooks.info

Understandingthelimitsofshorttermmemory Visuallyencodingdataforrapidperception Gestaltprinciplesofvisualperception Itisn'taccidentalthatwhenwebegintounderstandsomethingwesay,"Isee."Not"Ihear"or"Ismell,"but "Isee."Visiondominatesoursensorylandscape.Asasensophile,Icherishtherichabundanceofsounds, smells,tastes,andtexturesthatinhabitourworld,butnoneoftheseprovidestherichvolume,bandwidth, andnuanceofinformationthatIperceivethroughvision.Approximately70%ofthesensereceptorsinour bodiesarededicatedtovision,andIsuspectthatthereisastrongcorrelationbetweentheextensive brainpoweranddominanceofvisualperceptionthathavecoevolvedinourspecies.Howweseeisclosely tiedtohowwethink. I'velearnedaboutvisualperceptionfrommanysources,butonestandsoutabovetheothersinits applicationtoinformationdesign:thebookInformationVisualization:PerceptionforDesignbyColinWare. Dr.Wareexpressestheimportanceofstudyingvisualperceptionbeautifully: Whyshouldwebeinterestedinvisualization?Becausethehumanvisualsystemisa patternseekerofenormouspowerandsubtlety.Theeyeandthevisualcortexofthe brainformamassivelyparallelprocessorthatprovidesthehighestbandwidth channelintohumancognitivecenters.Athigherlevelsofprocessing,perceptionand cognitionarecloselyinterrelatedHowever,thevisualsystemhasitsownrules.We caneasilyseepatternspresentedincertainways,butiftheyarepresentedinother ways,theybecomeinvisibleThemoregeneralpointisthatwhendataispresented incertainways,thepatternscanbereadilyperceived.Ifwecanunderstandhow perceptionworks,ourknowledgecanbetranslatedintorulesfordisplaying information.Followingperceptionbasedrules,wecanpresentourdatainsucha waythattheimportantandinformativepatternsstandout.Ifwedisobeytherules, ourdatawillbeincomprehensibleormisleading.1 We'llconcentrateourlookatvisualperceptiononthefollowingareas: Thelimitsofshorttermvisualmemory Visualencodingforrapidperception Gestaltprinciplesofvisualperception

Eachofthesetopicscanbeapplieddirectlytothedesignofdashboards.

4.1.UnderstandingtheLimitsofShortTermMemory
Intruth,wedon'tseewithoureyes;weseewithourbrains.Oureyesarethesensorymechanismsthrough whichlightentersandistranslatedbyneuronsintoelectricalimpulsesthatarepassedontoandaroundin ourbrains,butourbrainsarewhereperceptiontheprocessofmakingsenseofwhatoureyes registeractuallyoccurs.

ColinWare,InformationVisualization:PerceptionforDesign,SecondEdition(SanFrancisco:MorganKauffman, 2004),xxi.
1

www.it-ebooks.info

Oureyesdonotregistereverythingthatisvisibleintheworldaroundus,butonlywhatlieswithintheir spanofperception.Onlyaportionofwhatoureyessensebecomesanobjectoffocus.Onlythroughfocus doeswhatweseebecomemorethanavaguesense.Onlyafractionofwhatwefocusonbecomesthe objectofattentionorconsciousthought.Finally,onlyalittlebitofwhatweattendtogetsstoredawayfor futureuse.Withouttheselimitsandfilters,perceptionwouldoverwhelmourbrains. Ourmemoriesstoreinformationstartingthemomentweseesomething,continuingasweconsciously processtheinformation,andfinallyaccumulatingoveryearsinapermanent(ornearlyso)storagearea whereinformationremainsreadyforuseifeverneededagainthatis,untilaccesstothatinformation eventuallybeginstoatrophy. Memorycomesinthreefundamentaltypes: Iconicmemory(a.k.a.thevisualsensoryregister) Shorttermmemory(a.k.a.workingmemory) Longtermmemory

Iconicmemoryisalotlikethevisualmemorybufferofacomputer:aplacewhereimagesarebrieflyheld untiltheycanbemovedtorandomaccessmemory(RAM),wheretheyresidewhilebeingprocessedbythe CPU.Eventhoughwhatgoesoniniconicmemoryispreconscious,acertaintypeofprocessingknownas preattentiveprocessingoccursnonetheless.Certainattributesofwhatweseearerecognizedduring preattentiveprocessingatanextraordinarilyhighspeed,whichresultsincertainthingsstandingoutand particularsetsofobjectsbeinggroupedtogether,allwithoutconsciousthought.Preattentiveprocessing playsapowerfulroleinvisualperception,andwecanintentionallydesignourdashboardstotake advantageofthisifweunderstandabitaboutit. Shorttermmemoryiswhereinformationresidesduringconsciousprocessing.Themostimportantthings toknowaboutshorttermmemoryare: Itistemporary.1 Aportionofitisdedicatedtovisualinformation. Ithasalimitedstoragecapacity.

Wecanstoreonlythreetoninechunksofvisualinformationatatimeinshorttermmemory.Whenits capacityisfull,forsomethingnewtobebroughtintoshorttermmemory,somethingthat'salreadythere musteitherbemovedintolongtermmemoryorsimplyremovedaltogether(thatis,forgotten).What constitutesa"chunk"ofvisualinformationvariesdependingonthenatureoftheobjectsweareseeing, aspectsoftheirdesign,andourfamiliaritywiththem.Forinstance,individualnumbersonadashboardare storedasdiscretechunks,butawelldesignedgraphicalpattern,suchasthepatternformedbyoneor morelinesinalinegraph,canrepresentagreatdealofinformationasasinglechunk.Thisisoneofthe greatadvantagesofgraphs(whenusedappropriatelyandskillfullydesigned)overtext.Dashboardsshould bedesignedinawaythatsupportsoptimalchunkingtogetherofinformationsothatitcanbeperceived andunderstoodmostefficiently,inbigvisualgulps.
Informationremainsinshorttermmemoryfromafewsecondstoaslongasafewhoursifperiodicallyrehearsed; thenitisflushed.Ifrehearsedinaparticularway,informationismovedfromshorttermmemorytolongterm memory,whereitisstoredmorepermanentlyforlaterrecall.Wheninformationisrecalledfromlongtermmemory, itistemporarilymovedonceagainintoshorttermmemory,whereitisprocessed.
1

www.it-ebooks.info

Thelimitedcapacityofshorttermmemoryisalsothereasonwhyinformationthatbelongstogethershould neverbefragmentedintomultipledashboards,andscrollingshouldn'tberequiredtoseeitall.Oncethe informationisnolongervisible,unlessitisoneofthefewchunksofinformationstoredinshortterm memory,itisnolongeravailable.Ifyouscrollorpagebacktoseeitagain,youthenloseaccesstowhatyou weremostrecentlyviewing.Aslongaseverythingyouneedremainswithineyespanonasingledashboard, however,youcanrapidlyexchangeinformationinandoutofshorttermmemoryatlightningspeed.

4.2.VisuallyEncodingDataforRapidPerception
Preattentiveprocessing,theearlystageofvisualperceptionthatrapidlyoccursbelowthelevelof consciousness,istunedtodetectaspecificsetofvisualattributes.Attentiveprocessingissequential,and thereforemuchslower.Thedifferenceiseasytodemonstrate.Takeamomenttoexaminethefourrowsof numbersinFigure41,andtrytodetermineasquicklyasyoucanthenumberoftimesthenumber5 appearsinthelist.

Figure41.Howmanyfivesareinthislist?Notetheslowspeedatwhichweprocessvisualstimulithatlackpreattentive attributes.

Howmanydidyoufind?Thecorrectanswerissix.Whetheryougottheanswerrightornot,theprocess tookyouawhilebecauseitinvolvedattentiveprocessing.Thelistofnumbersdidnotexhibitany preattentiveattributesthatyoucouldusetodistinguishthefivesfromtheothernumbers.Nowtryitagain, thistimeusingthelistofnumbersinFigure42.

Figure42.Howmanyfivesdoyouseenow?Notethefastspeedatwhichweprocessvisualstimulithatexhibitpreattentive attributes.

Mucheasierthistime,wasn'tit?Inthisfigurethefivescouldeasilybedistinguishedfromtheother numbers,duetotheirdifferingcolorintensity(oneofthepreattentiveattributeswe'lldiscussbelow):the fivesareblackwhilealltheothernumbersaregray,whichcausesthemtostandoutinclearcontrast.Why couldn'tweeasilydistinguishthefivesinthefirstsetofnumbers(Figure41)basedpurelyontheirunique shape?Becausethecomplexshapesofthenumbersarenotattributesthatweperceivepreattentively. Simpleshapessuchascirclesandsquaresarepreattentivelyperceived,buttheshapesofnumbersaretoo elaborate. InInformationVisualization:PerceptionforDesign,ColinWaresuggeststhatthepreattentiveattributesof visualperceptioncanbeorganizedintofourcategories:color,form,spatialposition,andmotion.Forour presentinterestrelatedtodashboarddesign,I'vereducedhislargerlistof17preattentiveattributestothe following11:

www.it-ebooks.info

Table41.

Category Attribute Color Hue

Illustration

Intensity

Position

2Dlocation

Form

Orientation

Linelength

Linewidth

Size

www.it-ebooks.info

Shape

Added marks

Enclosure

Motion

Flicker

Avisualattributeofanobject,suchascolor,continuouslychangesbackandforth betweentwovalues,ortheentireobjectitselfrepeatedlyappearsandthen disappears.

Eachofthesevisualattributescanbeconsciouslyappliedtodashboarddesigntogrouporhighlight information.Somecanbeusedtoencodequantitativeinformationaswell,aswe'lldiscussbelow. 4.2.1.AttributesofColor Acommonwaytodescribecolorcombinesthreeattributes:hue,saturation,andlightness/brightness.This issometimesreferredtoastheHSLorHSBsystemofdescribingcolor.Hueisamoreprecisetermforwhat wenormallythinkofascolor(red,green,blue,purple,etc.).Saturationmeasuresthedegreetowhicha particularhueexhibitsitsfull,pureessence.ThesaturationoftheredhueinFigure43rangesfrom0% saturationontheleftto100%saturationontheright.

Figure43.Thefullrangeofcolorsaturation,inthiscaseofthehuered,with0%saturationontheleftand100%saturationon theright.

Lightness(orbrightness)measuresthedegreetowhichanyhueappearsdarkorlight,rangingfromfully dark(black)tofullylight(white).ThefullrangeoflightnessisshownfortheredhueinFigure44.

www.it-ebooks.info

Figure44.Thefullrangeofcolorlightness,inthiscaseofthehuered,with0%lightnessontheleft(pureblack)and100% lightnessontheright(purewhite).

Intensityreferstobothsaturationandlightness.TheillustrationofcolorintensityonSection4.2showsa circlethatvariesfromtheothersnotasadifferenthuebutasalighter(thatis,lessintense)versionofthe samehue.Botharedifferentpointsalongacolorscalethatrangesfromwhite(nobrown)toarichdark shadeofbrown(fullybrown).Itreallyisn'tnecessarytofullyunderstandthetechnicaldistinctionbetween saturationandlightness,whichiswhyIdescribethembothmoresimplyasintensity. Oneoftheinteresting(buthardlyintuitive)thingsaboutcoloristhatwedon'tperceivecolorinanabsolute way.Whatweseeisdramaticallyinfluencedbythecontextthatsurroundsit.Takealookatthegray squaresinFigure45.Theyappeartovaryinintensity,butinfacttheyareallexactlythesameasthelone squarethatappearsagainstawhitebackgroundatthebottom.

Figure45.Contextaffectsourperceptionofcolorintensity.Thesmallsquareisactuallytheexactsameshadeofgray everywhereitappears.

Allfivesquareshaveacolorvalueof50%black,yetthesurroundinggrayscalegradient,rangingfromlight onthelefttodarkontheright,altersourperceptionofthem.Thisperceptualillusionappliesnotonlyto intensity,buttohue.InFigure46,theword"Text"appearsagainsttwobackgrounds:redandblue.Inboth cases,thecoloroftheword"Text"isthesame.However,itnotonlylooksdifferent,butit'smuchless visibleagainsttheredbackground.

Figure46.Contextalsoaffectsourperceptionofhue.Theword"Text"isexactlythesamehueinbothboxes.

Colormustbeusedwithafullawarenessofcontext.Wenotonlywantdatatobefullylegible,butalsoto appearthesamewhenwewishittoappearthesameanddifferentwhenwewishittoappeardifferent. 4.2.2.AttributesofForm Someofthevisualattributesofformhavenoobviousconnectiontodashboarddesign,buttheirrelevance shouldbecomeclearwithalittleexplanation.Themostcommonapplicationoforientationisintheformof

www.it-ebooks.info

italicizedtext,whichistextthathasbeenreorientedfromstraightupanddowntoslightlyslantedtothe right.Iusuallydiscouragetheuseofitalicizedtextasameansofmakingsomewordsstandoutfromthe rest,becauseitalicsarehardertoreadthannormalverticallyorientedtext.However,itissometimesuseful inapinch. Indashboarddesign,theattributeoflinelengthismostusefulforencodingquantitativevaluesasbarsina bargraph.Linewidth,ontheotherhand,canbeusefulforhighlightingpurposes.Youcanthinkofline widthasthethicknessorstrokeweightofaline.Whenlinesareusedtounderlinecontentor,intheform ofboxes,toformbordersaroundcontent,youcandrawmoreattentiontothatcontentbyincreasingthe thicknessofthelines. Therelativesizesofobjectsthatappearonadashboardcanbeusedtovisuallyranktheirimportance.For instance,largertitlesforsectionsofcontent,orlargertables,graphs,oricons,canbeusedtodeclarethe greaterimportanceoftheassociateddata.Simpleshapescanbeusedingraphstodifferentiatedatasets and,intheformoficons,toassigndistinctmeanings,suchasdifferenttypesofalerts.Addedmarksare mostusefulondashboardsintheformofsimpleiconsthatappearnexttodatathatneedattention.Any simplemark(suchasacircle,asquare,anasterisk,oranX),whenplacednexttoinformationonlywhenit mustbehighlighted,worksasasimplemeansofdrawingattention.Lastonthelistofformattributesis enclosure,whichisapowerfulmeansofgroupingsectionsofdataor,whenusedsparingly,highlighting contentasimportant.Tocreatethevisualeffectofenclosure,youcanuseeitheraborderorafillcolor behindthecontent. 4.2.3.AttributesofPosition Thepreattentiveattribute2Dpositionistheprimarymeansthatweusetoencodequantitativedatain graphs(forexample,thepositionofdatapointsinrelationtoaquantitativescale).Thisisn'tarbitrary.Ofall thepreattentiveattributes,differencesin2Dpositionaretheeasiestandmostaccuratetoperceive.1 4.2.4.AttributesofMotion AsItypethesewords,Iamawareofmycursorflickeringonandoffonthescreen.Flickerwaschosenas themeanstohelpuslocatethecursorbecauseitisapowerfulattentiongetter.Evolutionhasequippedus withaheightenedsensitivitytosomethingthatsuddenlyappearswithinourfieldofvision.Ourancient ancestorsfounditveryvaluabletobecomeinstantlyalertwhenasabertoothedtigersuddenlyspranginto theirperipheralvision.AsI'msureyou'reaware,flickeringobjectsonascreencanbequiteannoyingand thusshouldusuallybeavoided.Still,thereareoccasionswhenflickerisuseful.Thisisespeciallytruefor dashboardsthatareconstantlyupdatedwithrealtimedataandareusedtomonitoroperationsthat requireimmediateresponses. 4.2.5.EncodingQuantitativeVersusCategoricalData Someofthepreattentiveattributesthatwe'veexaminedcanbeusedtocommunicatequantitativedata, whileotherscanbeusedonlytocommunicatecategoricaldata.Thatis,whilesomeattributesallowusto perceiveonethingasgreaterthanothersinsomeway(bigger,taller,moreimportant),othersmerely indicatethatitemsaredistinctfromoneanother,withoutanysenseofsomebeinggreaterthanorless thanothers.Forexample,differentshapescanbeperceivedasdistinct,butonlycategorically.Squaresare
Perhapsyou'venoticedthatI'vespecified"2D"positionanobject'slocationrelativetotheverticalandhorizontal dimensionsonlyandhaveignored3Dposition,alsoknownasstereoscopicposition.3Dpositionisalsoapreattentive attribute,butthepseudo3Deffectthatcanbeproducedontheflatsurfaceofacomputerscreencomeswithabevy ofperceptualproblemsthatcomplicateitsuse.3Delementsaresorarelynecessarytocommunicatebusiness informationandsodifficulttodesigneffectivelythatIrecommendthatyouavoidusingthemaltogether.
1

www.it-ebooks.info

notgreaterthantrianglesorcirclesthey'rejustdifferent.Thefollowingtableagainlistseachofthe preattentiveattributesandindicateswhicharequantitativelyperceived: Table42. Category Color Position Form

Motion

Attribute Hue Intensity 2Dposition Orientation Linelength Linewidth Size Shape Addedmarks Enclosure Flicker

Quantitative No Yes,butlimited Yes No Yes Yes,butlimited Yes,butlimited No No No Yes,basedonspeed,butlimited

Note:Youmightarguethatwecanperceiveorientationandcurvaturequantitatively,butthereisno naturalassociationofgreaterorlesservaluewithdifferentorientationsordegreesofcurvature(for example,whichisgreater,averticalorhorizontalline?). Wecanusethoseattributeswithquantitativeperceptiondescribedas"Yes,butlimited"toencourage perceptionofonethingasgreaterthan,equalto,orlessthananother,butnotwithanydegreeof precision.Forexample,inFigure47,itisobviousthatthecircleontherightisbiggerthanthecircleonthe left,buthowmuchbigger?Ifthesmallcirclehasasizeofone,whatisthesizeofthebiggercircle?

Figure47.Thisillustratesourinabilitytoassignprecisequantitativevaluestoobjectsofdifferentsizes.

Thecorrectansweris16,butit'slikelythatyouguessedalowernumber.Humanstendtounderestimate differencesin2Dareas,andhenceyoumustbewaryofusing2Dareasofdifferentsizestoencode quantitativevaluesespeciallyonadashboard,wherespeedofinterpretationisessential.

www.it-ebooks.info

Itisimportanttounderstandthedifferentwaysthatthepreattentiveattributescanbeusedtogroupand encodedata,butbysplittingthemalongtheselinesquantitativeandcategoricaldonotmeantoimplythat onlythoseattributesthatenableviewerstomakequantitativecomparisonsareofusetodashboard designers.Ourinabilitytoperceivecertainpreattentiveattributesquantitativelydoesnotrenderthem uselesstous.Eachofthemcanbeusedtodividedataintodistinctcategories,tovisuallylinkdatatogether evenwhenitisseparatedspatially,andtohighlightdata. 4.2.6.LimitstoPerceptualDistinctness Whendesigningdashboards,bearinmindthatthereisalimittothenumberofdistinctexpressionsofa singlepreattentiveattributethatwecanquicklyandeasilydistinguish.Forexample,whenusingvarying intensitiesofthecolorgraytodistinguishdatasetsinalinegraph,youmustmakesurethatthecolorof eachlineisdifferentenoughfromthoseclosestincolortoittoclearlystandoutasdistinct.Whenyou placeenoughperceptualdistancebetweenthecolorintensitiesoftheseparatelinestomakethem sufficientlydistinct,there'sapracticallimitofaboutfivetothenumberofdistinctexpressionsthatare availableacrossthegrayscale.InFigure48,itiseasytoseethatitwouldbedifficulttoincludemoregray linesthatwouldstandoutasdistinctwithoutrequiringcareful,conscious,andthusslowexaminationon thepartoftheviewer.

Figure48.Thereisapracticallimitofaboutfivedistinctcolorintensitiesonthegrayscalecontinuumthatcanbeusedto encodeseparatelinesinagraph.

Similarlimitsapplytoeveryoneofthepreattentiveattributes,exceptlinelength(suchasthelengthofa baronagraph)and2Dlocation(suchasthelocationofadatapointonagraph).Whenorganizingdata intodistinctgroupsusingdifferentexpressionsofanypreattentiveattribute,youshouldbecarefulnotto exceedfivedistinctexpressions.Whenusingtheshapeattribute,inadditiontothislimityoumustalsobe carefultochooseshapesthataresimple,suchascircles,squares,triangles,dashes,andcrosses(orXs). Rememberthatcomplexshapes,includingmosticons,arenotperceivedpreattentively.Whenusinghue, keepinmindthateventhoughwecaneasilydistinguishmorethanfivehues,shorttermmemorycan't simultaneouslyretainthemeaningofmorethanaboutnineintotal.Also,theuseoftoomanyhuesresults inadashboardthatlookscluttered,withtoomanydistinctionstosortthroughquickly.Whendesigning dashboards,ithelpstopreparestandardsetsofhues,colorintensities,shapes,andsoonfromwhichto

www.it-ebooks.info

choose,andthensticktothem.Thiswillkeepthedisplayperceptuallysimpleandwilleliminatetheneedto selectvisualattributesfromscratcheachtimeyoumustchooseone. 4.2.7.UsingVividandSubtleColorsAppropriately ColorissooftenmisusedindashboarddesignthatI'mcompelledtoemphasizeonemoreprincipleofits use.Somecolorsaresoothing,andsometakeholdofusandshakeusaround.Knowingthedifferenceis quiteimportant.Therearetimeswhenparticularinformationneedstograbtheviewer'sattentioninan unavoidableway,butusingcolorforthispurposeworksonlyifit'sdonesparingly.Reservetheuseof bright,fullysaturatedcolorforthesespecialcases.Colorsthatarecommoninnature,suchassoftgrays, browns,oranges,greens,andblues,workverywellasastandardcolorpalettefordashboards.Theyallow theviewertoperusethedashboardcalmlywithanopenmind,ratherthanstressfully,withpinpoint attentioninresponsetoassaultingcolors.Figure49displaysexamplesofstandardandboldcolorpalettes.

Figure49.Examplesoftwocolorpalettes:oneforstandarduseandoneforemphasis.

4.3.GestaltPrinciplesofVisualPerception
Backin1912,theGestaltSchoolofPsychologybeganitsfruitfuleffortstounderstandhowweperceive pattern,form,andorganizationinwhatwesee.TheGermanterm"gestalt"simplymeans"pattern."These researchersrecognizedthatweorganizewhatweseeinparticularwaysinanefforttomakesenseofit. TheirworkresultedinacollectionofGestaltprinciplesofperceptionthatrevealthosevisualcharacteristics thatinclineustogroupobjectstogether.Theseprinciplesstillstandtodayasaccurateanduseful descriptionsofvisualperception,andtheyofferseveralusefulinsightsthatwecanapplydirectlyinour dashboarddesignstointentionallytiedatatogether,separatedata,ormakesomedatastandoutas distinctfromtherest. We'llexaminethefollowingsixprinciples: Proximity Closure Similarity Continuity Enclosure Connection

www.it-ebooks.info

4.3.1.ThePrincipleofProximity Weperceiveobjectsthatarelocatednearoneanotherasbelongingtothesamegroup.Figure410clearly illustratesthisprinciple.Basedontheirrelativelocations,weautomaticallyseethedotsasbelongingto threeseparategroups.Thisisthesimplestwaytolinkdatathatyouwanttobeseentogether.Whitespace aloneisusuallyallyouneedtoseparatethesegroupsfromtheotherdatathatsurroundsthem.

Figure410.TheGestaltprincipleofproximityexplainswhywesee3groupsinsteadofjust10dotsinthisimage.

Theprincipleofproximitycanalsobeusedtodirectviewerstoscandataonadashboardpredominantlyin aparticulardirection:eitherlefttorightortoptobottom.Placingsectionsofdataclosertogether horizontallyencouragesviewers'eyestogroupthesectionshorizontally,andthustoscanfromlefttoright. Placingsectionsofdataclosertogetherverticallyachievestheoppositeeffect. NoticehowsubtlythisworksinFigure411.Youarenaturallyinclinedtoscanthesmallsquaresthatappear onthelefthorizontallyasrowsandtheonesontherightverticallyascolumns,allbecauseofhowtheyare positionedinrelationtoeachother.

Figure411.TheGestaltprincipleofproximitycanbeusedtoencourageeitherhorizontalorverticalscanning.

4.3.2.ThePrincipleofSimilarity Wetendtogrouptogetherobjectsthataresimilarincolor,size,shape,andorientation.Figure412 illustratesthistendency.

www.it-ebooks.info

Figure412.Whenobjectssharesomevisualattributeincommon,wetendtoseethemasbelongingtothesamegroup.

Thisprinciplereinforceswhatwe'vealreadylearnedabouttheusefulnessofcolor(bothhueandintensity), size,shape,andorientationtoencodecategoricalvariables.Theprincipleofsimilarityappliesvery effectivelytogroupsofvisualobjectsthatvaryasdifferentexpressionsofpreattentiveattributessuchas these.Itworksespeciallywellasameansofidentifyingdifferentdatasetsinagraph(forexample,income, expenses,andprofits).Evenwhendatathatwewishtolinkresidesinseparatelocationsonadashboard, theprincipleofsimilaritycanbeappliedtoestablishthatlink. Forinstance,ifyouwishtotietogetherrevenueinformationthatappearsinvariousgraphs,youcandoso byusingthesamecolortoencodeitwhereveritappears.Thistechniquecanbeusefulforencouraging comparisonsofanydatathatappearinvariousplaces,suchasordercount,ordersize,andorderrevenue. 4.3.3.ThePrincipleofEnclosure Weperceiveobjectsasbelongingtogetherwhentheyareenclosedbyanythingthatformsavisualborder aroundthem(forexample,alineoracommonfieldofcolor).Thisenclosurecausestheobjectstoappear tobesetapartinaregionthatisdistinctfromtherestofwhatwesee.Noticehowstronglyyoureyesare inducedtogrouptheenclosedobjectsinFigure413.

Figure413.TheGestaltprincipleofenclosurepointsoutthatanyformofvisualenclosurecausesustoseetheenclosedobjects asagroup.

Thearrangementofthetwosetsofcirclesinthisfigureisexactlythesame,yetthedifferingenclosures directustogroupthecirclesinverydifferentways.Thisprincipleisexhibitedfrequentlyintheuseof bordersandfillcolorsorshadingintablesandgraphstogroupinformationandsetitapart.Asyoucansee, itdoesnottakeastrongenclosure(e.g.,bright,thicklinesordominantcolors)tocreateastrong perceptionofgrouping.

www.it-ebooks.info

4.3.4.ThePrincipleofClosure Humanshaveakeendislikeforlooseends.Whenfacedwithambiguousvisualstimuliobjectsthatcouldbe perceivedeitherasopen,incomplete,andunusualformsorasclosed,whole,andregularformswe naturallyperceivethemasthelatter.Theprincipleofclosureassertsthatweperceiveopenstructuresas closed,complete,andregularwheneverthereisawaythatwecanreasonablydoso.Figure414illustrates thisprinciple.

Figure414.TheGestaltprincipleofclosureexplainswhyweseetheseasclosedshapes,despitethefactthattheyarenot finished.

ItisnaturalforustoperceivewhatappearsontheleftinFigure414asarectangleratherthantwosetsof threeconnectedlinesconnectedatrightanglesandtoperceivetheobjectontherightasacompleteoval ratherthansimplyacurvedline. Wecanapplythistendencytoperceivewholestructuresindashboards,especiallyinthedesignofgraphs. Forexample,wecangroupobjects(points,lines,orbarsinagraph,etc.)intovisualregionswithouttheuse ofcompletebordersorbackgroundcolorstodefinethespace.Thisispreferable,becausetheneedto displayalargecollectionofdatainasmallamountofspacerequiresthatweeliminateallvisualcontent thatisnotabsolutelynecessary,toavoidclutter.AsshowninFigure415,itissufficienttodefinethearea ofagraphthroughtheuseofasinglesetofXandYaxes,ratherthanbylinesthatformacomplete rectanglearoundthegraph,withorwithoutafillcolor.

Figure415.TheGestaltprincipleofclosurealsoexplainswhyonlytwoaxes,ratherthanfullenclosure,arerequiredonagraph todefinethespaceinwhichthedataappears.

www.it-ebooks.info

4.3.5.ThePrincipleofContinuity Weperceiveobjectsasbelongingtogether,aspartofasinglewhole,iftheyarealignedwithoneanotheror appeartoformacontinuationofoneanother.InFigure416,forinstance,wetendtoseetheindividual linesasacontinuationofoneanother,moreasadashedlinethanseparatelines.

Figure416.TheGestaltprincipleofcontinuityexplainswhyweseethisasasinglewavyline.

Thingsthatarealignedwithoneanotherappeartobelongtothesamegroup.InthetableinFigure417,it isobviouswhichitemsaredivisionnamesandwhicharedepartmentnames,basedontheirdistinct alignment.Divisions,departments,andheadcountsareclearlygrouped,withoutanyneedforverticalgrid linestodelineatethem.Eventhoughthedivisionanddepartmentcolumnsoverlapwithnowhitespacein between,theirdistinctalignmentalonemakesthemeasytodistinguish.Thissametechniquecanbeused totietogetherseparatesectionsofdataonadashboard.

Figure417.TheGestaltprincipleofcontinuityalsoexplainshowtheindentationoftextworksasameanstogroupinformation.

4.3.6.ThePrincipleofConnection Weperceiveobjectsthatareconnectedinsomeway,suchasbyaline,aspartofthesamegroup.InFigure 418,eventhoughthecirclesarenearertooneanotherverticallythanhorizontally,thelinesthatconnect themcreateaclearperceptionoftwohorizontallyattachedpairs.

www.it-ebooks.info

Figure418.TheGestaltprincipleofconnectionexplainswhyweseethesedotsgroupedbyrowsratherthancolumns.

AsFigure419illustrates,theperceptionofgroupingproducedbyconnectionisstrongerthanthat producedbyproximityorsimilarity(color,size,andshape);itisweakeronlythanthatproducedby enclosure.Theprincipleofconnectionisespeciallyusefulfortyingtogethernonquantitativedatafor example,torepresentrelationshipsbetweenstepsinaprocessorbetweenemployeesinanorganization.

Figure419.Whenobjectsareconnected,suchasbythelinesintheseexamples,theyaregroupedtogethermorepowerfully thanbyjustaboutanyothervisualmeans.Onlytheenclosureintherightmostexamplemorestronglygroupsthetwosquares ontherightthantheconnectionsformedbythelines.

4.4.ApplyingthePrinciplesofVisualPerceptiontoDashboardDesign
Twoofthegreatestchallengesindashboarddesignaretomakethemostimportantdatastandoutfrom therest,andtoarrangewhatisoftenagreatdealofdisparateinformationinawaythatmakessense,gives itmeaning,andsupportsitsefficientperception.Anunderstandingofthepreattentiveattributesofvisual perceptionandtheGestaltprinciplesprovidesausefulconceptualfoundationforfacingthesechallenges.It ismuchmorehelpfultounderstandhowandwhysomethingworksthantosimplyunderstandthat somethingworks.Ifyouunderstandthehowandwhy,whenyou'refacedwithnewchallengesyou'llbe abletodeterminewhetherornottheprinciplesapplyandhowtoadaptthemtothenewcircumstances.If you'vesimplybeentoldthatsomethingworksinaspecificsituation,you'llbestuckwhenfacedwith conditionsthatareevenslightlydifferent. Asyouproceedintothecomingchapters,you'llhaveseveralopportunitiestoreinforceyourgraspofvisual perceptionbyapplyingwhatyou'velearnedtoseveralrealworlddashboarddesignproblems.

www.it-ebooks.info

Chapter5.EloquenceThroughSimplicity
Nowthatyou'refamiliarwithsomeofthesciencebehinddashboarddesign,it'stimetotakealookatafew strategiesyoucanemploytocreateeffectivedisplays.Theguidingprincipleindashboarddesignshould alwaysbesimplicity:displaythedataasclearlyandsimplyaspossible,andavoidunnecessaryand distractingdecoration.

www.it-ebooks.info

Characteristicsofawelldesigneddashboard Reducingthenondatapixels Enhancingthedatapixels Inearlierchapters,weconcentratedonwhatdoesn'twork.Nowit'stimetoshiftourfocustowhatdoes, beginningwiththedesignprocessitselfthegoalsandstepsnecessarytoproducedashboardsthatinform rapidlywithimpeccableclarity.

5.1.CharacteristicsofaWellDesignedDashboard
Thefundamentalchallengeofdashboarddesigninvolvessqueezingagreatdealofusefulandoften disparateinformationintoasmallamountofspace,allthewhilepreservingclarity.Thiscertainlyisn'tthe onlychallengeothersabound,suchasselectingtherightdatainthefirstplacebutitistheprimary challengethatisparticulartodashboards.Limitedtoasinglescreentokeepallthedatawithineyespan, dashboardrealestateisextremelyvaluable:youcan'taffordtowasteaninch.Fittingeverythinginwithout sacrificingmeaningdoesn'trequiremuscles,itrequiresfinesse.

Figure51.Thefundamentalchallengeofdashboarddesignistoeffectivelydisplayagreatdealofoftendisparatedatainasmall amountofspace.

Unlessyouknowwhatyou'redoing,you'llendupwithaclutteredmess.Thinkforamomentaboutthe cockpitofacommercialjet.Yearsofeffortwentintoitsdesigntoensurethatdespitethemanythings

www.it-ebooks.info

pilotsmustmonitor,theycanseeeverythingthat'sgoingonataglance.EverytimeIboardaplane,I'm gratefulthatskilleddesignersworkedhardtopresentthisinformationeffectively.Similarcareisneededfor thedesignofdashboards,butunlikeaircraftcockpitdesign,fewofthosewhocreatedashboardshave actuallystudiedthescienceofdesign.Youcanbecomeanexceptiontothisunfortunateandcostlynorm.It isunlikelythatpeoplewilllosetheirlivesifyoufail,butbusinessesdooccasionallycrashandburnand frequentlylosemoneyduetofailedcommunicationofjustthissort. HenryDavidThoreauoncepennedthesamewordthreetimesinsuccessiontoemphasizeanimportant qualityoflifethatappliestodesignaswell:"Simplify,simplify,simplify!"1ThoughIoftenfail,Istrivetolive mylifeandtodesignallformsofcommunicationaccordingtoThoreau'ssageadvicetokeepthingssimple. Eloquenceincommunicationisoftenachievedthroughsimplification.Toooftenwesmearathicklayerof gaudymakeupoverdatainanefforttoimpressorentertain,ratherthanfocusingoncommunicatingthe truthofthematterintheclearestpossibleway. Whendesigningdashboards,youmustincludeonlytheinformationthatyouabsolutelyneed,youmust condenseitinwaysthatdon'tdecreaseitsmeaning,andyoumustdisplayitusingvisualdisplay mechanismsthat,evenwhenquitesmall,canbeeasilyreadandunderstood.Welldesigneddashboards deliverinformationthatis: Exceptionallywellorganized Condensed,primarilyintheformofsummariesandexceptions Specifictoandcustomizedforthedashboard'saudienceandobjectives Displayedusingconciseandoftensmallmediathatcommunicatethedataanditsmessagein theclearestandmostdirectwaypossible

Dashboardstellpeoplewhat'shappeningandshouldhelpthemimmediatelyrecognizewhatneedstheir attention.Justlikethedashboardofacar,whichprovideseasilymonitoredmeasuresofspeed,remaining fuel,oillevel,batterystrength,enginetrouble,andsoon,abusinessinformationdashboardprovidesan overviewthatcanbeassimilatedquickly,butdoesn'tnecessarilygiveyoualltheinformationyoumight needtothoroughlyrespondtoanyproblemsoropportunitiesthatarerevealed. Afulldiagnosistodeterminehowtorespondtothedatagleanedfromadashboardoftenrequires additionalinformation.Thisisasitshouldbe,becauseadashboardthattriedtogiveyoueverythingyou needtodoyourjob,includingallthedetails,wouldbeunreadable.Instead,dashboardsshouldprovidea broadandhighleveloverviewthatinformsyouinstantlyaboutthestateofthings.Iftheygofurtherby providingquickandeasyaccesstotheadditionalinformationthatyoumightneed,that'swonderfulbut thatjourneytakesyoubeyondthedashboarditself. 5.1.1.CondensingInformationviaSummarizationandException Thebestwaytocondenseabroadspectrumofinformationtofitontoadashboardisintheformof summariesandexceptions.Summarizationinvolvestheprocessofreduction.Summariesrepresentasetof numbers(oftenalargeset)asasinglenumber.Thetwomostcommonsummariesthatappearon dashboardsaresumsandaverages.Measuresofdistributionandcorrelationaresometimesappropriate, butthesearerelativelyrare. Giventhepurposeofadashboardtohelppeoplemonitorwhat'sgoingon,muchoftheinformationit presentsisnecessaryonlywhensomethingunusualishappening;somethingthatfallsoutsidetherealmof
1

HenryDavidThoreau,Walden(originallypublishedin1864).

www.it-ebooks.info

normality,intotherealmofproblemsandopportunities.Whymakesomeonewadethroughhundredsof valueswhenonlyoneortworequireattention?Wecallthesecriticalvaluesexceptions. Thebestdashboardsaredesignedtospecificallyaddressinformationneedsrelatedtoaparticularobjective orsetofobjectives.Notonlyshouldtheinformationbenarrowedtowhatdirectlyapplies,butthe communicationofthatinformationshoulduseitsaudience'svocabulary.Youwouldn'texpressthe relationshipbetweenthecostsofmarketingandresultingrevenuesasalinearcorrelationcoefficientifthe audiencehadnoideawhatthatwasorhowtomakesenseofit.Afamiliargraphwoulddoabetterjob. Likewise,youwouldn'tbreakthedataintomonthsiftheaudiencewerecomposedofsalesmanagerswho thinkentirelyintermsofweeks.Customizationisvitaltothesuccessofadashboard. Anaspectofcustomizationthatisoftenoverlookedinvolvesexpressingquantitativedataatalevelof precisionthatisappropriatetothetaskathand.Thegreaterthenumericprecision,themoretimeitwill takeviewerstoabsorbthedata.Whenexaminingfinancials,mostexecutivesrarelyneedtoseenumbers downtothelevelofcentsorevenbeyondthenearestthousand,tenthousand,hundredthousand,oreven million,butthemanagerofaccountingmightneedtoseeeverypenny. Displaymediamustbedesignedtosayexactlywhattheyneedtosaynomoredirectly,clearly,andwithout anyformofdistraction,inawaythatcommunicatesthemaximummeaningintheminimumamountof space.Ifadisplaymechanismthatlookslikeafuelgauge,thermometer,ortrafficsignalcommunicatesthe necessaryinformationinthismanner,thenthat'swhatyououghttouse.If,however,itfailsanyofthese tests,itoughttobereplacedwithsomethingthatdoesthejobbetter.Insistingoncutedisplayswhenother meanswouldworkbetteriscounterproductive,evenifeveryoneseemstobeinlovewiththem.Thisloveis fickle.Theappealofcutenesswillfadequickly,andtheonlythingthatwillmatterthenishowwellthe displaydeviceworks:howefficientlyandeffectivelyitcommunicates. Twofundamentalprinciplesshouldguidetheselectionoftheidealdashboarddisplaymedia: Itmustbethebestwaytodisplayaparticulartypeofinformationthatiscommonlyfoundin dashboards. Itmustbeabletoserveitspurposeevenwhensizedtofitintoasmallspace.

Inthenextchapter,we'llexamineanideallibraryofdashboarddisplaymediathatfulfillthese requirements.Fornow,let'sexaminesomedesignprinciples.

5.2.KeyGoalsintheVisualDesignProcess
EdwardR.Tufteintroducedaconceptinhis1983classicTheVisualDisplayofQuantitativeInformationthat hecallsthe"datainkratio."Whenquantitativedataisdisplayedinprintedform,someoftheinkthat appearsonthepagepresentsdata,andsomepresentsvisualcontentthatisnotdata(a.k.a.nondata). Figure52showstwodisplaysofquantitativedata:oneintheformofatableandtheotherintheformofa graph.Takeaminutetoexaminethemandtrytodifferentiatethedatainkfromthenondataink.

www.it-ebooks.info

Figure52.Thistableandgraphconsistofbothdatainkandnondataink.

Thereisn'tmuchnondatainkineitherthetableorthegraph,becausetheywereintentionallydesignedto keepittoaminimum.Figure53showsthesametableandgraph,thistimewiththenondatainkencoded asred.

Figure53.Here,thenondatainkishighlightedinred.

Tuftedefinesthedatainkratiointhefollowingway: Alargeshareofinkonagraphicshouldpresentdatainformation,theinkchanging asthedatachange.Datainkisthenonerasablecoreofagraphic,thenon redundantinkarrangedinresponsetovariationinthenumbersrepresented.Then, Datainkratio =dataink/totalinkusedtoprintthegraphic =proportionofagraphic'sinkdevotedtothenonredundantdisplayofdata information =1.0proportionofagraphicthatcanbeerasedwithoutlossofdatainformation.1 Hethenappliesitasaprincipleofdesign:"Maximizethedatainkratio,withinreason.Everybitofinkona graphicrequiresareason.Andnearlyalwaysthatreasonshouldbethattheinkpresentsnewinformation."2 Thisprincipleappliesperfectlytothedesignofdashboards,withonesimplerevision:becausedashboards arealwaysdisplayedoncomputerscreens,I'vechangedtheword"ink"to"pixels."Acrosstheentire dashboard,nondatapixelsanypixelsthatarenotusedtodisplaydata,excludingablankbackground shouldbereducedtoareasonableminimum.TakeamomenttoexaminethedashboardinFigure54on thenextpageandtrytoidentifythenondatapixelsthatcanbeeliminatedwithoutsacrificinganything meaningful.
1 2

EdwardR.Tufte,TheVisualDisplayofQuantitativeInformation(Cheshire,CT:GraphicsPress,1983),93. Ibid.,96.

www.it-ebooks.info

Figure54.Thisdashboarddisplaysanexcessiveamountofnondatapixels.

Thenondatapixelsthatyoucouldeasilyeliminatewithoutanylossofmeaninginclude: Thethirddimensionofdepthonallthepiechartsandonthebarsintheupperbargraph Thegridlinesinthebargraphs Thedecorationinthebackgroundoftheupperbargraph Thecolorgradientsinthebackgroundsofthegraphs,whichvaryfromwhiteatthetopthrough shadesofblueastheyextenddownward

Someofthedatapixelsonthisdashboardcouldalsoberemovedwithoutalossofusefulmeaningwe'll comebacktothatinamoment. Reducingthenondatapixelstoareasonableminimumisakeyobjectivethatplacesusonthepathto effectivedashboarddesign.Muchofvisualdashboarddesignrevolvesaroundtwofundamentalgoals: 1. Reducethenondatapixels. 2. Enhancethedatapixels.

www.it-ebooks.info

Youstartbyreducingthenondatacontentasmuchaspossible,andthenproceedtoenhancethedata contentwithasmuchclarityandmeaningaspossible,workingtomakethemostimportantdatastandout abovetherest(Figure55).

Figure55.Keygoalsandstepsofvisualdashboarddesign.

5.2.1.ReducetheNonDataPixels Thegoalofreducingthenondatapixelscanbebrokendownintotwosequentialsteps: 1. Eliminateallunnecessarynondatapixels. 2. Deemphasizeandregularizethenondatapixelsthatremain. Let'stakealookathowtoaccomplishthesetwogoals. 5.2.1.1.Eliminateallunnecessarynondatapixels Dashboarddesignisusuallyaniterativeprocess.Youbeginbymockingupasampledashboard,andthen youimproveitthroughaseriesofredesigns,eachfollowedbyafreshevaluationleadingtoanother redesign,untilyouhaveitright.Asyougetbetterandbetteratthis,thenumberofiterationsthatwillbe requiredwilldecrease,partlybecauseyouwon'tbeincludingunnecessarynondatapixelsinthefirstplace. Nomatterhowfaryouadvance,however,thestepoflookingforunnecessarynondatapixelswillnever ceasetobeproductive. Thenextfewfiguresprovideexamplesofnondatapixelsthatoftenfindtheirwayontodashboardsbutcan usuallybeeliminatedwithoutloss. Graphicsthatservemerelyasdecoration(Figure56).

Figure56.Youshouldeliminategraphicsthatprovidenothingbutdecoration.

Variationsincolorthatdon'tencodeanymeaning(Figure57).

Figure57.Thesebarsvaryincolorfornomeaningfulreason.

www.it-ebooks.info

Bordersthatareusedtodelineatesectionsofdatawhenthesimpleuseofwhite/blankspacealonewould workaswell(Figure58).

Figure58.Unnecessarybordersaroundsectionsofdatafragmentthedisplay.

Fillcolorsthatareusedtodelineatesectionsofcontentsuchasatitle,thedataregionorlegendofagraph, thebackgroundofatable,oranentiresectionofdata,whenaneutralbackgroundwouldworkaswell (Figure59).

Figure59.Fillcolorstoseparatesectionsofthedisplayareunnecessary.

Gradientsoffillcolorwhenasolidcolorwouldworkaswell(Figure510).

www.it-ebooks.info

Figure510.Gradientsofcolorbothonthebarsofthisgraphandacrosstheentirebackgroundadddistractingnondatapixels.

Gridlinesingraphs(Figure511).

Figure511.Gridlinesingraphsarerarelyuseful.Theyareoneofthemostprevalentformsofdistractingnondatapixelsfound indashboards.

Gridlinesintables,whichdividethedataintoindividualcellsordivideeithertherowsorthecolumns, whenwhitespacealonewoulddothejobaswell(Figure512).

Figure512.Gridlinesintablescanmakeotherwisesimpledisplaysdifficulttolookat.

www.it-ebooks.info

Fillcolorsinthealternatingrowsofatabletodelineatethemwhenwhitespacealonewouldworkaswell (Figure513).

Figure513.Fillcolorsshouldbeusedtodelineaterowsinatableonlywhenthisisnecessarytohelpviewers'eyestrackacross therows.

Completebordersaroundthedataregionofagraphwhenonehorizontalandoneverticalaxiswould sufficientlydefinethespace(Figure514).

Figure514.Acompleteborderaroundthedataregionofagraphshouldbeavoidedwhenasinglesetofaxeswouldadequately definethespace.

3Dingraphswhenthethirddimensiondoesn'tcorrespondtoactualdata(Figure515).

Figure515.3Dshouldalwaysbeavoidedwhentheaddeddimensionofdepthdoesn'trepresentactualdata.

Visualcomponentsorattributesofadisplaymediumthatservenopurposebuttomakeitlookmorelikea realphysicalobjectormoreornate(Figure516).

www.it-ebooks.info

Figure516.Thisdashboardisfilledwithvisualcomponentsandattributesthatservethesolepurposeofsimulatingrealphysical objects.

Thisisbynomeansacomprehensivelist,butitdoescovermuchofthenondatacontentthatIroutinely runacrossondashboards.Whenyoufindthatyou'veincludeduselessnondatapixelssuchasthoseinany oftheaboveexamples,simplyremovethem. 5.2.1.2.Deemphasizeandregularizethenondatapixelsthatremain Notallnondatapixelscanbeeliminatedwithoutlosingsomethinguseful.Somesupportthestructure, organization,orlegibilityofthedashboard.Forinstance,whendataistightlypacked,sometimesitis necessarytouselinesorfillcolorstodelineateonesectionfromanother,ratherthanwhitespacealone.In thesecases,ratherthaneliminatingtheseusefulnondatapixels,youshouldsimplymutethemvisuallyso theydon'tattractattention.Focusshouldalwaysbeplacedontheinformationitself,notonthedesignof thedashboard,whichshouldbealmostinvisible.Thetrickistodeemphasizethesenondatapixelsby makingthemjustvisibleenoughtodotheirjob,butnomore. Beginningonthenextpageareafewexamplesofnondatapixelsthatareeitheralwaysoroccasionally useful.I'veshowneachoftheseexamplesintwoways:1)aversionthatistoovisuallyprominent,which illustrateswhatyoushouldavoid;and2)aversionthatisjustvisibleenoughtodothejob,whichisthe objective. Axislinesthatareusedtodefinethedataregionofagraph(Figure517).

www.it-ebooks.info

Figure517.Axislinesusedtodefinethedataregionofagrapharealmostalwaysuseful,buttheycanbemuted,likethoseon theright.

Lines,borders,orfillcolorsthatareusedtodelineatesectionsofdatawhenwhitespaceisnotenough (Figure518).

Figure518.Linescanbeusedeffectivelytodelineateadjacentsectionsofthedisplayfromoneanother,buttheweightofthese linescanbekepttoaminimum.

Gridlinesingraphswhennecessarytoreadthegrapheffectively(Figure519).

Figure519.Gridlinesareusefulwhentheyhelpviewerscomparespecificsubsectionsofgraphs,suchastherangeofvaluesthat fallwithin65to75ontheverticalscaleand35,000to45,000onthehorizontalscale.

Gridlinesand/orfillcolorsintableswhenwhitespacealonecannotadequatelydelineatecolumnsand/or rows(Figure520).

www.it-ebooks.info

Figure520.Gridlinesandfillcolorscanbeusedintablestoclearlydistinguishsomecolumnsfromothers,butthisshouldbe doneinthemutedmannerseenbelowratherthantheheavyhandedmannerseenabove.

Fillcolorsinthealternatingrowsofatablewhenwhitespacealonecannotadequatelydelineatethem (Figure521).

Figure521.Fillcolorscanbeusedtodelineaterowsinatablewhennecessarytohelpviewers'eyesscanacrosstherows,but thisshouldalwaysbedoneinthemutedmannerseenbelowratherthanthevisuallyweightymannerseenabove.

Theseexamplesdemonstratehowthevisualprominenceofnondatapixelscanusuallybedeemphasized byusinglight,lowlysaturatedcolors,suchaslightgrays,andminimalstrokeweights(thatis,thinlines). Nondatapixelsalsocanbepushedfurtherfromnoticebyregularizingthem(thatis,bymakingthem consistent).Iftheaxislinesofallgraphslookthesamesay,ifyouusethesamelightgraylineswherever theyappearnoonegraphislikelytocatchaviewer'seyesmorethantheothers.Differencesseldomgo unnoticed,evenwhentheyareexpressedinmutedtones.Don'tvarythecolor,weight,orshapeofnon datapixelsthatservethesamepurposeinthedashboard. Anothercategoryofcontentoftenfoundondashboardsthatcanbeconsiderednondatapixelsisthat whichsupportsnavigationanddataselection.Buttonsandselectionboxesareoftenusedtoallowusersto navigatetoanotherscreenortochoosethedatathatappearsonthedashboard(forexample,byselecting adifferentsubset,suchashardwareratherthansoftware).Theseelementsmightserveanimportant function,buttheydon'tdisplaydata.Assuch,theyshouldnotbegivenprominence.Iftheymustexist,

www.it-ebooks.info

placetheminanoutofthewaylocationsuchasthebottomrightcornerofthescreenandmutethem visually,sotheywon'tcompetewiththedataforattention.NoticehowmuchofthedashboardinFigure5 22isdedicatedtobuttonsanddataselectioncontrols,whichI'vehighlightedwithredborders.These elementstakeupfarmorevaluableandprominentrealestateonthedashboardthanisrequired.

Figure522.Thisdashboardgivesnavigationalanddataselectioncontrolsfarmoredominanceandspacethantheydeserve.

Similarly,whileitmaysometimesbenecessarytoincludeonthedashboardinstructionsthatprovide importantsupportinformation,anynonessentialtextjusttakesupspacethatcouldbeusedbydata, attractsattentionawayfromthedata,andcluttersthedashboard'sappearance.Itusuallyworksbestto placemostinstructionalordescriptivecontenteitheronaseparatescreenthatcaneasilybereachedwhen neededor,ifpossible,intheformofpopupsthatcanbeaccessedwhennecessarywithaclickofthe mouse.NoticehowmuchprimerealestateiswastedonthedashboardinFigure523toprovide instructionsthatviewerswillprobablyonlyneedthefirsttimetheyusethedashboard.

www.it-ebooks.info

Figure523.Asyoucanseeintheareahighlightedinred,thisdashboardusesupvaluablespacetodisplayinstructionsthat couldhavebeenprovidedonlywhenneededthroughaseparatescreenorapopupmenu.

5.2.2.EnhancetheDataPixels Justlikethereductionofnondatapixels,theprocessofenhancingthedatapixelscanbebrokendowninto twosequentialsteps: Eliminateallunnecessarydatapixels. Highlightthemostimportantdatapixelsthatremain.

Let'sexaminethesetwotasks. 5.2.2.1.Eliminateallunnecessarydatapixels Whenyou'redesigningadashboard,itistemptingtothroweverythingyouthinkanyonecouldever possiblywantontoit.Thoseofuswhohaveworkedinthefieldofbusinessintelligenceforawhilehave grownwearyofbeingaskedformore(alwaysmore!),sothethoughtofheadingoffthisdemandbygiving folkseverythingupfrontcanbeappealing.Onadashboard,however,whereimmediateinsightisthegoal, thisisacostlymistake.I'mnotsuggestingthatyouforcepeopletogetbywithlessthantheyreallyneed, butratherthatyouhonortheconsiderationofwhattheyreallyneedforthetaskathandasastrict

www.it-ebooks.info

criterionfortheselectionofdata.Byremovinganyinformationthatisn'treallynecessary,you automaticallyincreasefocusontheinformationthatremains. Eliminationofunnecessarydatapixelsisachievednotonlythroughthecompleteremovaloflessrelevant databutalsobycondensingdatathroughtheuseofsummariesandexceptions,sothatthelevelofdetail thatisdisplayeddoesn'texceedwhat'snecessary.Formostapplications,itwouldbeabsurdtoinclude detailedinformationsuchastransactionlevelsalesdataonadashboardsomelevelofsummarizationis needed,anditisoftenuptoyoutodeterminewhatthatlevelis.Youmightchoosetodisplayasingle quartertodatevalue,avalueperregion,oravaluepermonth,justtonameafewpossibilities. Exceptionsareanespeciallyusefulmeanstoreducethedataonadashboardtowhatisessentialforthe taskathand.Often,thestateofsomethingneednotbepresentedunlessthereisaproblemoran opportunitythatrequiresaction.Ifyoucareaboutstaffexpensesonlywhensomeonehasexceededa definedthreshold,whyclutterthedashboardwithacompletelistofallstaffmembersandtheirexpenses? Bewareoftakingthisusefulpracticeofmanagingbyexceptiontoofar,however.Ireceivedanemail recentlyfromanexecutiveofasoftwarecompanythatspecializesindashboards.Wewerediscussingmy definitionofadashboard,andinthecourseofthisdiscussionhestatedthatacustomeronceassertedthat hisidealdashboardwoulddisplayasingletrafficsignaltoindicateifeverythingwasallrightorifanything neededattention.Theideawasthathedidn'twanttobebotheredwithunnecessaryinformationifallwas well,andwhensomethingwaswrong,hecoulddrilldownfromthatsinglealerttoadditional,more detaileddashboardsorreportstodetermineexactlywhatwaswrongbeforetakingaction.ForaninstantI foundmyselfenamoredwiththisidea,attractedtoitsSpartansimplicitybutonlyforamoment.Thenext momentmymindbecamehauntedbyvisionsofexecutivestryingtoruntheirbusinessesinignorantbliss, completelyoutoftouchunlessthresholdsbuiltintothesoftwaredeterminedthattheyoughttobe informed.Anyonewhohasajobtodoneedstokeepupwithabasicpictureofwhat'sgoingon,evenwhen alliswell.Toooftenleaderswhetherinbusiness,academia,religion,orpoliticsforgeaheadwiththeir agendas,relyingentirelyonotherstotellthemwhattheythinktheyshouldknow,onlytodiscoverafterthe dustofsomedestructiveeventsettlesthattheyknewfartoolittletoleadeffectively. Beforedepartingfromthetopicofsummariesandexceptions,Iwanttofocusinonaparticular summarizingtechniquethatIfindusefulonoccasion.ThistechniqueinvolveswhatIcallmultifocidisplays. Whenitisusefultodisplayhistoricalcontextforameasure,suchasthelast12monthsorthelast5years, ofteninformationthatismoredistantfromthepresentislessimportantthanrecenthistory.Insuchcases, thereisnoreasontodisplaythefullrangeofdataatthesamelevelofdetail.Forinstance,youmightwant todisplaythecurrentmonthasdailymeasures,thepreceding12monthsasmonthlymeasures,andthe preceding4yearsasannualmeasures.Thisdisplaywouldconsistofthreesections,eachexpressedin differentintervalsoftime,withlongerintervalsandmoresummarizationusedfortheperiodthemost distantfromthepresent.Graphicdisplayscanbedesignedtopresenttimeseriesinthismanner,as illustratedinFigure524.

www.it-ebooks.info

Figure524.Thesethreetimeseriesgraphsdisplayingpublictransportationriderstatisticscontainthreelevelsofdetail:dailyfor thecurrentmonth,monthlyforthecurrentyear,andyearlyforthelast10years.

Varyinginterestcancorrespondtodistancesinspaceaswellastime.Forinstance,aviewermightbemost interestedindatafromhisimmediategeographicalregion,andgraduallylessinterestedindatafrom increasinglydistantgeographicalareas.1 5.2.2.2.Highlightthemostimportantdatapixelsthatremain Alltheinformationthatfindsitswayontoadashboardshouldbeimportant,butnotalldataiscreated equal:somedataismoreimportantthanotherdata.Themostimportantinformationcanbedividedinto twocategories: Informationthatisalwaysimportant Informationthatisonlyimportantatthemoment

Whenyouconsidertheentirecollectionofinformationthatbelongsonadashboard,youshouldbeableto prioritizeitaccordingtowhatisusuallyofgreatestinteresttoviewers.Forinstance,adashboardthat servestheneedsofacorporation'sexecutivesmightdisplayseveralcategoriesoffinancial,sales,and personneldata.Onthewhole,however,theexecutivesusuallycareaboutsomekeymeasuresmorethan others. Theothercategoryofespeciallyimportantinformationisthatwhichisimportantonlywhenitreveals somethingoutoftheordinary.Ameasurethathasfallenfarbehinditstarget,anopportunitythathasjust arisenandwon'tlastforlong,oranoperationalconditionthatdemandsimmediateattentionallfallinto thiscategory. Thesetwocategoriesofimportantinformationrequiredifferentmeansofhighlightingonadashboard.The firstcategoryinformationthatisalwaysimportantcanbeemphasizedusingstaticmeans,butthesecond categoryinformationthatisimportantonlyatthemomentrequiresadynamicmeansofemphasis. Thelocationofdataonthescreenthelayoutisanaspectofadashboard'sappearancethatdoesn't,orat leastshouldn't,changedynamically.Thisistruenotonlybecauseitwouldbetechnicallydifficultto dynamicallyrearrangetheplacementofdataonthescreen,butalsobecauseaftersomeuseviewerswill cometoexpectspecificdatatoappearinspecificlocations,whichisgoodbecauseithelpsthemtoscanthe

Multifocidisplaysarenotexclusivelyrelevanttodashboards.Ihaveafondnessforbeautifullyrenderedmaps,andI enjoyexploringgeographyandtracingmytravelsacrossthesurfaceofmaps.Infact,Ikeepthreemapsmountedon thewallsofmyoffice:anextremelylargeoneofCalifornia,myhomestate;aslightlysmalleroneoftheentireUnited States;andanevensmalleroneoftheentireworld.Thismightseemcountertothelogicalarrangement,becausethe worldiscertainlylargerthanCalifornia,butitservesmyneedsprecisely.Iwanttoseegreatdetailinplacescloseto home,whereIspendmostofmytime,andgraduallylessandlessdetailasthedistancefromhomegrows.


1

www.it-ebooks.info

dashboardquickly.Becauselocationisstatic,thisisavariablethatwecanleveragetohighlightinformation thatisalwaysimportant. Fewaspectsofvisualdesignemphasizesomedataabovetherestaseffectivelyasitslocation.Figure525 identifiestheemphasizingeffectthatdifferentregionsofadashboardprovide.Thetopleftandcenter sectionsofthedashboardaretheareasofgreatestemphasis.Thegreateremphasistiedtotheupperleftis primarilyduetotheconventionsofmostwesternlanguages,whichsequencewordsonapagefromleftto rightandtoptobottom.Contrarytotheinfluenceofreadingconventions,however,theverycenterofthe screenisalsoaregionofstrongemphasis,duetoamorefundamentalinclinationofvisualperception.I've found,however,thatplacinginformationinthecenterresultsinemphasisonlywhenitissetapart somewhatfromwhatsurroundsit,suchasthroughtheuseofwhitespace.

Figure525.Differentdegreesofvisualemphasisareassociatedwithdifferentregionsofadashboard.

Asmuchaspossible,placetheinformationthatisalwaysofgreatimportanceintheupperleftorcenter regionsofthedashboard.Neverwastethisvaluablerealestatebyplacingacompanylogoorcontrolsfor navigationordataselectionintheseareas.Figure526providesavividexampleofwhatyoushouldavoid whendesigningthelayoutofadashboard.

www.it-ebooks.info

Figure526.Themostvaluablerealestateonthisdashboardisdedicatedtoacompanylogoandmeaninglessdecoration.

Visualattributesotherthanlocationonthescreenareusuallyeasytomanipulateinadynamicmannerona dashboard.Assuch,dynamictechniquescanbeusedtohighlightinformationthatisofgreatimportance onlyatparticulartimes.Thesetechniquescanalsobeusedtohighlightinformationthatisalways important,onceyou'veuseduptheprimescreenlocationsforotherimportantdata. ManyofthevisualattributesthatweexaminedinChapter4,TappingintothePowerofVisualPerception, canbeusedeffectivelytohighlightdata,bothstaticallyanddynamically.Herearetwoapproachesthatyou cantake: Useexpressionsofvisualattributesthataregreaterthanthenorm(forexample,brighteror darkercolors). Useexpressionsofvisualattributesthatsimplycontrastwiththenorm(forexample,bluetext whenthenormisblackorgray).

Expressionsofvisualattributesdon'tneedtobegreaterthanotherstostandout;contrastfroma predominantpatternisallittakes.Visualperceptionishighlysensitivetodifferencesandevervigilantto assignmeaningtothemwhentheyaredetected. Someusefulexpressionsofvisualattributesthatareperceivedasgreaterthanothersincludethefollowing: Table51. Visual attribute Color Usefulexpressions Illustrations

Adarkerormorefullysaturatedversionofanyhueisnaturallyperceivedas

www.it-ebooks.info

intensity Size Linewidth

greaterthanalighterorlesssaturatedversion. Biggerthingsclearlystandoutasmoreimportantthansmallerthings. Thickerlinesstandoutasmoreimportantthanthinnerlines.

Someusefulexpressionsofvisualattributesthatstandoutmerelythroughcontrasttothenormincludethe following: Table52. Visual attribute Hue Orientation Enclosure Usefulexpressions Illustrations

Anyhuethatisdistinctfromthenormwillstandout. 1 Anythingorienteddifferentlythanthenormwillstandout. Anythingenclosedbybordersorsurroundedbyafillcolorwillstandoutif differentfromthenorm. Anythingwithsomethingdistinctlyaddedtoitoradjacenttoitwillstand out.

Addedmarks

Anyofthesevisualattributescanbeusedtomakethemostimportantinformationstandoutfromtherest. Colorisespeciallyusefulbecausedistinctdifferencesincolorstandoutveryclearlyandbecauseitisa variablethatisnormallyeasytochangedynamicallyusingdashboardsoftwarebasedonpredefineddata conditions. I'vealsofoundthatoneofthebestwaystodrawattentiontoparticularitems,especiallythoseexpressed astext,involvestheuseofanaddedmarkwithadistinctcolor.Forexample,causingasimplesymbolsuch asacircle,checkmark,orasterisktoappearnexttoitemsthatneedattentiondoesthejobnicely.Choosing onecolorandvaryingitsintensitytoindicatevaryingdegreesofimportanceorurgencyworksbetterthan usingdifferentcolors,becauseeventhosewhoarecolorblindcandetectdistinctintensitiesofthesame color.Figure527illustratesthispractice.Differentsymbolscouldalsobeusedtoindicatedifferentlevels ofimportanceorurgencywithnoneedtovarytheircolors,butincreasingcolorintensitiescorresponding toincreasinglevelsofimportanceorurgencyareunderstoodmoreintuitively.

Reddoesnotsignifythatsomethingisimportant,urgent,oraprobleminallcultures.Forexample,inChina,red connoteshappiness.Bearinmindalsowhenchoosingsymboliccolorsthatasignificantchunkofthepopulationis colorblind.


1

www.it-ebooks.info

Figure527.Simplesymbolscanbeusedalongwithvaryingcolorintensitiestodynamicallyhighlightdata.

Whenhighlightingimportantinformation,youmustalwaysbecarefultorestrictthedefinitionofwhat's important.Ifyouhighlighttoomuchinformation,nothingwillstandoutandyourefforttocommunicate willfail.Whenusedwithdiscretion,however,visualhighlightingcanachievethegoalofimmediate recognitionandquickresponse.

www.it-ebooks.info

Chapter6.EffectiveDashboardDisplayMedia
Dashboardsmustbeabletocondensealotofinformationontosinglescreenandpresentitataglance withoutsacrificinganythingimportantorcompromisingclarity.Consequently,theyrequiredisplaymedia thatcommunicateeffectively,despitetheseconditions.Everysectionofdataonadashboardshouldbe displayedusingtheclearestandrichestpossiblemeans,usuallyinsmallamountofspace.Thisrequiresan availablelibraryofdisplaymediathathavebeenselected,customized,andsometimescreatedespeciallyfor dashboards,andanunderstandingofthecircumstancesinwhicheachmediumofdisplayshouldbeapplied.

www.it-ebooks.info

Selectthebestdisplaymedium Anideallibraryofdashboarddisplaymedia Adashboardmustbebuiltusingappropriatelychosenanddesigneddisplaymediatoreachitsunique potentialforclearandimmediatecommunication.We'llbeginthischapterwithsomebasicguidelinesfor matchingyourdataandmessagetotherightformofdisplay,andthenproceedtotheheartofthechapter: thedescriptionofafulllibraryofdisplaymediathatareidealfordashboards.

6.1.SelecttheBestDisplayMedium
Thebestmediumfordisplayingdatawillalwaysbebasedonthenatureoftheinformation,thenatureof themessage,andtheneedsandpreferencesoftheaudience.Asingledashboardgenerallydisplaysa varietyofdataandrequiresavarietyofdisplaymedia,eachmatchedtospecificdata.Inthenextsection we'llpairspecificdataandmessageswiththegraphicmediathatdisplaythembest,butlet'sbeginhere withamorefundamentalquestion:"Shouldtheinformationbeencodedastext,graphics,orboth?"The appropriatenessofeachmediumforagivensituation,eitherverballanguageinwrittenform(text)orvisual language(graphics),isn'tarbitrary. Verballanguageisprocessedserially,onewordatatime.Somepeoplearemuchfasterreadersthan othersanabilitythatIenvybuteveryoneprocesseslanguageserially.Especiallywhencommunicating quantitativeinformation,thestrengthofwrittenwordsandnumberscomparedtographicsistheir precision.Ifyoursolepurposeistopreciselycommunicatecurrentyeartodateexpensesof$487,321,for example,nothingworksbetteronadashboardthanasimpledisplaylikethis: Table61. YTDExpenses

$487,321

Displayingindividualvaluesdoesnotrequiregraphicsindeed,theirusewouldonlyretardcommunication. Let'scontinuetoenhancethisdatatoseeifthereisapointwhereswitchingfrompuretexttotheaddition ofgraphicsaddsclearvalue. Sometimesjustprovidinganindividualnumberandlabelisappropriate,butoftenyouwanttosaymore. Let'senhancethedatawithasimpleevaluativeremarkthatthisyeartodateexpensefigureishigherthan itshouldbe: Table62. YTDExpenses

$487,321

Thiscertainlyisn'ttheonlywaytocommunicatethisevaluativeinformation,butitissufficient.Aslongas onlymeasuresinthisconditionaredisplayedinthisfashion,eventhosewhoarecolorblindwillbeableto recognizethatwearecallingattentiontothisexpenseamount(becausewe'veboldfacedthenumber). Nowlet'saddtothegeneraldeclarationthatthisexpenseamountisbadthespecificcriterionthatwas usedtodeterminethis,whichinthiscaseisthetargetforyeartodateexpenses: Table63. YTDExpenses

Actual $487,321

Target $450,000

www.it-ebooks.info

Atthisstagewe'rebeginningtoventureintotheterritorywhereagraphicaldisplaymightbeuseful,butit certainlyisn'timperativeyet.Theviewermustdoalittlemathtointerprettheextentoftheexpense overage,butinthiscasethemathissimpleandfast.Youcouldevenremovetheneedfortheviewertodo thecalculationbyaddingtheamountofvariancefromthetarget,orperhapsbydisplayingthevariance alone,withouttheactualexpenseamount,ifthevarianceisallthat'sneeded.Herearesomeexamplesof howyoucouldchoosetopresentthisdata,usingtextalone: Table64. YTDExpenses Table65. YTDExpenses Table66. YTDExpenses Table67. YTDExpenses YTDExpensesVariance YTDExpensesVariance

Actual $487,321

Target $450,000

Varience +$37,321

Actual $487,321

Target $450,000

Varience% +$8%

Actual $487,321

VariancetoTarget +$37,321

Actual $487,321 +$37,321 +8%

VariancetoTarget% +8%

Anyoneoftheseapproachesmightbeappropriateforasinglemeasurethathasbeenenhancedwith contextualdatasuchasthetargetandsomeindicationofwhetheritisgoodorbad. Anentiredashboardfullofindividualmeasuresexpressedtextuallyinthismannerwouldworkfineifits purposeweretodrawattentiontoindividualmeasuresoneatatime,butwhatifyouwantabiggerpicture ofthewholeorcomparisonsofmultiplemeasurestoemerge?Textalonedoesn'tsupportthis. Text,especiallywhenorganizedintotables(thatis,asrowsandcolumnsofdata),isasuperbmediumfor lookingupinformation.Busschedules,taxratetables,andtheindexesofbooks,tonamebutafew examples,areallorganizedastablestosupportthisuse.IfyouneedtolookuptheConsumerPriceIndex (CPI)rateforSeptember1996usingthetableinFigure61,forexample,youcaneasilyfindtheprecise valueof157.8.Graphsdon'tsupportlookingupindividualvaluesasefficiently,andcertainlynotas precisely.

www.it-ebooks.info

Figure61.ThisCPItableillustratesthestrengthoftablesasameanstolookuppreciseindividualvalues.

NowlookattheCPItableagain,butthistimetrytodeterminetheshapeofthevaluesastheychange throughthecourseoftheyear1996.Textdoesn'tsupportthisviewofthedata,butlookathowclearlythe graphinFigure62onthenextpagepresentsit.

Figure62.ThisgraphoftheCPIfortheyear1996illustrateshowwellgraphsrevealtheshapeofdata,inthiscaseasitchanges throughtime.

Noticealso,however,thattheprevioustaskoflookinguptheindexvalueforSeptemberisnotsupported verywellbythegraph. When,inthelate18thcentury,theBritishsocialscientistWilliamPlayfairinventedmanyofthegraphsthat westillusetoday,hecreatedapowerfullanguageforcommunicatingquantitativeinformation.Giving valuesshapethroughtheuseofgridcoordinatesalongtwoaxesenabledustovisualizenumbers,which dramaticallyextendedourabilitytothinkquantitatively.Thisisthestrengthofgraphs:theygiveshapeto numbersand,indoingso,bringtolightpatternsthatwouldotherwiseremainundetected. Let'sseesomeoftheseconceptsatworkonadashboard.Lookatthepredominantlytextbaseddashboard inFigure63.

www.it-ebooks.info

Figure63.Apredominantlytextbaseddashboard.

Noticehowthetextualmediumprimarilysupportstheprocessoflookup.Eachmeasureisisolatedfromthe rest,andcomparisonsaredifficult. Theonlybigpictureinformationthatisprovidedisconveyedthroughthevisualattributeofhue.Assuming thatyouarenotcolorblindandcandistinguishthesehues,withaquickscanthemanyredandyellow boxesrevealthatmuchiswrong.Beyondthat,youareforcedtoconsidereachmeasureindividually.Ifno comparisonsorpatternsareusefulforthisdashboard,thepredominanceoftextisfine.Butevenifthis werethecase,whichisunlikely,thetextualdisplayofthisinformationcouldhavebeenpresentedinaless fragmentedway,suchastheredesignthatyouseeinFigure64.Here,themeasuresarearrangedintables tomakescanningeasier.Thered,yellow,andgreencolorcodinghasbeenreplacedwithboldface,black, andgraytext,respectively,toenableperceptionbypeoplewhoarecolorblind.Notethatthisredesignhas improvedthedashboard'suseforlookup,butnotforgleaningadditionalmeaning.

www.it-ebooks.info

Figure64.RedesignofthetextbaseddashboardinFigure63,arrangedintablestobettersupportlookup.

Effectivedashboardsneedtocombinetextandgraphicsinawaythatsupportsarichandmeaningful displayofdata,alongwiththedesiredlevelofquantitativeprecision,inawaythatcanbeperceived efficiently.Witheachmeasureorsetofrelatedmeasures,youmustaskwhattheviewerneeds,howthe datawillbeused,andwhatmessagethedatamustconvey,andthenblendtheuseoftextandgraphicsto achievethesecommunicationobjectives.

6.2.AnIdealLibraryofDashboardDisplayMedia
Sofarwe'veconsideredonlythefirst,mostfundamentalstepinselectingthebestmediumofdisplay.Once you'vechosenbetweentext,graphics,orsomecombinationofthetwo,youmustthendeterminehowto organizethetextand/orwhatkindsofgraphicstouse.Thesechoicesarevital.Apoorlychosengraph,for example,couldcompletelyobscureotherwisecleardata.Inthissection,we'llfocusspecificallyonthebest choiceofgraphicaldisplaytousewhenyoudeterminethatavisualratherthanatextualdisplayis appropriate. Mostdisplaymediathatworkwellondashboardsareprobablyfamiliartoyoualready.Quantitativegraphs andseveralothertypesofchartsthatarecommonlyusedinbusinessreporting(forexample,processflow andorganizationcharts)workwellondashboards,providedtheirdesigniskeptclearandsimple. Thisdiscussionfocusesondashboarddisplaymediathatareusedtopresentactualdata.Otherdisplay media,suchascommandbuttons,aresometimesneeded,buttheyfalloutsideourscopeofinterest.Two fundamentalprincipleshaveguidedtheselectionofeachdisplaymediuminthisproposedlibrary: Itmustbethebestmeanstodisplayaparticulartypeofinformationthatiscommonlyfound ondashboards. Itmustbeabletoserveitspurposeevenwhensizedtofitintoasmallspace.

Thelibraryisdividedintosixcategories: Graphs Images Icons

www.it-ebooks.info

Drawingobjects Text Organizers

6.2.1.Graphs Mostdashboarddisplaymediafallintothegraphcategory.Giventhepredominanceofquantitativedataon mostdashboards,thisisn'tsurprising.Allbutoneoftheitems(treemaps)inthiscategorydisplay quantitativedataintheformofa2DgraphwithXandYaxes.Mostofthesearefamiliarbusinessgraphs, butoneortwowillprobablybenewtoyou,becausetheyweredesignedoradaptedspecificallyforusein dashboards.Here'sthelist: Bulletgraphs Bargraphs(horizontalandvertical) Stackedbargraphs(horizontalandvertical) Combinationbarandlinegraphs Linegraphs Sparklines Boxplots Scatterplots Treemaps

6.2.1.1.Bulletgraphs Thisistheonegraphonthelistthatisalmostcertainlynewtoyou.Iassumethisbecauseabulletgraphisa simpleinventionofmyown,createdspecificallyfordashboards.Itismyanswertotheproblemsexhibited bymostofthegaugesandmetersthathavebecomesynonymouswithdashboards.Gaugesandmeters typicallydisplayasinglekeymeasure,sometimescomparedtoarelatedmeasuresuchasatarget,and sometimesinthecontextofquantitativerangeswithqualitativelabelsthatdeclarethemeasure'sstate (suchasgoodorbad).Figure65providestwoexamplesofthegaugesandmetersthatarecommonly foundondashboards.Bothdisplayakeymeasureincomparisontoatarget,whichisrepresentedbyzero onthegaugeontherightand,Iassume,bythetopofthethermometerontheleft.

www.it-ebooks.info

Figure65.Thesearetypicalexamplesofmetersandgaugeswithcontextualdata.1

Thequestionthatyoushouldaskwhenconsideringgaugesandmeterssuchastheseis:"Dotheyprovide theclearest,mostmeaningfulpresentationofthedataintheleastamountofspace?"Inmyopinion,they donot.RadialgaugessuchastheexampleontherightinFigure65wasteagreatdealofspace,dueto theircircularshape.Thisproblemismagnifiedwhenyouhavemanyradialdisplaymechanismsonasingle dashboard,fortheycannotbearrangedtogetherinacompactmanner.Thelinearnatureofthe thermometerstyleofdisplaypotentiallyavoidsthisproblem,butindisplayssuchasthis,spacetendstobe wastedonmeaninglessrealism.Ifdashboarddisplaymediaweredesignedbyexpertcommunicators, ratherthanbygraphicartistswhoclearlyhaven'tfocusedonthecommunicationneeds,theywouldlook muchdifferent. Thebulletgraphachievesthecommunicationobjectivewithouttheproblemsthatusuallyplaguegauges andmeters.Itisdesignedtodisplayakeymeasure,alongwithacomparativemeasureandqualitative rangestoinstantlydeclareifthemeasureisgood,bad,orinsomeotherstate.Figure66providesasimple example.

Figure66.Asimplehorizontallyorientedbulletgraph.

Now,Iamwellawarethatitsoundsabittoohighandmightyformetocallthebulletgraphmyinvention. It'snotmuchmorethanabargraphwithasinglebar,orathermometerwithoutthereservoirattheendto holdthemercurywhileatrest.Simpleasitis,whyhasn'tanyoneelsecomeupwiththisideabefore?Any softwarevendorwhowantstouseitcanbemyguest,freeofcharge.I'llevensupplythedesign specification.Figure67showsthesamebulletgraph,thistimewitheachofitscomponentsidentified.

Figure67.Asimplebulletgraphwitheachofitscomponentslabeled.

Thelineardesignofthebulletgraph,whichcanbeorientedeitherhorizontallyorvertically,allowsseveral tobeplacednexttooneanotherinarelativelysmallspace.Figures68and69showhowcloselytheycan bepackedtogetherimaginehowmuchroomwouldberequiredtodisplaythesamedatausingcircular gauges.

CanyoumakesenseofthethermometerontheleftinFigure65?Dosalesincreaseastheyriseorastheyfallonthe thermometer?Giventhefactthatactualsalesare75.93%oftargetandthemercuryinthethermometerextends about75%ofthewaytothetopofthethermometer,wemustassumethatsalesriseasthemercuryrises,butthen, asredonadashboardusuallymeansbad,whyistheredrangeatthetop?


1

www.it-ebooks.info

AsyouscanacollectionofbulletgraphssuchasthoseinFigures68and69,noticehoweasyitistodetect thosemeasuresthathavemetorexceededthecomparativemeasuresrepresentedbytheshortlinethat intersectseachbar.Whenameasureexceedsthisbar,acrossshapeisformed.Thisformiseasytosee becauseitisperceivedpreattentively.Youcanscanthebulletgraphsonadashboardandimmediately knowwhichmeasuresaredoingwellandwhicharenotsimplybythepresenceorabsenceofthesecross shapes.

Figure68.Acollectionofhorizontallyorientedbulletgraphs.

Figure69.Acollectionofverticallyorientedbulletgraphs.

Noticealsothatthebackgroundfillcolorsthatencodethequalitativecategories(suchasbad,satisfactory, andgood)arevariablesofcolorintensityratherthanofhue.Thisassuresthatviewerswhoarecolorblind

www.it-ebooks.info

canstillseethedistinctions.Eventhoughvariousshadesofgrayhavebeenusedintheexamplessofar,any huewillwork.Figure610usesvariousintensitiesofbeige.

Figure610.Thisbulletgraphusesvariousintensitiesofbeigetoencodequalitativestates.

Youcanencodemorethanthreequalitativestateswithbackgroundfillcolors,buttoavoidcomplexitythat cannotbeperceivedefficientlyandtomaintainacleardistinctionbetweenthecolors,youshouldn't exceedfive.Figure611illustratesthispracticallimit.

Figure611.Thisbulletgraphusesfivedistinctcolorintensitiestoencodequalitativestates.

Itissometimesusefultocompareakeymeasuretomorethanoneothermeasure.Forinstance,youmight wanttocomparerevenuetotherevenuetargetandtotherevenueamountatthistimelastyear.The bulletgrapheasilyhandlesmultiplecomparisonsbyusingadistinctmarkerforeach.Thesedistinctionscan bedisplayedusingvariablesofcolorintensity,linewidth(a.k.a.strokeweight),orevensymbolshapesina pinch.Figure612illustrateshowtwocomparisonscanbeincludedusingmarkerswithdifferentstoke weights.

Figure612.Thisbulletgraphincludestwocomparisons,whichhavebeenmadevisuallydistinctthroughtheuseofdifferent strokeweights.

WhenIoriginallydevelopedthedesignspecificationforthebulletgraph,Icalleditbyadifferentname:a performancebar.Thisoriginalnamepossessedchutzpahandevokedasenseofgoodhealth,duetoits similaritytothosepopularultraperformancenutritionsnackslikethePowerBar.Ihadtochangethename, however,becauseIeventuallyrealizedthatthereweretimeswhenthekeymeasureshouldbeencoded usingsomethingotherthanabar. Wheneveryouuseabartoencodeaquantitativevalue,asyou'veseenineachoftheexamplesofbullet graphssofar,thequantitativescaleshouldstartatzero.Thelengthofthebarrepresentsthevalue,notjust thelocationofitsendpoint,soascalethatstartsanywherebutzerowillproduceabarwithalengththat doesn'tcorrespondtoitsvalue.Thismakesaccuratecomparisonsbetweenbarsverydifficult. Itissometimesusefulwithbulletgraphs,however,toavoidstartingthequantitativescaleatzerosothat thescalecanbenarrowedtodisplaymorequantitativedetail.Forinstance,supposethatallofthevalues thatneedtobeincludedinthebulletgraphfallbetweentherangeof$150,000and$300,000,andyou wanttofocusexclusivelyonthisrangeofvaluestoshowmoresubtletyinthedifferencesbetweenthekey measureanditscomparisons(forexample,atarget).Inthiscase,youshouldusesomemeansotherthana bartoaccuratelyencodethekeymeasure.Forexample,youcanuseamarker(asimplesymbolshape)to

www.it-ebooks.info

encodethekeymeasureanddifferentlyshapedmarkersforanycomparativemeasures.Figure613 illustratesthisapproach.

Figure613.Becausethequantitativescaleofthisbulletgraphdoesnotbeginatzero,itusesasymbolmarkerratherthanabar toencodethekeymeasure.Inthiscase,thekeymeasureisencodedasacircleandthetargetmeasureisencodedasashort line.

Usingabartoencodethekeymeasurehastheadvantageofsuperiorvisualweighttohighlightthekey value,butasymbolmarkerallowsyoutonarrowthequantitativescaletodisplaygreatersubtletyinthe valuesandtheirdifferences(usingthesymbolmarkerservesasavisualalerttotheviewerthatthescale doesnotstartatzero).Bothworkwellonadashboard. Let'slookatonemorewayyoucanusebulletgraphs.Wheneveryoucompareacurrentmeasuretoa futuretarget,suchasrevenueasofJanuary15comparedtoaQuarter1target,youcaneasilyseehowfar youarefromthetarget,butit'snotalwayssoeasytotellifyouareontracktomeetorsurpassthatfuture target,whichcouldstillbeweeksorevenmonthsaway.Thisistruewhetheryouareusingabulletgraphor anyothergraphicalmeanstodisplaythisinformation.Thisshortcomingintheusefulnessofthe comparisoncanbeamelioratedbyaddingaprojectionofwhereyou'llbeattheendoftheperiodoftime thatisrelevanttothetarget.ThebulletgraphinFigure614onthenextpagesplitstherevenuemeasure intotwosegments:theactualmeasureasoftodayandtheprojectedmeasureofrevenuebasedoncurrent performance.Thisprovidesarichdisplaythattellsyounotonlyhowfaralongyouareonthepathtothe futuretarget,butalsohowwellyou'redoingtodayinrelationtothattarget.

Figure614.Thisbulletgraphdisplaysboththeactualquartertodaterevenueandaprojectionofexpectedquarterendrevenue basedoncurrentperformance.

Icanstatewithsomeconfidencethatbulletgraphsworkwell,becauseI'vetestedthemincontrolled experimentstocomparethemtosimpleradialgauges.Inmytests,bulletgraphsoutperformedradial gaugesbothinefficiencyandaccuracyofperception.Thenumberoftestsubjectswasfartoosmallto satisfyscientificstandards,soI'llrefrainfromclaimingspecificmeasuresofsuperiorperformance.These testsweresufficient,however,toenablemetostatewithoutreservationthatbulletgraphsworkeverybit aswellondashboardsasradialgaugesandareabletoconveythesameinformationinmuchlessspace.I believethatmakesthemsuperior. 6.2.1.2.Bargraphs Unlikebulletgraphs,bargraphsaredesignedtodisplaymultipleinstances,ratherthanasingleinstance,of oneormorekeymeasures.Infact,everygraphinthisproposedlibraryotherthanthebulletgraphis designedtodisplaymorethanoneinstanceofoneormoremeasures.Bargraphsaregreatfordisplaying measuresthatareassociatedwithitemsinacategory,suchasregionsordepartments.ThegraphinFigure 615isatypicalexamplethatcouldbefoundonadashboard:itdisplaystwokeymeasuresbookingsand billingsrevenuesubdividedintosalesregions.

www.it-ebooks.info

Figure615.Atypicalbargraph.

Iusetheterm"bargraph"inreferencetoallgraphsthatusebarstoencodedata,whethertheyare orientedverticallyorhorizontally. Figure616showsanotherexampleofatypicalbargraph,thistimewiththebarsrunninghorizontally.

Figure616.Abargraphwithhorizontallyorientedbars.

Tofullyunderstandwhenitisappropriatetoencodedatainagraphasbarsratherthanaslines(asinaline graph),youmustunderstandalittleaboutthethreetypesofcategoricalscalesthatappearcommonlyin graphs: Nominalscalesconsistofdiscreteitemsthatbelongtoacommoncategorybutreallydon'trelatetoone anotherinanyparticularway.Theydifferinnameonly(thatis,nominally).Theitemsinanominalscale,in andofthemselves,havenoparticularorderanddon'trepresentquantitativevaluesinanyway.Typical examplesindashboardsincluderegions(forexample,TheAmericas,Asia,andEurope)anddepartments (forexample,Sales,Marketing,andFinance).

www.it-ebooks.info

Ordinalscalesconsistofitemsthat,unliketheitemsinanominalscale,dohaveanintrinsicorder,butin andofthemselvesstilldonotcorrespondtoquantitativevalues.Typicalexamplesinvolverankings,suchas "A,B,andC,""small,medium,andlarge,"and"poor,belowaverage,average,aboveaverage,and excellent." Intervalscales,likeordinalscales,alsoconsistofitemsthathaveanintrinsicorder,butinthiscasethey representquantitativevaluesaswell.Anintervalscalestartsoutasaquantitativescalethatisthen convertedintoacategoricalscalebysubdividingtherangeofvaluesintheentirescaleintoasequential seriesofsmallerrangesofequalsizeandgivingeachrangealabel.Considerthequantitativerangemade upofvaluesextendingfrom55to80. Thisrangecouldbeconvertedintoacategoricalscaleoftheintervaltypeconsistingofthefollowing sequenceofsmallerranges: Greaterthan55andlessthanorequalto60 Greaterthan60andlessthanorequalto65 Greaterthan65andlessthanorequalto70 Greaterthan70andlessthanorequalto75 Greaterthan75andlessthanorequalto80

Figure617showsanexampleofeachtypeofscale.

Figure617.Thethreetypesofcategoricalscalesfoundingraphs.

Here'saquick(andsomewhatsneaky)testtoseehowwellyou'vegraspedtheseconcepts.Canyouidentify thetypeofcategoricalscalethatappearsinFigure618?
Figure618.Thisisacategoricalscalethatiscommonlyusedingraphs.Canyoudeterminewhichofthethreetypesitis?

Monthsoftheyearobviouslyhaveanintrinsicorder,whichbegsthequestion:"Dotheitemsinatime seriescorrespondtoquantitativevalues?"Infact,theydo.Unitsoftimesuchasyears,quarters,months, weeks,days,hours,andsoonaremeasuresofquantity,andtheindividualitemsinanygivenunitof measureforexample,yearsrepresentequalintervals.(Actually,monthsaren'texactlyequal,andevenyears varyinsizeoccasionallyduetoleapyears,buttheyarecloseenoughinsizetoconstituteanintervalscale forreportingpurposes.) Bargraphsneverlinegraphsarethebestmeanstodisplaymeasuressubdividedintodiscreteinstances alonganominalorordinalscale.Thevisualweightofbarsplacesemphasisontheindividualvaluesinthe

www.it-ebooks.info

graphandmakesiteasytocompareindividualvaluestooneanotherbysimplycomparingtheheightofthe bars.Lines,ontheotherhand,emphasizetheoverallshapeofthevalues,andbyconnectingtheindividual valuestheygiveasenseofcontinuityfromonevaluetothenextthroughouttheentireseries.Thissenseof connectionbetweenthevaluesisappropriateonlyalonganintervalscale,whichsubdividesacontinuous rangeofquantitativevaluesintoequal,sequentialintervals;it'snotappropriatealonganominalorordinal scale,wherethevaluesarediscreteandnotintimatelyconnected.Figure619showssomeexamplesof inappropriateandappropriateusageoflinestoencodedataingraphs.

Figure619.Examplesofinappropriate(toptwo)andappropriate(bottomtwo)usesoflinestoencodedataingraphs.

Linegraphsareusefulforencodingvaluesalonganintervalscale,butthereareoccasionswhenitis preferabletouseabargraphtodisplaysuchmeasures.Forexample,whenyouwishtoemphasizethe individualvaluesratherthantheoveralltrendsorotherpatternsofthevalues,orwhenyouwishtoenable closecomparisonsofvaluesthatarelocatednexttooneanother,abargraphisabetterchoice.Figure620 onthenextpagedisplaysthesameintervaldataintwoways:asabargraphandasalinegraph.Noticethe differencesinwhatthetwoimagesemphasize,despitethefactthatthedataarepreciselythesame.The bargraphemphasizestheindividualvaluesineachintervalandmakesiteasytocomparethosevaluesto oneanother,whilethelinegraphdoesamuchbetterjobofrevealingtheoverallshapeofthedistribution.

www.it-ebooks.info

Figure620.Thesetwographsoneabargraphandonealinegraphdisplayexactlythesamedatabuthighlightdifferentaspects ofit.

Becausebargraphsemphasizeindividualvalues,theyalsoenableeasycomparisonsbetweenadjacent values.Figure621illustratestheeasewithwhichyoucancomparemeasuresinthiscasetheproductivity ofthedaytimeandthenighttimecrewsinanygivenmonthusingthistypeofgraph.

Figure621.Barsarepreferabletolinesforencodingdataalonganintervalscaleinthiscase,atimeseriesdividedinto monthswhenthegraphisintendedtosupportcomparisonsofindividualmeasures.

Evenwhenyouwishtodisplayvaluesthatrepresentpartsofawhole,youshoulduseabargraphrather thantheeverpopularpiechart.Thiswillpresentthedatamuchmoreclearlyjustbesuretoindicate somewhereintext(forexample,inthegraph'stitle)thatthebarsrepresentpartsofawhole.Figure622 providesanexampleofbothapiechartandabargraphusedtopresentthesameparttowholedata. Noticehowmucheasieritistomakeaccuratevisualjudgmentsoftherelativesizesofeachpartinthebar graph.

www.it-ebooks.info

Figure622.Youcanuseabargraphtomoreclearlydisplaythesameparttowholedatathatiscommonlydisplayedwithapie chart.

6.2.1.3.Stackedbargraphs Avariationofthebargraphthatissometimesusedtodisplaybusinessdataisthestackedbargraph.This typeofgraphisusefulforcertainpurposes,butitcaneasilybemisused.Irecommendagainsteverusinga stackedbargraphtodisplayasingleseriesofparttowholedata.Aregularbargraphworksmuchbetter. Asyoucansee,itismuchharderandmoretimeconsumingtoreadthestackedbargraphinFigure623 thanthebargraphshowingthesamedatainFigure622.

Figure623.Astackedbargraphisnotthebestwaytodisplayasingleseriesofparttowholedata.

www.it-ebooks.info

Stackedbargraphsaretherightchoiceonlywhenyoumustdisplaymultipleinstancesofawholeandits parts,withemphasisprimarilyonthewhole.Figure624providesanexamplewithaseparateinstanceof salesrevenueperquarter,eachsubdividedbysaleschannel.

Figure624.Theonlycircumstancewhenastackedbargraphisusefuliswhenyoumustdisplaymultipleinstances(forexample, oneforeachquarter)ofawhole(totalsales)anditsparts(inthiscase,persaleschannel),withagreateremphasisonthewhole thantheparts.

Thechangesinthedistributiontendtobesomewhatdifficulttodetectforallthesegmentsexcepttheone thatappearsatthebottomofeachbar(inthiscase,"Direct"sales),whichiswhyastackedbargraph shouldnotbeusedifthesechangesmustbeshownmoreprecisely.Noticethedetailregardingthechanges indistributionofsalesthatcaneasilybeseeninthebargraphsinFigure625(especiallytheoneonthe right).Ifyouwanttoclearlydisplayboththewholeanditsparts,youcanuseeithertwographsnexttoone anotheroneforthewholeandoneforitspartsoracombinationbarandlinegraphwithtwoquantitative scalesonefortheparts,encodedasindividualbars,andoneforthewhole,encodedasaline.

Figure625.Thesebargraphsrevealtheshiftsinthedistributionofsalesbetweenthefourchannelsmuchmoreclearlythanthe stackedbargraphinFigure624.

6.2.1.4.Combinationbarandlinegraphs Whenyoucombinebarsandlinestogetherinasinglegraph,youshouldn'tdosoarbitrarily.This combinationshouldbeusedonlywhensomedatacanbedisplayedbestusingbars,withanemphasison

www.it-ebooks.info

individualvaluesandlocalcomparisons,andsomeusingaline,withanemphasisontheoverallshapeof thedata.Acommonexampleinvolvesdisplayingrevenuesandexpenses(usingbarstohighlightthe individualmonths)alongwithprofits(usingalinetohighlightthetrend),asseeninFigure626.

Figure626.Thisgraphcombinesbarsandalinetohighlightmonthlyrevenuesandexpensesontheonehandandtheoverall trendofprofitsontheother.

AlesscommonuseofcombinationbarandlinegraphsisonethatIsuggestedinthebargraphsection aboveasawaytoclearlydisplaymultipleinstancesthroughtimeofboththeindividualpartsofawhole andthewholeitself.TheexampleinFigure627solvesthisproblem.

Figure627.Exampleofacombinationbarandlinegraphthatdisplaysquarterlyinstancesofrevenuebysaleschannel,encoded asbars,andtotalrevenue,encodedasaline.

Thisisacombinationbarandlinegraphwithonequantitativescaleforthebarsandanotherfortheline.It isn'tnecessarytousetwoquantitativescales,oneontheleftaxisandoneontheright,butdoingso

www.it-ebooks.info

eliminatesthewastedspacethatwouldotherwiseappearinthegapbetweenthetotalsalesvaluesandthe muchsmallervaluesfortheindividualsaleschannels. AnotherusefulcombinationofbarsandalinebreaksarulethatIdeclaredearlier,whenIsaidthatyou shoulduselinesonlytoencodedataalonganintervalscale.Thereisoneexceptiontothisrule,which involvesaspecialkindofgraphcalledaParetochart(namedafteritsinventor,VilfredoParetothesame fellowwhoformulatedthewellknown80:20ruleofdistribution1).Let'slookatanexample,andI'llexplain whytheParetochartdeservestobeanexceptiontomygeneralruleabouttheuseoflinesingraphs. Paretochartsdisplayindividualvaluesasbarsandthecumulativetotalofthosevaluesasalinealonga categoricalscale.ThecategoricalscaleinaParetochartmaybeatimeseries,suchasmonthsoftheyear; thisisanintervalscale,sotheuseofalineinthiscasedoesn'tneedanexplanation.TheexampleinFigure 628doesnothaveanintervalscale,butalinestillworkswellinthisexample.

Figure628.ThisParetochartdisplayssalesrevenuebysalesrepresentatives,encodedasbars,aswellasthecumulative revenue,encodedasaline.

Thisgraphhasbeendesignedtoclearlyshowthatthetop3of10totalsalesrepresentativeswere responsiblefor75%oftotalrevenueforthequarter.Thecategoricalscale,consistingofsales representatives,isanordinalscalebyvirtueofthefactthatthesalespeoplehavebeenarrangedinorderby rank,basedontheirsales.Cumulativesales,astheyincreasefromonesalespersontothenextinranked order,representmeaningfulchange.Eachsuccessivevalueisintimatelyconnectedtotheonethat precedesit,becauseitisthesumofitselfandthepreviousvalue.Thisintimateconnectionmeritstheuse ofalinetoencodechangesinvaluesfromonetothenext.Theslopeofthelineprovidesusefulinformation inthiscontext:thesteeperthelinefromonesalespersontothenext,thegreaterthatsalesperson's revenuecontributionwasrelativetothenextbestsalesperson's.Byviewingthelineasawhole,youcan easilyseehowevenlydistributedthecontributionsofthesalespeopleare,orhowmuchtheyareskewed towardthetopperformers.
The80:20ruleofdistributionisoftenusedinreferencetoacompany'srevenue,usuallystatingthat80%ofthe revenuecomesfrom20%ofthecustomers.Pareto'soriginalobservationthatledtotheformulationofthisrulein 19thcenturyItalywasthat80%ofthecountry'swealthwasownedby20%ofthepopulation.
1

www.it-ebooks.info

6.2.1.5.Linegraphs Linegraphsdoanexceptionaljobofrevealingtheshapeofdataitsmovementupanddownfromonevalue tothenextespeciallyasitchangesthroughtime.Anytimethatyouwishtoemphasizepatternsinthedata, suchastrends,fluctuations,cycles,ratesofchange,andhowtwodatasetsvaryinrelationtooneanother, linegraphsprovidethebestmeans.Keepinmindthatwhenyoudisplaytimeseriesdataonadashboard, theshapeofthedata("Isitgoingupordown?""Isitvolatile?""Doesitgothroughseasonalcycles?")is generallythepicturethatisneeded,ratherthantheemphasisonindividualvaluesthatbargraphsprovide. Inthecontextofdashboards,linegraphsareoftenthebestmeanstopresentaquickoverviewofatime series. Figure629showsthesametimeseriesdataintwoways:ontheleftusingabargraphandontheright usingalinegraph.Noticehowmuchmorequicklyandclearlytheoverallshapeofthedatacomesthrough inthelinegraph.Unlikeabargraph,thequantitativescaleofalinegraphneednotbeginatzero,butitcan benarrowedtoarangeofvaluesbeginningjustbelowthelowestandjustabovethehighestvaluesinthe data,therebyfillingthedataregionofthegraphandrevealinggreaterdetail.Alwaysbesuretomakethe linesthatencodethedatamoreprominentthananyotherpartofthegraphsothatthedatastandsout aboveallelse.

Figure629.Twographsofthesametimeseriesdata:abargraphontheleftandalinegraphontheright.Noticehowthe overallshapeofthedataismucheasiertoseeinthelinegraph.

6.2.1.6.Sparklines SparklinesarethebrainchildofEdwardR.Tufte,atrueaficionadoofdatadisplay.Hehasdedicatedafull chaptertotheminhisbookBeautifulEvidence(asyetunpublishedbutexpectedin2006).Mytreatmentof thesubjectisbriefandfarfromdefinitive;mypurposehereistodescribesparklinesonlytotheextent necessarytodemonstratetheirvaluablecontributiontodashboards.Figure630providesanexampleofa simplesparkline.


Figure630.Asimplesparklinethatdisplaysthe12monthhistoryofacheckingaccountbalance.

Tuftecreatedthesparklinetoprovideabarebonesandspaceefficienttimeseriescontextformeasures. AssumingthatthesparklineinFigure630encodesarolling12monthhistoryofanaccountbalance,the upsanddownsareinstantlyavailabletotheviewerwhowishestoconsiderthemeaningofthecurrent balanceinlightofitshistory.

www.it-ebooks.info

Tuftedescribessparklinesas"dataintense,designsimple,wordsizegraphics."1Assuch,theyareidealfor dashboardsandanythingelsethatrequireshighlycondensedformsofdatadisplay,suchasmedical diagnosticreportsthatincludepatienthistories. Youmightbewondering,"Where'sthequantitativescale?"It'snowheretobeseen,andthat'sintentional. Sparklinesarenotmeanttoprovidethequantitativeprecisionofanormallinegraph.Theirwholepurpose istoprovideaquicksenseofhistoricalcontexttoenrichthemeaningofthemeasure.Thisisexactlywhat's requiredinadashboard.Insteadofdetails,youmustdisplayaquickviewthatcanbeassimilatedata glance.Thedetailscancomelater,ifneeded,intheformofsupplementalgraphsandreports. Althoughalwayssmallandsimple,sparklinescanincludeabitmoreinformationthanwhatI'veillustrated sofar.Figure631showsasparklinethatincludesalightgrayrectangletorepresentthenumberof manufacturingdefectsthatareacceptable,whichrevealsthatinthelast30days(thefullrangeofthe sparkline)thenumberofdefectshasexceededtheacceptablerangeonthreeoccasions.Theoptionalred dotmarkingthefinalvalueinthesparklinetiestheendofthesparklinetothecurrentvalueoffiveby makingthembothred.

Figure631.Thissparklinedisplays30daysofmanufacturingdefecthistorycomparedtotheacceptablerange.

Peoplecommonlyusesimpleupordowntrendarrowstodisplaythedirectioninwhichameasureis moving,buttheseareoftenambiguous.InlookingattheMTDRevenuemeasureinFigure632,for example,itisn'tobviousiftheupwardtrendarrowindicatesthatrevenueistrendingupwardoverallfor theyear,thequarter,themonth,orjustsinceyesterday.


Figure632.Simpletrendarrowsareoftenusedondashboards,butwhattheymeanissometimesunclear.

Asparkline,however,asshowninFigure633,isnotambiguous,becauseitdisplaystheentireperiodof historyacrosswhichthetrendapplies.
Figure633.Thissparklineprovidesaclearpictureofthehistoricaltrendleadinguptothepresentmeasure.

Asyoucansee,sparklinesareidealfordashboards.Everydashboardvendoroughttosupportthem. 6.2.1.7.Boxplots Theboxplotisafairlyrecentadditiontothelexiconofgraphs.Itwasinventedinthe1970sbyan extraordinarymathematiciannamedJohnWilderTukey,whospecializedindatadisplay.Thisparticular typeofgraphdisplaysthedistributionofvaluesetsacrosstheentirerange,fromthesmallesttothe largest,withmanyusefulmeasuresinbetween. Itisofteninadequatetodescribeasetofvaluesasasinglesummarizedmeasuresuchasasumoraverage. Attimesitisimportanttodescribehowthosevaluesaredistributedacrosstheentirerange.Forinstance, tofullyunderstandthenatureofemployeecompensationinyourcompanyineachofthesalarygrades (thatis,specifiedlevelsofcompensationwithprescribedranges),youwouldcertainlyneedtoseemore
1

EdwardR.Tufte,BeautifulEvidence(Cheshire,CT:GraphicsPress,2006).

www.it-ebooks.info

thanthesumofsalariesforeachsalarygrade.Evenameasureofaveragecompensation,suchasthemean ormedian,wouldn'ttellyouenough.Let'slookatafewdifferentwaysthatthisdatacouldbepresented. Figure634presentsthemediansalaryineachgradethatis,thevaluethat'sinthemiddleofeachrange.

Figure634.Thisgraphdisplaysemployeesalariespersalarygradeasasinglemedianvalueforeachgrade.

Theadequacyofthisdisplaydependsonyourpurpose.Ifyourpurposerequiresasenseofhowsalariesare distributedacrosseachrange,thisdisplaywon'ttellenoughofthestory.Themedianexpressestheexact centeroftherange,butnothowthevaluesaredistributedaroundthatcenter.Figure635showssixquite differentexamplesofhowtheindividualsalariesinasinglesalarygradewithapotentialrangeof$20,000 to$30,000andamedianpreciselyinthemiddleat$25,000mightbedistributedacrossthatrange.Asyou cansee,themedianalonetellsalimitedstory,soitisoftenusefultodisplaythedatainawaythatreveals moreabouthowthevaluesaredistributed.

www.it-ebooks.info

Figure635.Sixexamplesofhowasetofsalarieswiththesamemedianvaluemightbedifferentlydistributed.Thescaleonthe verticalaxesrepresentsthenumberofemployeeswhosesalariesfallintoeachoftherangesthatrunalongthescaleonthe horizontalaxes.

ThegraphinFigure636onthenextpageillustratesthesimplest(andleastinformative)waytodisplay howsetsofvaluesaredistributed.Itusesrangebarstodisplaytwovaluesforeachsalarygrade:thelowest andthehighest.Althoughitisusefultoseethefullrangeofeachsalarygrade,thissimpleapproachstill tellsusnothingabouthowindividualvaluesaredistributedacrossthoseranges.Dothevaluesclusternear thebottom,center,ortop,oraretheyevenlydistributed?

www.it-ebooks.info

Figure636.Thisisthesimplestbutleastinformativewaytodisplayrangesofvalues.Itusesrangebarsthatencodethelowest andhighestsalariesineachsalarygrade.

Withacombinationofrangebarsandameasureofthemedian,asshowninFigure637,abitmoreinsight beginstoemerge.

Figure637.Thisgraphcombinesrangebarswithdatapointstomarkthemediansaswellasthehighandlowsalariesineach salarygrade.

Knowingthatbydefinitionhalfofthevaluesarelargerthanthemedianandhalfaresmaller,weknowthat whenthemedianisclosertothelowendofarangeofvalues,morevaluesfallintothelowerhalfthanthe upperhalfoftherange.Thecloserthemedianistothebottomoftherange,themoreskewedthevalues areintheoppositedirection.Theoppositeistruewhenthemedianliesclosertothetopoftherange.The understandingofthedistributionthatisrevealedbythisrelativelysimpledisplaycertainlyisn'tcomplete, butit'sdefinitelygettingbetterandisprobablysufficientformanypurposesonadashboard.

www.it-ebooks.info

Youcanthinkofthecombinationofrangebarswithdatapointstomarkthemediansasasimplisticversion ofaboxplot.Atrueboxplot,asintroducedbyTukey,providesmoreinformation.Theboxportionofabox plotissimplyarectangle(orbar)withorwithoutafillcolor.Aswitharangebar,thebottomofthebox representsavalueandthetoprepresentsavalue,buttheseareusuallynotthelowestandhighestvalues intherange.Figure638illustratesafullgrownversionofasingleboxplotwith"whiskers"(knownasa boxandwhiskerplot).Thisisjustoneofthemanyvariationsthatarecommonlyused.

Figure638.Anindividualboxplotwithwhiskers.Outliersareindividualdatavaluesthatfalloutsidetherangethatisdefinedby thewhiskers.

Agraphwithboxeslikethisconveysarichpictureofdatadistributionperhapstoorichformostdashboards andmostofthefolkswhousethem.Asimplerversionoftheboxplot,suchastheoneinFigure639onthe nextpage,maybepreferablefordashboarduse

www.it-ebooks.info


Figure639.Asimplifiedversionofaboxplotsuchasthisoneisusuallymoreappropriatefordashboardsthantheoneshownin Figure638.

6.2.1.8.Scatterplots Ascatterplotdoesonlyonething,butitdoesitquitewell:itdisplayswhetherornot,inwhatdirection,and towhatdegreetwopairedsetsofquantitativevaluesarecorrelated.Forinstance,ifyouwanttoshowthat thereisarelationshipbetweenthenumberofbroadcastadsandsalesrevenues,ascatterplotsuchasthe oneinFigure640wouldworknicely.

Figure640.Thisscatterplotdisplaysthecorrelationbetweenthenumberofbroadcastadsandtheamountofsalesrevenuefor 24months.

Inthiscase,boththenumberoftimesadswereairedandthesalesrevenuesforeachmonthwere collectedasapairedsetofvaluesfor24months.Thisgraphtellsusthefollowing:

www.it-ebooks.info

Thereisacorrelationbetweenadsandsalesrevenue,indicatedbythefactthatachangeinthe numberofadsalmostalwayscorrespondedtoachangeinsalesrevenue. Thecorrelationispositive(upwardslopingfromlefttoright),indicatingthatasthenumberof adsincreasedthesalesrevenuealsousuallyincreased. Thecorrelationisfairlystrong.Thisisindicatedbythetightgroupingofthedatavaluesaround thetrendline,showingthatanincreaseordecreaseinadsfromonemeasuretoanother almostalwayscorrespondedtoasimilaramountofincreaseordecreaseinsalesrevenue.

Giventhateachpairofmeasureswascollectedforagivenmonthacross24consecutivemonths,thisdata couldhavebeendisplayedasatimeserieslinegraph,butthenatureofthecorrelationwouldnothave stoodoutasclearly. Thescatterplotwillstillworknicelyifyousplitthemeasuresintomultiplesets.Forinstance,youcouldsplit theadsintotwotypesradioandtelevisionasshowninFigure641.Aquickexaminationofthisdisplaytells usthatthecorrelationoftelevisionadstosalesrevenueismorepositive(upwardsloping)thanthatof radioads,thoughthestrengthofeachcorrelation(theproximityofthedatavaluestothetrendline) appearstobeaboutthesame.

Figure641.Thisscatterplotdisplaysthecorrelationbetweenthenumberofradioandtelevisionadsandtheirrespective amountofsalesrevenuefor24months.

Scatterplotsaresometimesrenderedthreedimensionally,inordertodisplaythecorrelationofthree quantitativevariables,ratherthanjusttwo.Othermethodsaresometimesusedaswelltoincreasethe numberofcorrelatedvariablesinasinglescatterplot.Irecommendagainstusinganyoftheseapproaches onadashboard,however,becauseevenwhentheyaredesignedaswellaspossible,theyrequiretoomuch studytounderstandtimethatdashboardviewersdon'thave. OneotherpointI'dliketomentionisthattheuseofastraighttrendline(alsoknownasalineofbestfit)in ascatterplotmakesthedirectionandstrengthofthecorrelationstandoutmorethanjusttheindividual datapointsbythemselves.ThegraphinFigure642ispreciselythesameastheoneinFigure641,except thatitlackstrendlines.Itiseasytoseethatthedirectionandespeciallythestrengthofthecorrelations wouldrequiremoretimetodiscernwithoutthetrendlines.Linesofbestfitcomeinseveraltypes,someof

www.it-ebooks.info

whicharecurved,andeachworksbestfordatasetsthatexhibitparticularpatterns.Knowledgeofwhento usethemandhowtointerpretthem,however,isnotcommonexceptamongstatisticians,soitisbestto avoidallbutthesimplestraightlineofbestfitunlessyouandthedashboard'susershavethenecessary trainingtounderstandtheotherforms.

Figure642.Thisscatterplotdisplaysthecorrelationbetweenthenumberofradioandtelevisionadsandtheirrespective amountofsalesrevenuefor24months,thistimewithoutthetrendlinesthatappearinFigure641.

6.2.1.9.Treemaps Treemaps,developedinthe1990sbyBenShneidermanoftheUniversityofMaryland,aregraphsusedto displaylargesetsofhierarchicallyorcategoricallystructureddatainthemostspaceefficientwaypossible. Shneidermanisoneofthemostinspiringresearchersandinnovatorsworkingininformation visualizationonewhoplayedamajorroleindefiningthedomain.Treemapscompletelyfillavailablescreen spacewithasetofcontiguousrectanglesthathaveeachbeensizedtoencodeaquantitativevariable. Hierarchiesandcategoriesarerepresentedasrectanglescontainedwithinlargerrectangles.Inadditionto thequantitativevariablethatisassociatedwithrectanglesize,colorcanalsobeusedtoencodeasecond quantitativevariableforprovidingarichermultivariatedisplay. Thepurposeoftreemapsisnottomakefinequantitativecomparisonsortorankitems,butrathertospot particularconditionsofinterest.The2Dareasofrectanglesandvariationsincolordonotsupporteasy, efficient,oraccuratevaluecomparisons,butwhenthesevisualattributesarecombinedinthetreemap, theycanmakeparticularconditionsjumpoutandtherebyenabletheprocessofdiscovery. Duetotheirspaceefficientdesign,treemapscanbeusedquiteeffectivelyondashboards,buttheyshould bereservedforthosecircumstancesforwhichtheyweredeveloped,and,whenused,shouldbedesigned withcare.TheexampleinFigure643illustratesanappropriatelyappliedandeffectivelydesignedtreemap forabusinessdashboard.

www.it-ebooks.info

Figure643.Thistreemap,createdusingTreemap4.3softwaredevelopedattheUniversityofMaryland'sHumanComputer InteractionLab(HCIL),displayssalesdata(revenueandpercentageofquota)byregion.

Itdisplayssalesbyregion,withrevenueencodedasrectanglesizeandthepercentageofsalesquota achievedencodedascolor(rangingfrombrightredasthelowestpercentageandpurewhiteasthe highest).Noticehowyoureyesaremostlydrawntothelargeredrectangles,whichrepresentstateswith largerevenuesthatareperformingpoorlyinotherwords,stateswhoseperformanceresultsinthegreatest negativeaffectonrevenue(forexample,California).Ifyou'reinterestedinspottingthosestateswhose goodperformanceishavingthegreatestpositiveaffectonrevenue,yousimplylookforthelargestlight coloredrectangles(forexample,Florida). Ichosetouseasinglehueratherthanseveraltoencodethepercentageofsalesquota,varyingthevalues byintensityfromcompletelyunsaturatedred(thatis,white)tofullysaturated,brightred.Itiscommonfor thistypeofdatatobeencodedinatreemapusingmultiplehues,suchasredforvaluesthatarebelow quotaandgreenforthosethatareabovequota.Typically,thesecolorswouldrangefrombrightredatthe lowendthroughdarkeranddarkershades,reachingblackinthemiddle(forvaluesclosetothequota),and proceedingthroughdarkshadesofgreenallthewaytobrightgreenatthehighend.Ifacleardistinction betweenvaluesthatarebelowquotaandthoseaboveisnecessary,thenmultiplehueswouldwork,butI believethatoftenwhensuchdistinctionsaredisplayed,theyareunwarranted.Ifyouareresponsiblefor monitoringsalesperformancebystate,doyoureallywanttoseeaqualitativedistinctionbetweenastate thatisslightlybelowquota(darkred)andonethatisslightlyabovequota(darkgreen)?Arethesevalues reallythatdifferent? Treemapsareusuallyinteractive,providingthemeanstoselectaparticulariteminthehierarchyandthen drilldownintothenextlevelofitemsthatbelongtothehigherlevelitemthatyouselected.Thisenables easynavigationthroughthehierarchytoinvestigateparticularconditionsofinterest,potentiallyrevealing

www.it-ebooks.info

whatisgoingonatlowerlevelsthatiscreatingtheseconditions.Thisprovidesasimplepathformorefully exploringandrespondingtothoseconditionsthatjumpoutonthedashboardasneedingattention. 6.2.1.10.Finalthoughtsaboutgraphs Youmightbewonderingwhysomeoftheothergraphsthatarefamiliartoyouaremissingfromthis proposedlibrary.Eachismissingforoneofthefollowingtworeasons: ItcommunicateslesseffectivelythananalternativethatI'veincluded. Itistoocomplexforthetypicalneedsofadashboard.

Thepiechartprobablytopsthelistofoftenusedgraphsthatwereleftoutofthislibraryofgraphsbecause theycommunicatelesseffectivelythanothermeans.Piechartsweredesignedtodisplayparttowhole information,suchastheindividualproductsthatmakeupanentireproductline.Aswe'vealready discovered,however,parttowholeinformationcanbecommunicatedmoreclearlyusingabargraph. Anothercomparisonofthetwotypesofgraphusedtodisplaythesamesetofparttowholedataisshown inFigure644.

Figure644.Thispiechartandbargraphbothdisplaythesameparttowholedata.Thevaluesaremucheasiertointerpretand comparewhenabargraphisused.

Viewerscanprocesstheinformationinthebargraphontherightmuchmorequicklyandeasilythaninthe piechartontheleft.Why?Whereasabargraphusesthepreattentivevisualattributeoflinelength(thatis, thelengthsorheightsofthebars)toencodequantitativevalues,piechartsencodevaluesasthetwo dimensionalareasoftheslicesandtheiranglesastheyextendfromthecentertowardthecircumference ofthecircle.Ourvisualperceptiondoesapoorjobofaccuratelyandefficientlycomparing2Dareasand angles.Theonlythingthatapiecharthasgoingforitisthatwhenyouseeoneyouautomaticallyknowthat youarelookingatmeasuresthatarepartsofawhole.Becausebargraphscanbeusedforothertypesof comparisons,whenyouusethemtodisplayparttowholedata,youmustlabeltheminamannerthat makesthisclear.Aslongasthisisdone,bargraphsarefarsuperior. Apiechartfallsintoalargerclassofgraphscalledareagraphs.Areagraphsuse2Dspacetoencode quantitativevalues,whichispronetoinaccurateinterpretationandoftentoocclusion(aproblemthatis causedwhenoneobjectishiddenentirelyorinpartbehindanother).TheareagraphinFigure645onthe

www.it-ebooks.info

nextpageillustratestheproblemofocclusionrevenuesforQuarters2and3intheWestandQuarter4in theNortharecompletelyhidden.

Figure645.Areagraphscansufferfromtheproblemofocclusion.

Anothertypeofgraphthat'ssurfacingmoreandmoreoftenthesedaysistheradargraph,acirculargraph thatencodesquantitativevaluesusinglinesthatradiatefromthecenterofthecircletomeettheboundary formedbyitscircumference.Itisnothingbutalinegraphwiththecategoricalscalearrangedalonga circularaxis,asyoucanseeontheleftinFigure646.Forcommonbusinessdataaradargraphisnotas effectiveasabargraph(shownontherightinFigure646),becauseitismoredifficulttoreadvalues arrangedinacircularfashion.TheonlytimeI'vefoundaradargraphtobetolerablefordisplayingtypical businessdatawaswhenthecategoricalscalecouldnaturallybeenvisionedascircularforexample,when themeasuresonthescalearethehoursofaday,duetothefamiliarcirculararrangementoftimeona clock.

Figure646.Thisradargraph(left)andbargraph(right)displaythesameexpensedata.Intheradargraph,departmentsare arrangedalongthecircumferenceandthequantitativescaleforexpensesresidesalongtheradialaxesthatextendfromthe center.Thebargraphismucheasierandfastertoread.

6.2.2.Icons Iconsaresimpleimagesthatcommunicateaclearandsimplemeaning.Onlyafewareneededona dashboard.Themostusefuliconsaretypicallythosethatcommunicatethefollowingthreemeanings:

www.it-ebooks.info

Alert Up/down On/off

6.2.2.1.Alerticons Itisoftenusefultodrawattentiontoparticularinformationonadashboard.Thisisespeciallytruewhen somethingiswrongandrequiresattention.Aniconthatworksasanalertshoutsattheviewer,"Hey,look here!"Foranicontoplaythisrolewell,itneedstobeexceptionallysimpleandnoticeable.Tenvariations ofanalerticon,eachwithitsownslightlydifferentmeaning,arefartoocomplexforadashboard.Tryto limitalertlevelstoamaximumoftwo,andideallytoone.Asinglealerticoncatchestheeyemuchmore effectivelythanmultiplealertswithvariousmeanings. Acommonalertschemeondashboardsusesthetrafficlightmetaphor,composedofthreecolorswith differentmeanings.Greenistypicallyusedtoindicatethatalliswellbutwhat'sthepoint?Ifeverything's fine,youdon'tneedtodrawattentiontothedata.Alertsthatarealwaystheredrawlessattentionthan alertsthatappearonlywhenattentionisrequired.Thisisbecauseasimpleiconthatappearsonlyincertain circumstancesisperceivedpreattentivelyasan"addedmark."Thispreattentiveattributeisnottappedinto whenthetrafficlightalertsystemisused,becausealthoughthecolorusedtoencodethedatamaychange, nothingisbeingadded. I'vefoundthatasimpleshape,suchasacircleorsquare,usuallyworksbestasanalerticon.Ifyoumust communicatemultiplelevelsofalerts,ratherthanusingdistincticons,stickwithoneshapeandvarythe color.Trafficsignalcolorsofred,yellow,andgreenareconventional,buttheydon'tworkforthe10%of malesand1%offemaleswhoarecolorblind.Figure647illustratesthispointbyshowingthecolorsgreen, yellow,andredontheleftandwhatapersonwiththepredominantformofcolorblindnesswouldseeon theright.

Figure647.Theiconsontherightsimulatewhatsomeonewhoiscolorblindwouldseewhenlookingatthoseontheleft.

Asolutionthatworksforeveryoneinvolvesdistinctintensitiesofthesamehue,suchaslightred(inplace ofyellow)anddarkred,asshowninFigure648.

Figure648.Thesimplealerticonsontheleftusevaryingintensitiesofasinglehuetoencodedifferentmeanings.Thetwoon therightsimulatewhatapersonwhoiscolorblindwouldsee.Varyingintensitiesofanysinglehuearedistinguishableby everyone.

6.2.2.2.Up/downicons Up/downiconsconveythesimplemessagethatameasurehasgoneupordowncomparedtosomepoint inthepastorisgreaterorlesserthansomethingelse,suchasthetarget.Financialinformationiscommon

www.it-ebooks.info

ondashboards,andaquickwaytoindicatetheupordownmovementofstocks,profits,andsoonisoften useful.Fortunately,aconventionalsymbolisalreadyinusetocommunicatethesemeanings:atriangleor arrowwiththetippointingeitherupordown.Thecoloroftheiconsmayvaryaswell(usuallygreenfor goodandredforbad)sotheystandoutmoreclearly,butagain,thisisaproblemforthosewhoarecolor blind.Thisproblemcanbeavoidedbyusingcolorsthatvarygreatlyinintensityaswellashue,suchasfully saturatedredfortheiconthatindicatesmovementinthewrongdirectionandlesseyecatchingpalegreen fortheother.Figure649illustratesapossiblepresentationoftwoversionsofthissimpleicon.

Figure649.Simpleupanddownicons.

6.2.2.3.On/officons On/officonsserveasflagstoidentifysomeitemsasdifferentfromothers.Forexample,ifyoudisplayalist ofthetop10currentsalesopportunitiesandyouwanttoflagsomeasbeingclosertoclosingthanothers,a simpleon/officonwoulddothisnicely.Othertypicalusesincludemarkingfeatureditems,suchasproducts inalist,andpointingoutwhereyoucurrentlyareonaschedulethatincludeseventsthatextendintothe pastandfuture.Anyoneofmanysimpleiconscouldbeusedtoservethispurpose,butcheckmarks, asterisks,andXs(Figure650)areprobablythemostcommonandintuitivelyunderstood.Regardlessof whichiconyouchooseforthispurpose,itisbesttopickoneandsticktoit.Consistencymightseemboring, butondashboardsitmakesthingsclear.

Figure650.Sampleon/officons.

6.2.3.Text Alldashboards,nomatterhowgraphicallyoriented,includesomeinformationthatisencodedastext.This isbothnecessaryanddesirable,forsomeinformationisbettercommunicatedtextuallyratherthan graphically.Textisusedforthecategoricallabelsthatidentifywhatitemsareongraphs,butitisoften appropriateinotherplacesaswell.Anytimeitisappropriatetoreportasinglemeasurealone,without comparingittoanything,textcommunicatesthenumbermoredirectlyandefficientlythanagraph(Figure 651).Notethatintheseinstancessomemeanstodisplaythetextonadashboard,suchasasimpletext box,isnecessary.1

Figure651.Textcanbeusedonadashboardtoclearlyconveyasinglemeasureonitsown.

6.2.4.Images Themeanstodisplayimagessuchasphotos,illustrations,ordiagramsissometimesusefulonadashboard, butrarely,inmyexperience.Adashboardthatisusedbyatrainermightincludephotographsofthepeople


11

SeeChapter7,DesigningDashboardsforUsability,foradiscussionofchoosingfontsforuseonadashboard.

www.it-ebooks.info

scheduledtoattendtheday'sclass,oneusedbyamaintenanceworkermighthighlighttheareasofthe buildingwherelightbulbsneedtobereplaced,oroneusedbyapolicedepartmentmightuseamapto showwherecrimeshaveoccurredinthelast24hours.However,imageswillbeunnecessaryformost typicalbusinessuses. 6.2.5.DrawingObjects Itissometimesusefultoarrangeandconnectpiecesofinformationinrelationtooneanotherinwaysthat simpledrawingobjectshandlewithclarityandease.Forinstance,whendisplayinginformationabouta process,itcanbehelpfultoarrangeseparateeventsintheprocesssequentiallyandtoindicatethepath alongwhichtheprocessflows,especiallywhenbranchingalongmultiplepathsispossible.Another exampleiswhenyouneedtoshowconnectionsbetweenentities,perhapsincludingahierarchical relationship,suchasinanorganizationchart.Entitiescaneasilybedisplayedasrectanglesandcircles,and relationshipscanbedisplayedusinglinesandarrows.Forinstance,rectanglesorcirclescouldrepresent tasksinaproject,witharrowsconnectingthemtoindicatetheirrelationshipsandorder. Figures652and653provideexamplesofhowsomeoftheseobjectsmightbeused.Theycanalsobeused tohighlightandgroupinformation,whichisacommonneedindashboarddesign.Switchingbetween rectanglesandcirclesprovidesaneasywaytodistinguishdifferenttypesofentities.Linesandarrowsboth showconnectionsbetweenentities,butarrowsdisplaytheadditionalelementofdirection.

Figure652.Simpledrawingobjectscanbeusedtoclarifyrelationshipsbetweenthecomponentsofnetrevenue.

Figure653.Simpledrawingobjectscanbeusedtodisplayrelationshipsbetweentasksinaprojectplan.

www.it-ebooks.info

6.2.6.Organizers Itisoftenthecasethatsetsofinformationneedtobearrangedinaparticularmannertocommunicate clearly.Threeseparatewaysoforganizingandarrangingrelatedinformationstandoutasparticularlyuseful whendisplayingbusinessinformationondashboards: Tables Spatialmaps Smallmultiples

6.2.6.1.Tables Tablesarrangedataintocolumnsandrows.Thisisafamiliararrangementfortext(Figure654),butitcan alsobeusedtoarrangeanyoftheotherdisplaymediathatwe'vealreadyexamined.Arranginggraphs, icons,andimagesintocolumnsandrowsisoftenuseful.

Figure654.Atabulararrangementoftext.

6.2.6.2.Spatialmaps Spatialmapsofferamorespecializedandlessoftenneededformoforganization.Theycanbeusedto associatedatabothcategoricalandquantitativewithphysicalspace.Whendataistiedtophysicalspaceand itsmeaningcanbeenhancedbymakingthatarrangementvisible,spatialmapsareuseful. Themostcommonarrangementofdatarelatedtophysicalspaceisageographicalarrangementintheform ofamap.Whenthegeographicallocationofthethingbeingmeasuredmustbeseentounderstandthe data,placingthemeasuresonamapsupportsthisunderstanding.However,thisdoesn'tmeanthatany timemeasurescanbeshowninrelationtogeography,theyshouldbe;onlywhenthemeaningofthedata istiedtogeographyandthatmeaningcannoteasilybeunderstoodwithoutactuallyseeingthedata arrangedonamapshouldthisapproachbetaken.Forexample,salesrevenuecanbeunderstoodin relationtoasmallnumberofsalesregionswithoutdisplayingthedataonamap(seeFigure655),but displayingconcentrationsofabsenteeismamongemployeesinstoreslocatedthroughouttheUnitedStates onamapcouldrevealpatternsrelatedtolocationthatmightnotbeobviousotherwise.

www.it-ebooks.info

Figure655.Spatialmapscanbeusefulwhentheyaddtoourunderstandingofthedata,but,asinthiscase,theyareoftenused unnecessarily.

Thesecondmostusefultypeofspatialmaponadashboardisprobablythefloorplanofabuilding.If,for example,itisyourjobtomonitortemperaturesthroughoutalargebuildingandrespondwhenever particularareasexceedestablishednorms,seeingthetemperaturesarrangedonafloorplancouldbring relationshipsbetweenadjacentareastolightthatyoumightmissotherwise. 6.2.6.3.Smallmultiples ThelastorganizerarrangesgraphsinamannerthatEdwardTuftecalls"smallmultiples."Thisarrangement istabular,consistingofasingleroworcolumnofrelatedgraphs,ormultiplerowsandcolumnsofrelated graphsarrangedinamatrix.Ilistsmallmultiplesseparatelyfromtablesbecauseorganizersthatdisplay smallmultiplesoughttohavesomeintelligencebuiltintothemtohandleaspectsofthisarrangementthat wouldbetimeconsumingtoarrangemanuallyinatable. Inadisplayofsmallmultiples,thesamebasicgraphappearsmultipletimes,eachtimedifferingalonga singlevariable.Let'slookatanexample.Ifyouneedtodisplayrevenuedataasabargraphacrossfoursales regions,withbookingsandbillingsrevenueshownseparately,youcoulddosoinasinglegraph,asshown inFigure656.

www.it-ebooks.info

Figure656.Thisbargraphdisplaysthreevariables.

If,however,youmustsimultaneouslydisplaytherevenuesplitbetweenthreesaleschannels(forexample, solddirectly,throughdistributors,andthroughresellers),asinglegraphwon'twork.Totherescuecomes thesmallmultiplesdisplay.AsshowninFigure657,byarrangingthreeversionsofthesamegraphnextto oneanotheronegraphpersaleschannelyoucanshowtheentirepicturewithineyespan,making comparisonseasy.Toeliminateunnecessaryredundancy,youcouldavoidrepeatingtheregionlabelsin eachgraph,aswellasthelegendandtheoveralltitle.Thisnotonlysavesvaluablespace,whichisalways importantonadashboard,butitalsoreducestheamountofinformationthattheviewermustreadwhen examiningthedisplay.

Figure657.Thisseriesofhorizontallyalignedsmallmultiplesdisplaysrevenuesplitbetweenthreesaleschannels.

Anintelligentorganizerforsmallmultiplesbuiltintothesoftwarewouldallowyoutoreferencethedata, indicatewhichvariablegoesonwhichaxisofthegraph,whichshouldbeencodedaslinesofseparate colors,whichshouldvarypergraph,andfinallywhetheryouwantthegraphstobearrangedvertically, horizontally,orinamatrix;theorganizerwouldthenhandletherestforyou.Asofthiswriting,Ihaveyetto seedashboardsoftwarethatmakesthiseasytodo.Ireservethehope,however,thatthiswillsoonchange.

6.3.Summary
ThelibraryofdashboarddisplaymediathatI'veproposedinthischapteriscertainlynotcomprehensive, norwillitremainunchangedastimegoeson.Asnewgraphicinventionsemergethatsuitthepurposeand designconstraintsofdashboards,thislibrarywillcontinuetogrow,butIexpectthatitwilldososlowly.

www.it-ebooks.info

Justbecauseavendorintroducesanewvisualizationtechniquedoesn'tmeanitbelongsonadashboard. Let'skeepthevisiontruetoformandeffectiveforenlighteningandefficientcommunication.

Chapter7.DesigningDashboardsforUsability
Afewimportantaspectsofdashboard'svisualdesignremaintobeconsidered.Oneofthemostchallenging istheneedtoarrangemanyitemsofinformationoftenrelatedsolelybytheviewer'sneedtomonitorthem allinamannerthatdoesn'tresultinaclutteredmess.Thisarrangementmustsupporttheintrinsic relationshipsbetweenthevariousitemsandthemannerinwhichtheymustbenavigatedandusedto supportthetaskathand.Adashboard'sdesignmustoptimallyandtransparentlysupportitsuse.Thewhole alsomustbepleasingtolookupon,oritwillbeignored.

www.it-ebooks.info

Organizetheinformationtosupportitsmeaninganduse Maintainconsistencyforquickandaccurateinterpretation Maketheviewingexperienceaestheticallypleasing Designforuseasalaunchpad Testyourdesignforusability Beyondselectingappropriatedisplaymediaandreducingthenondatapixelstoaminimum,attentionalso mustbegiventoseveralotheraspectsofdesigntoguaranteethatyourdashboardsareeasytouseanddo everythingtheycantosupporttheviewer'sneedtorespondtotheinformation.Havingknowledgeofafew moredesignstrategiesunderyourbeltwillhelpyoublendallthevisualaspectsofyourdashboardintoa pleasingandfunctionaldisplay.

7.1.OrganizetheInformationtoSupportItsMeaningandUse
Youcan'tjusttakeinformationandthrowitontothedashboardanywayyouplease.Howthepiecesare arrangedinrelationtooneanothercanmakethedifferencebetweenadashboardthatworksandonethat endsupbeingignored,eventhoughtheinformationtheypresentisthesame.Keepthefollowing considerationsinmindwhenyoudeterminehowtoarrangedataonthescreen: Organizegroupsaccordingtobusinessfunctions,entities,anduse. Colocateitemsthatbelongtothesamegroup. Delineategroupsusingtheleastvisiblemeans. Supportmeaningfulcomparisons. Discouragemeaninglesscomparisons.

7.1.1.OrganizeGroupsAccordingtoBusinessFunctions,Entities,andUse Agoodfirstcutatorganizingdataistoformgroupsthatarealignedwithbusinessfunctions(forexample, orderentry,shipping,orbudgetplanning),withentities(departments,projects,systems,etc.),orwithuses ofthedata(forinstance,theneedtocomparerevenuesandexpenses).Thesearethenaturalwaysto organizemostbusinessdata. Inabusiness,becauseentitiesandfunctionsarepartsofaninterconnectedsystem,someonewhoserole spansmanyoftheseindividualunitsmightprefertoseedataorganizedinawaythatismoreintegrated andalignedwiththewaysheusesthatinformation.Forinstance,aCEOstandsabovethedivisionsfoundin anorganization'sstructureandusuallywantstoseerelationshipsamongdatathataremoreholistic, perhapsbasedontherelativeimportanceofeachitemtothecompany'sbottomline,fromgreatestto least.Inacaselikethis,itemsthatothersmightnaturallyseeasbelongingtodistinctgroupsmightbe groupedtogethertobetterservetheneedsoftheCEO.Ifthereisaparticularorderinwhichthedataought tobescannedtobuildthedesiredoverviewasefficientlyaspossible,groupingandorderingitems accordinglymightworkbest. Whenorganizingdataonadashboard,startbylearningpreciselyhowtheinformationwillbeusedandhow thepiecesoughttobearrangedtobestservetheseuses. 7.1.2.ColocateItemsThatBelongtotheSameGroup Onceyou'vedeterminedthoseitemsthatbelongtogetherrelativetothetaskathand,thebestmeansto connectthemistoplacethemclosetooneanother,yetdelineatedinsomesimplemannerfrom

www.it-ebooks.info

surroundinggroups.Usingpositiontogroupitemsvisuallyisastrategythatispreattentivelyandthus rapidlyperceived. 7.1.3.DelineateGroupsUsingtheLeastVisibleMeans Visualmeansthatareusedtodelineategroupsofdata,suchasgridlines,borders,andbackgroundfill colors,qualifyasnondatapixels.Assuch,theyshouldbeonlyasvisibleasnecessarytodothejob.Whatis theleastvisiblemeanstovisuallydelineategroupsofdata?Theansweriswhitespace.Whenenoughblank spacesurroundsagroupofdatatosetitapartfromtheothergroups,theobjectiveisaccomplished withoutaddinganyvisualcontenttothedashboardthatmightdistractattentionfromthedata.Usewhite spacetodelineategroupsofdatawheneverpossible. Ofcourse,asdashboardsareoftenhighdensitydisplays,theydonotalwayshavethesparespace necessarytousewhitespacealonetodelineatethegroups.Whenthatisthecase,subtlebordersare usuallythebestmeanstodistinguishthegroups.Youmightbesurprisedathowlightlinescanbeandstill dothejob.TakealookatFigure71foranexampleofhowyoucanusewhitespaceorlightbordersto delineatethesamegroupsofdata.

www.it-ebooks.info

Figure71.Thefourtablesonthetophavebeenseparatedeffectivelyusingwhitespacealone,butthefouronthebottom, becausetheyareclosertogether,havebeenseparatedusinglightborders.

7.1.4.SupportMeaningfulComparisons Measuresofperformancecomealiveonlywhenyoucomparethemtoothermeasures.Forexample, knowingthatquartertodatesalesrevenueis$92,354ismeaningfulonlywhencomparedtooneormore othermeasuresthatcanbeusedasyardstickstodetermineitsmerit,suchasatargetortheamountof revenuethathadcomeinatthispointinthepriorquarter.Youcanencouragemeaningfulcomparisonsby doingthefollowing: Combiningitemsinasingletableorgraph(ifappropriate) Placingitemsclosetooneanother Linkingitemsindifferentgroupsusingacommoncolor Includingcomparativevalues(forexample,ratios,percentages,oractualvariances)whenever usefulforclarityandefficiency

Figure72illustratestwoofthesepractices.Thegraphontopshowsseveralmeasuresthatsharethesame unitofmeasure,displayedinasinglegraphtoencouragecomparison.Thegraphonthebottomcombines twodatasetswithdifferentunitsofmeasureinasinglegraphbyplacingonequantitativescaleontheleft verticalaxisandanotherontheright.

Figure72.Twoexamplesofcombiningmultiplemeasuresinasinglegraphtoencouragecomparisons.

www.it-ebooks.info

ThetableinFigure73illustrateshowvaluescanbeexpresseddirectlyascomparativeunitsofmeasureto encouragecomparisons.Boththe"%ofTotal"and"%ofFcst"columnscontainvaluesthatarecomparative bytheirverynature.Especiallywhenyouwanttocommunicatethedegreetowhichonevaluediffersfrom another,percentagesexpressthismoredirectlythanrawvalues.

Figure73.Youcanusecomparativevaluestodirectlysupportcomparisons.

7.1.5.DiscourageMeaninglessComparisons Evenifit'sallimportanttosomejoborsetofobjectives,notallthedatathatappearsonadashboardis meanttobecompared.However,withoutvigilance,youmightinadvertentlymakedesignchoicesthat encouragethecomparisonofunrelateddata.Forinstance,inFigure74,someofthecolorchoicesproduce thisunintendedeffect.Thecolorsgreenandredmean"good"and"bad"wherevertheyappear,which encouragesustoassumethatallthecolorsusedonthisdashboardmeanthesamewherevertheyappear. However,thisisn'tthecasenoticethatthecoloryellowmeans"satisfactory"insomecontexts,butinone graphitrepresentsforecastbalancesandinanotherthemonthofJune.Inthiscase,ournaturalinclination tolinklikecolorsismisleading.

Figure74.Thisdashboardinadvertentlyencouragesmeaninglesscomparisons.

www.it-ebooks.info

Youcandiscouragemeaninglesscomparisonsbydoingtheoppositeofthepracticesmentionedinthe previoussection: Separateitemsfromoneanotherspatially(ifappropriate). Usedifferentcolors.

7.2.MaintainConsistencyforQuickandAccurateInterpretation
Differencesinappearancealwayspromptustosearch,whetherconsciouslyorunconsciously,forthe significanceofthosedifferences.Anythingthatmeansthesamethingorfunctionsinthesamewayought tolookthesamewhereveritappearsonadashboard.Evensomethingassubtleasarbitrarilyusingdark axislinesononegraphandlightaxislinesonanotherwillleadviewerstosuspectthatthisdifference, whichisinfactarbitrary,issignificant. It'simportanttomaintainconsistencynotonlyinthevisualappearanceofthedisplaymedia,butinyour choiceofdisplaymediaaswell.Iftwosectionsofdatainvolvethesametypeofquantitativerelationship (suchasatimeseries)andareintendedforsimilaruse(forexample,tocompareameasuretoatarget measureforeachmonth),youshouldusethesametypeofdisplayforboth(forexample,abargraph). Nevervarythemeansofdisplayforthesakeofvariety.Alwaysselectthemediumthatbestcommunicates thedataanditsmessage,evenifthatmeansthatyourdashboardconsistsofthesametypeofgraph throughout.

7.3.MaketheViewingExperienceAestheticallyPleasing
In1988DonaldNorman,acognitivescientist,wroteawonderfulbookentitledTheDesignofEveryday Things(NewYork:BasicBooks).Itisaclassicinthefieldofdesignthatconvincinglyarguesthatthe effectivenessofsomething'sdesignshouldbejudgedbyhowwellitworksandhoweasyitistouse.Inthe yearssinceitspublication,designershaveoftenaccusedNormanofignoringthevalueofaesthetics.This frequentcritiquewasoneofhismotivesforwritingtherecentbookentitledEmotionalDesign:WhyWe Love(orHate)EverydayThings(NewYork:BasicBooks,2004). Inthisbook,Normandescribesthepsychologicalandphysiologicalbenefitsofaestheticallypleasingdesign. Ifappliedtodashboarddesign,Norman'spointwouldarguethataestheticallypleasingdashboardsare moreenjoyable,whichmakesthemmorerelaxing,whichpreparestheviewerforgreaterinsightand creativeresponse.ThisisnotadeparturefromhisearlierassertionsinTheDesignofEverydayThings,but ratheranextensionassertingthataesthetics,whennotinconflictwithaproduct'susability,possess intrinsicqualitiesthatalsocontributetousability.Thisnewbookconvincinglyreframesthediscussion abouttheimportanceofusabilityasamatternotofusabilityversusaestheticsbutofusabilityversus anythingthatflagrantlyunderminesusability,whichgood,aestheticallypleasingdesignmanagestoavoid. Ilovevisualart.Iappreciatebeautyforitsownsake.Momentsofgreatbeautyexaltme.Information design,however,isaboutcommunication:gettinganintendedmessageacrossinawaythatresultsin usefulunderstanding.Aestheticsareanimportantcomponentofinformationdesign,butnotinthesame waythattheyareinart.Ifadashboardisnotdesignedinanaestheticallypleasingway,theunpleasant experiencethatresultsfortheviewerunderminesthedashboard'sabilitytocommunicate.Ona dashboard,youraesthetictalentoughttobeapplieddirectlytothedisplayofthedataitself,notto meaninglessanddistractingornamentation.Theaestheticsofdashboarddesignshouldalwaysexpress themselvessimply,strivingfortheeloquencethatemergesuniquelyfromsimplicity.

www.it-ebooks.info

ThedashboardshowninFigure75,whilesimpleenough,isaglaringexampleofdesignthatisanythingbut aestheticallypleasing.Howcanyouavoidcreatingasimilarmonstrosity?Let'slookatafewguidelinesthat willhelpyouachieveasimpleaestheticwithoutcompromisingthedata.

Figure75.Anexampleofadownrightuglydashboard.

7.3.1.ChooseColorsAppropriately Pooruseofcolorisperhapsthemostcommonoffensetoadashboard'sappearance.Colorsthatarebright ordarknaturallydemandmoreattention.Toomanybrightordarkcolorscanquicklybecomevisually exhausting.Whenselectingcolors,keepthefollowingguidelinesinmind: Keepbrightcolorstoaminimum,usingthemonlytohighlightdatathatrequiresattention. Exceptforcontentthatdemandsattention,uselesssaturatedcolorssuchasthosethatare predominantinnature(forexample,thecolorsoftheearthandsky). Useabarelydiscernablepalebackgroundcolorotherthanpurewhitetoprovideamore soothing,lessstarklycontrastingsurfaceonwhichthedatacanreside.

Figure76illustratestheseprinciples.

www.it-ebooks.info

Figure76.Avoidtheuseofbrightcolorsexcepttohighlightparticulardatastickwithmoresubduedcolorsformostofwhat's displayed.Useabackgroundcolorthatisslightlyoffwhitetoavoidthestarkcontrastbetweenforegroundcolorsagainstapure whitebackground.

7.3.2.ChooseHighResolutionforClarity Thehighdensityofinformationthattypicallyappearsonadashboardrequiresthatthegraphicalimagesbe displayedwithexceptionalvisualclarity.Imageswithpoorresolutionarehardtoread,whichslowsdown theprocessofscanningthedashboardforinformation(andisjustplainannoying).Visualclaritydoesnot requirefancyshadingorphotorealism;simplehighresolutionimageswilldo. 7.3.3.ChoosetheRightText Myfinalrecommendationregardingdashboardaestheticsinvolvestheuseoftext.Usethemostlegible fontyoucanfind.Youdon'tneedtosetamoodorreinforceathemebyusinganunusualfont.Ornatetext mightbeappropriateforaposteradvertisingthecircus,butnotforadashboard.Youwantafontthatcan bereadthefastestwiththeleastamountofstrainontheeyes.Findonethatworksandstickwithit throughoutthedashboard.Youcanuseadifferentfontforheadingstohelpthemstandoutifyouwish,but that'sthepracticallimit.Figure77illustratesafewofthegoodandbadchoicesthatareavailable.

Figure77.Examplesofsomefontsthatareeasytoreadyandsomethatarenot.

7.4.DesignforUseasaLaunchPad
Assinglescreendisplays,dashboardsdonotalwaysprovidealltheinformationneededtoperformajobor topursueaparticularsetofobjectives.Theycanprovidetheinitialoverviewthatisneededformonitoring atahighlevel,buttheymightneedtobesupplementedwithadditionalinformationformore comprehensiveunderstandingandresponse.Dashboardsshouldalmostalwaysbedesignedforinteraction. Themostcommontypesofdashboardinteractionare: Drillingdownintothedetails Slicingthedatatonarrowthefieldoffocus

www.it-ebooks.info

Whicheveroftheseyouintend,whenyourdashboardservesasalaunchpadtoadditional,complementary information,besuretokeepthefollowingprinciplesinmind: Allowtheviewertoinitiatethelaunchbyclickingthedataitself. Useconsistentlaunchactions.

Enablingtheviewertoaccessadditionaldata(suchasthedetailsbeneaththeoverview)viadirect interactioniseasyandintuitive,anditsavesspaceonthedashboardbyeliminatingseparatecontrolssuch asbuttons.Ifyoudisplayabargraphinwhicheachbarrepresentstherevenueofadifferentsalesregion, forexample,itmightbeidealtoallowtheviewertoclickdirectlyonaparticularbartoseeagraphthat furthersubdividesthatregion'srevenueaccordingtotheindividualstatesthatbelongtotheregion. Likewise,iftherearetimeswhenaviewermightwanttoknowtheprecisevalueforaparticulardatapoint alongalinegraph,theabilitytohoveroverthatpositionandhavethevaluepopuptemporarilyastextis ideal.Whatevermechanismyoudecidetobuildintothedashboardtoinitiatelinkstoadditionaldata,make surethatitisconsistentwhereveritappears,toavoidconfusion.

7.5.TestYourDesignforUsability
Nomatterhowwelldesignedyourfinalproductturnsouttobe,itisalwayshardtodissuadepeoplefrom predeterminednotionsofhowitshouldlook.Doyourbesttopreventthosewhowilleventuallyuseyour completeddashboardfromdevelopingexpectationsaboutitslookandfeelapartfromyourinputand expertadvice.Presentyouruserswithasingleprototypeofthemosteffectivedesignthatyoucancreate, andletthatbethestartingpointfordiscussionsabouthowitmightbetweakedtobetterservetheirneeds. Don'tpresentthemwithseveralalternativedesigns,becauseeventhoughyourusersprobablyknowwhat theyneedtoaccomplish,theydon'tknowhowthedashboardoughttobevisuallydesignedtoachievethat result.Youarethedesigner,soitisuptoyoutobringthisexpertisetotheprocess. Youwillnevergeteverythingrightonthefirsttry,nomatterhowskilledyouare.Youmustputyourdesign tothetest.Onlythosewhowillactuallyusethedashboardarequalifiedtodetermineifitactuallyworks andworkswell.Showittothempopulatedwithrealdata,andobservethemastheylookitoverandlearn tomakesenseofthedata.Ifyouareintroducingdisplaymediathatarenewtothem,beginwithsimple instructioninhowtheyworkandexplainwhyyouchosethosemechanismsratherthanothersthatmight bemorefamiliar.Ifyou'vedoneyourhomeworkandyourusersreallycareaboutdoingtheirjobswell ratherthandoingtheminaparticularway,usabilitytestingwillusuallyresultinrelativelyminoradditions andtweakstorefinetheeffectivenessofthedashboard,ratherthanmajorrevisions.Althoughthereare certainlyexceptionswhendealingwiththefoiblesofhumanbeings,gooddesignusuallyresultsinagood reception.

www.it-ebooks.info

Chapter8.PuttingItAllTogether
Agreatdealofinformationhasbeenamassedasthelessonsinthisbookhavebeenunveiledstepbystep, conceptbyconcept,andprinciplebyprinciple.Nowitistimetotieitalltogether,toseetheseprinciples combinedintheformofsampledashboards.Theproofisintheefficacyoftheresult:dashboardsthatcan bemonitoredandunderstoodataglance.We'lllookatfourexamplesofeffectivelydesigneddashboards, andputourknowledgetothetestbycritiquingeightalternatesolutionstooneofthesedesignproblems.

www.it-ebooks.info

Samplesalesdashboards SampleCIOdashboard Sampletelesalesdashboard Samplemarketinganalysisdashboard Inthisfinalchapter,we'llbringtogethertheprinciplesandpracticestaughtthroughoutthebook.We'll examinesomedashboardsthatillustratetheclearandefficientcommunicationthatresultsfrominformed design,andwe'lltestyourknowledgebycritiquingseveralothers.Thesesamplesaddressfourdifferent businessscenarios,includingdashboardsthatsupportstrategic,analytical,andoperationalpurposes: SamplesalesdashboardAsalesmanagermightusethisdashboardtomonitorsalesperformanceand opportunities(strategic). SampleCIOdashboardAChiefInformationOfficer(CIO)mightusethisdashboardtomonitorseveral aspectsofacompany'sinformationsystems(strategicandoperational). SampletelesalesdashboardThesupervisorofateamofsalesrepresentativeswhotakeordersandanswer questionsbyphonemightusethisdashboardtomonitorperformance(operational). SamplemarketinganalysisdashboardAmarketinganalystmightusethisdashboardtomonitorthe marketingperformanceofthecompany'swebsite(analytical). TheseexampleswillnotonlyputfleshonthebonesofthedesignprinciplesthatI'vetaughtinthisbook, but(Ihope)willalsosuggestideasforthetypesofinformationyoumightdisplayonadashboardandsome interestingandeffectivewaystodoso.

8.1.SampleSalesDashboard
Apartfromexecutivedashboards,Isuspectthatnoonetypeofdashboardisimplementedmoreoftenthan asalesdashboard.Salesactivityisthelifegivingheartofmostbusinesses.Thoseinchargeofsalesneedto keeptheirfingersonthepulseatalltimes,evenwhenalliswell.Salesstrategiesmightneedtochange quicklywhennewopportunities,problems,orcompetitivepressuresarise.Awelldesigneddashboardcan beapowerfultoolforasalesmanager. Ibegandesigningthesamplesalesdashboardbyselectingtheinformationthatseemedmostimportantfor asalesmanagertomonitor.EachitemthatIselectedisameasureofwhat'scurrentlygoingoninsales. Here'sthelist:1

1

Salesrevenue Salesrevenueinthepipeline(expectedrevenuedividedintocategoriesofprobability) Profit Customersatisfactionrating Top10customers Marketshare

Keepinmindthatthepurposeofthesamplesinthischapterisnottodefinethedatathatyoushouldincludeonany particulartypeofdashboard,butrathertoillustratehowthevisualdesignprinciplesthatyou'velearnedinthisbook canbeappliedtorealworldsituations,andhowtheymightlook.Itisn'tpossibletodeterminetheprecisedatathat willbeappropriateforalldashboardsofanyparticulartype,suchasasalesdashboard.

www.it-ebooks.info

Foreachoftheseitems,Ineededtomakeseveraldecisions,including:

AtwhatlevelofsummarizationshouldIexpressthismeasure? WhatunitofmeasureshouldIusetoexpressthismeasure? WhatcomplementaryinformationshouldIincludeascontexttoenhancethismeasure'smeaning? Whatmeansofdisplaywouldbestexpressthismeasure? Howimportantisthismeasuretoasalesmanagercomparedtotheothermeasures? Atwhatpointinthesequenceofviewingtheitemsonthedashboardmightasalesmanagerwant toseethismeasure? Towhatothermeasuresmightasalesmanagerwanttocomparethismeasure?

IfIweredesigningasalesdashboardforaparticularpersonorgroup,Iwouldinvolvetheminanswering thesequestions.Formypresentpurposes,however,Imadeseveralassumptionsbasedonmyknowledge ofsalesandproducedthedashboardinFigure81.

Figure81.Asamplesalesdashboardthatputsintopracticetheprincipleswe'vediscussedthroughoutthisbook.

Examinethisdashboardonyourown,througheyesthatcannowrecognizewhatworksandwhatdoesn't, withanunderstandingofwhy.Lookateachmeasure,atwhatIincludedascontext,andateveryaspectof thevisualdesign,bothonitsownandinrelationtothewhole.Askyourself,"Whywasitdesignedinthis way?"Takesometimenowtodothisbeforereadingon.Hopefully,you'llbeabletoidentifyandexplain thereasonsformostofmydesignchoices. Hereareafewofthehighlights: Colorhasbeenusedsparingly.Otherthanthelightbrownheadingstoclearlygroupthedatainto meaningfulsections,theonlyothercolorthatisnotagraytoneappearsontheredalerts.Thisjudicious useofcolormakesthoseitemsthatmustgrabattentiondosoclearly,withoutcompetitionfromother colorsthatmightalsoattractattention. Theprimerealestateonthescreenhasbeenusedforthemostimportantdata.Assumingthatthe measuresthathavebeenidentifiedasthe"keymetrics"aregenerallythemostimportantitemsonthe dashboard,placingthemintheupperleftcornerofthescreengivesthemtheprominencethatthey deserve. Small,concisedisplaymediahavebeenusedtosupportthedisplayofadensesetofdatainasmall amountofspace.Thisdashboarddisplaysagreatdealofinformation,yetitisn'tcluttered.Spaceefficient andsimpledisplaymediasuchassparklinesandbulletgraphsarerequiredtoachievethiseffect. Somemeasureshavebeenpresentedbothgraphicallyandastext.Peoplewhomonitorsalesactivityare

www.it-ebooks.info

generallyinterestedinknowingboththeactualsalesamountsandhowwellsalesaredoingcomparedto targets. Thedisplayofquartertodaterevenueperregioncombinestheactualandpipelinevaluesintheformof stackedbars.Thisapproachenablesviewerstoeasilyseetheresultofaddinganticipatedtoactualrevenue inrelationtothetarget. Whitespacealonehasbeenusedtodelineateandgroupdata.Borders,gridlines,andbackgroundfill colorsareunnecessaryandwouldseverelyclutterthescreen. Thedashboardhasnotbeenclutteredwithinstructionsanddescriptionsthatwillseldombeneeded.A singlehelpbuttonhasbeenprovidedtoallowtheviewertoaccessinformationthatwillprobablybe neededonlyonceortwice,atthebeginningofthedashboard'suse. Lookingatthissampledashboard,youmightseewaysthatdifferentchoicescouldhavebeenmadeto furtherimproveitseffectiveness.Ifullyexpectandevenhopetoreceivefeedbackfromreaderslikeyouto pointoutimprovementsthatcouldbemade. Youmightfinditusefultocomparemysalesdashboardtoseveralothersthatweredesignedtomeetthe sameexactsetofrequirements.IrecentlyjudgedadatavisualizationcompetitionforDMReview magazine.Oneofthefourbusinessscenariosthatparticipantswereaskedtoaddresswithdata visualizationsolutionsrequiredasalesdashboardwiththesamemeasuresthatIincludedinmine.The contestantsweregiventherequirementswithoutanydesigninstructionorsamplesolutions.I'dliketo showyouafewofthesolutionsthatweresubmitted,allofwhicharequitedifferentfrommine.Examine themtojudgehowthechoicestheirdesignersmademighthavebeenimproved.Ibelievethatbydoingthis youwillseehowapplyingthedesignprinciplesthatyou'velearnedinthisbookwillofferclearadvantages overtheseotherapproaches. I'veincludedafewcommentsfollowingeachofthesealternativesalesdashboardsolutions,buttakethe timetoexamineeachofthemonyourownbeforereadingmycritique.Thiseffortwillstrengthenyour understandingofdashboarddesignandhelptomoreseamlesslyintegratetheprincipleswe'vecovered intoyourthinking.Ihaven'tbotheredtolisteveryoneoftheproblemsthatI'vediscoveredineachofthe dashboards,buthavefocusedprimarilyonuniqueproblems.

www.it-ebooks.info

CritiqueofSalesDashboardExample1

Figure82.Thistextbasedsamplesalesdashboardcouldbeimproved.

Thissalesdashboardusesanapproachthatreliesalmostentirelyontexttocommunicate,usingvisual meansonlyintheformofgreen,lightred,andvibrantredhuestohighlightitemsas"good,""satisfactory," or"poor."Expressingquantitativedatatextuallyprovidesprecisedetail,butthisisn'tusuallythepurposeof adashboard.Dashboardsaremeanttoprovideimmediateinsightintowhat'sgoingon,buttextrequires readingaserialprocessthatismuchslowerthantheparallelprocessingofavisuallyorienteddashboard thatmakesgooduseofthepreattentiveattributesofvisualperception. Tocompareactualmeasurestotheirtargets,mentalmathisrequired.Graphicalsupportofthese comparisonswouldhavebeeneasierandfastertointerpret. Numbershavebeencenterjustifiedinthecolumns,ratherthanrightjustified.Thismakesthemharderto comparewhenscanningupanddownacolumn. Someimportantmeasuresaremissing.Thisdashboarddoesnotincludepipelinerevenueorthetop10 customers. Allfourquartersofthecurrentyearhavebeengivenequallevelsofemphasis.Asalesmanagerwouldhave greaterinterestinthecurrentquarter.Thedesignofthedashboardshouldhavefocusedonthecurrent quarterandcomparativelyreducedemphasisontheotherquarters. Propercarehasnotbeengiventomakeimportantdistinctions.Thegreaterintensityofthevibrantredhue thatisusedtohighlightmeasuresthatareperformingpoorlywillstandoutclearlyeventocolorblind users,butthesubduedshadeofredandtheequallysubduedshadeofgreenmightnotbedistinguishable.

www.it-ebooks.info

Also,thenumbersthatoughttostandoutmostandbeveryeasytoreadthepoorlyperformingmeasures arethehardesttoreadagainstthedarkredbackground. CritiqueofSalesDashboardExample2

Figure83.Thissolutionexhibitssomeofthesameproblemsasthepreviousexample,butalsoafewdifferentones.

Thegridlinesthatappearinthetablesarenotneededatall.Eveniftheywereneeded,theyshouldhave beenmutedvisually.Intheircurrentheavyform,theyimprisonthenumbers. Thegridlinesthatappearinthegraphsarealsounnecessary.Theydistractfromthedata.Especiallyinthe contextofadashboard,youcan'taffordtoincludeanyunnecessaryvisualcontent. Thedropshadowsonthebarsandlinesintwoofthegraphsandonthepiechartarevisualfluff.These elementsserveonlytodistract.

www.it-ebooks.info

Allofthenumbersinthetableshavebeenexpressedaspercentages.Ifthosewhousethisdashboardonly careaboutperformancerelativetotargets,thisisfine,butitislikelythattheywillwantasenseofthe actualamountsaswell. Thepiechartisnotthemosteffectivedisplaymedium.Assumingthatitisworthwhiletodisplayhowthe 90%probabilityportionoftherevenuepipelineisdistributedamongtheregions,abargraphwiththe regionsinrankedorderwouldhavecommunicatedthisinformationmoreeffectively. Overall,thisdashboardexhibitstoomanybrightcolors.Thedashboardasawholeisvisuallyoverwhelming andfailstofeaturethemostimportantdata. Thereisnocomparisonoftrendsintherevenuehistory.The12monthrevenuehistoryshownintheline graphisuseful,butitwouldalsohavebeenusefultoseethishistoryperregionandperproduct,toallow thecomparisonoftrends. CritiqueofSalesDashboardExample3

Figure84.ThissolutionillustratesseveraloftheproblemsthatIpointedoutinChapter3,ThirteenCommonMistakesin DashboardDesign.

Thisdesignfragmentsthedatathatasalesmanagerwouldwanttoseeintoseparateinstancesofthe dashboard.Noticetheradiobuttonsabovethegraphontheleft,whichareusedtoselectthequarterthat youwanttosee.Thisgivesyounomeanstocomparesalesperformanceovertime. Thephotographsarechartjunk(atermcoinedbyEdwardTuftetodescribevisualcontentinaninformation displaythatservesonlyasdecoration).Thisuselessdecorationservesonlytodistractfromthedata.After seeingthesefacesforacoupleofdays,viewerswilltireofthemandwishthespacehadbeenbetterused.

www.it-ebooks.info

Furthermore,themostimportantrealestateonthescreen(atthetopleft)istakenupbyphotographsand acompanylogo.Thisisawasteofvaluablespace. Thebargraphintheupperleftfailstovisuallydisplayclearcomparisonstothetargets.Youmustreadthe numbersprintedonthebarstodeterminetherelationshipstothetargets. Thetwographsontherightmakeanattempttovisuallycomparetherevenuemeasurestotheirtargets, buttheyusealinetoencodethetargets,whichisinappropriateforthisdata.Usingalinetoconnectvalues inagraphsuggestsarelationshipofchangebetweenthevalues,butrevenuevaluesforindividualproducts orregionsarenotintimatelyconnectedtooneanothertheyarediscretevaluesalonganominalscale.The patternsformedbythelinesaremeaningless. CritiqueofSalesDashboardExample4

Figure85.Thisexampleusedheadacheinducingcolors.

Theuseofcoloristoodramatic,especiallyintheareaswiththedarkbackgrounds.Alight,slightlyoffwhite backgroundthroughoutwouldhaveworkedbetter.Also,theuseofextremelydifferentbackgroundcolors toseparatethedataintofoursectionsisn'tnecessary. Whitespaceisoverused.Ratherthansurroundingthetwotablesontheleftinalargeamountofwhite space,thetablescouldhavebeenenlargedtomakethemeasiertoread.

www.it-ebooks.info

CritiqueofSalesDashboardExample5

Figure86.Thissolutionexhibitssomeoftheproblemsfoundinpreviousexamples,andafewothers.

Onceagain,wehaveadesignthathasfragmentedthedata.Noticetheradiobuttonsorslidersnexttoeach ofthegraphs.Wecanonlyseeonemeasureatatimeineachgraph,yetmuchofthisdataoughttobe displayedtogethertoenableustomakeusefulcomparisons(suchasbetweentheregions). Thebeautiful,brightlycoloredpiechartslooksomuchlikecandy,Igetasugarrushjustlookingatthem. Thecolorsaremuchtoobright,andthephotorealisticshadingtogivethema3Dappearanceissimplynot necessary.Thiseffectmakesthepiechartsjumpoutasthedominantfeaturesofthedashboard,whichis notwarranted.Also,onceagain,piechartsarenotthemosteffectivemeansofdisplayingdataona dashboard,becausetheydon'tallowforcomparisonsaseasilyasbargraphs. Thevisualshadingonthebarsandbuttons,likethatonthepiecharts,isunnecessaryanddistracting.This contributestotheeffectofmakingtheseobjectspopoutinappropriately.

www.it-ebooks.info

CritiqueofSalesDashboardExample6

Figure87.Whilevisuallyappealinginsomeways,thissolutionhassomeseriousweaknesses.

Despitethevisualappealofthelefthalfofthisdashboard,thedisplaymediawerenotwellchosen.The circularrepresentationsoftimeseriesdatausinghuestoencodestatesofperformance(good,satisfactory, andpoor)areclever,butforthepurposeofshowinghistory,thesearenotasintuitiveorinformativeasa lineardisplay,suchasalinegraph. Noneofthemeasuresthatappearontheleftsideofthedashboardisrevealedbeyonditsperformance state.Knowingtheactualrevenueamountandmoreabouthowitcomparestothetargetwouldcertainly beusefultoasalesmanager.Unlikesomeofthepreviousexamplesthatusedhuestoencodestatesof performance,however,Ibelievethatthesehueswerecarefullychosentoberecognizablebythosewho arecolorblind. Thecirculardisplaymechanismstreatallperiodsoftimeequally.Thereisnoemphasisonthecurrent quarter. Gradientfillcolorsinthebargraphsaddmeaninglessvisualinterest.Theyalsoinfluenceperceptionofthe valuesencodedbythebarsinsubtleways.Barsthatextendintothedarkersectionsofthegradientappear slightlydifferentfromthosethatextendonlyintothelightersections.Dashboarddesignersshouldbe consciousofeventhesesubtleeffectsandavoidthem.

www.it-ebooks.info

CritiqueofSalesDashboardExample7

Figure88.Thisdashboardhasaproblemthatwehaven'tseensofarthatunderminesitseffectiveness.

Someofthesegraphsaretoocomplexforeasyinterpretation.Therevenueperformancebyproductand regiongraphsatthelowerleftandthequartertodatesalespipelinebyregiongraphinthecenterbottom positionallusebarsthatencodevaluesintwodimensions,usingboththeheightandwidthofeachbar. Thisisaworthwhileattempttosavespace,butonethatrequirestoomuchstudytointerpretdueto limitationsinvisualperception.ThetwographsontheleftbothusetheX(horizontal)axistoencode revenueperformancecomparedtotarget,andtheY(vertical)axistoencodetheportionofeachproductor regiontothewhole,functioninglikestackedbargraphs.Thepipelinerevenuegraphinthecenterdisplays thedifferentpartsofthepipeline(90%probability,etc.)assegmentsofthebarrunninghorizontallyfrom lefttoright,andtheregionalportionsofthetotalpipelineasverticalsegments.Usingboththeheightand widthofthebarstoencodequantitativevaluesrectanglesthattemptustocomparetheir2Dareastoone anotherresultsininaccuratecomparisons.

www.it-ebooks.info

CritiqueofSalesDashboardExample8

Figure89.Thisfinalexampleisquiteadeparturefromtheothersandhassomeserious(andprobablyobvious)flaws.

Thisdashboard,whileaninterestingcontrasttotheothers,isconfusingatfirstglanceandlikelytoremain thatwayforsomewhile.Muchofthedataitpresentsisalsofairlyimprecise.Colorsandshapeshavebeen usedtoencodevaluesintherectanglesthatappearthroughoutthedashboard.Intheupperrightcorner, youfindalegendthattellsyouwhateachoftherectanglesrepresents(revenue,marketshare,profit,etc.) whereveryouseethemarrangedinthisparticularconfiguration.Althoughakeyforthemeaningofthe variouscolorsandshapesthatappearintherectanglesdoesnotappearonthedashboard,thekeyshown inFigure810wasprovidedseparatelywhenitwassubmittedforthecompetition.Youcouldcertainly memorizethemeaningsofthevariousrectanglelocationsandofthecolorsandshapesinsidethem,but evenafterthateffort,theserectangleswouldstillnevergiveyoumorethanaroughsenseofhowthe measurescomparetotheirtargets.Forinstance,seeingthesemeasuresencodedinthiswayandarranged sidebysidetorepresentmonthsorquartersdoesnotcomeclosetoprovidingtheunderstandingof historicaltrendthatasimplelinegraphcouldconvey.

www.it-ebooks.info

Figure810.KeyforinterpretingthedatainFigure89.

Nowthatyou'vetakenthislittletourthroughseveralsolutionstothesamedashboarddesignchallenge,go backandtakealookoncemoreatthedashboardinFigure81.Asyoucansee,thereisaneloquenceto datadisplayedsimplythatcannotbeachievedifwestrayfromtheessentialgoalofcommunication.

8.2.SampleCIODashboard
AChiefInformationOfficermustkeeptrackofmanyfactsregardingtheperformanceofthecompany's informationsystemsandactivities,includingprojectsthatservethecompany'sinformationneeds.Ichose toincludethefollowingdatainmysampledashboard:

Systemavailability(uptime) Expenses Customersatisfaction Severeproblemcount CPUusagerelativetocapacity Storageusagerelativetocapacity Networktraffic Applicationresponsetime Majorprojectmilestones Topprojectsinthequeue Othercriticalevents

ThisisamixtureofstrategicandfrequentlyupdatedoperationalinformationthataCIOmightneed. ExamineFigure811closelyandtrytogetasenseforhowitmightworkintherealworld. Onlyonesectionofthisdashboardtheupperleftcornerdisplaysnearrealtimedata.Thissectionconsistsof aseriesoffivealerts:oneforeachofthesystemsthattheCIOmightneedtorespondtoimmediatelywhen aproblemarises.Ifnoredcirclesappearinthissection,nothingcriticaliscurrentlywrongwithanyofthese systems.TobettergrabtheCIO'sattention,redalertsthatappearinthissectioncouldblinkuntilclicked,or

www.it-ebooks.info

evenemitalongwiththeblinksasoundthatgraduallyincreasesinvolume.Theredalertobjectscouldalso serveaslinkstootherscreensthatdescribepreciselywhatiswrong. TherestofthisdashboardprovidestheCIOwithinformationthatismorestrategicinnature.Noticethata greatdealofcontextualinformationhasbeenprovidedtocomplementthemeasuresespecially comparisonstomeasuresofacceptableperformance.ThisisthekindofcontextthatcouldhelptheCIO easilymakesenseofthesemeasures.

Figure811.AsampleCIOdashboard.

Thereisagreatdealofinformationonthisdashboard,yetitdoesn'tseemcluttered.Thisislargelydueto thefactthatnondatapixelshavebeenreducedtoaminimum.Forinstance,whitespacealonehasbeen usedtoseparatethevarioussectionsofthedisplay.Ajudicioususeofcolorhasalsocontributedtothis effect.Besidesgrayscalecolors,theonlyotherhuesyouseeareamutedgreenforthenameofeach sectionandtwointensitiesofred,whichineverycaseservesasanalert.Itiseasytoscanthedashboard andquicklyfindeverythingthatneedsattention,becausetheredalertobjectsareunique,visuallyunlike anythingelse. Includinginformationaboutprojectmilestones,pendingprojects,andothercriticaleventsonthis dashboardnotonlylocatesallthemostimportantinformationtheCIOneedsinoneplace,butalso supportsusefulcomparisons.Beingremindedaboutcomingeventsthatmightaffectexistingsystemsand

www.it-ebooks.info

beingabletolookimmediatelyatthecurrentperformanceofthosesystemscouldraiseusefulquestions abouttheirreadiness.

8.3.SampleTelesalesDashboard
Thissampledashboardwasdesignedtomonitorrealtimeoperationssothatatelesalessupervisorcan takenecessaryactionswithoutdelay.Thisisn'tadashboardthat'slikelytobelookedatonceaday,butone thatwillbekeptavailableandexaminedthroughouttheday.Itdoesn'tdisplayasmanymeasuresasthe examplesyou'veseensofarinthischapter,becausetoomanymeasurescanbeoverwhelmingwhenthe dashboardisusedtomonitorrealtimeoperationsthatrequirequickresponses.Onlythefollowingsix measuresareincluded:

Callwaittime Callduration Abandonedcalls(thatis,callerswhogottiredofwaitingandhungup) Callvolume Ordervolume Salesrepresentativeutilization(representativesonlinecomparedtothenumberavailable)

That'sitandthat'splentyforadashboardofthistype. Imaginethatyou'reresponsibleforateamofaround25telesalesrepresentativesandareusingthe dashboardinFigure812tokeepontopoftheiractivitiesthroughouttheday. Theprimarymetricsthatyoumustvigilantlymonitorarethelengthoftimecustomersarewaitingto connectwithasalesrepresentative,thelengthoftimesalesrepresentativesarespendingoncalls,andthe numberofcustomerswhoaregettingdiscouragedandhangingupwhilewaitingtogetthrough.Becauseof theirimportance,thesethreemetricsarelocatedintheupperleftcornerofthedashboardandare extremelyeasytoread. Whenproblemsarise,suchasthelengthyholdtimesandexcessivelylengthycallsshowninthisexample, youmustquicklydeterminethecausebeforetakingaction.Thisiswhenyouwouldswitchyourfocustothe performanceoftheindividualsalesrepresentatives,whichyoucanseeontherightsideofthedashboard. Individualsarerankedbyperformance,withthoseperformingpoorlyatthetopandaredrectangle highlightingthosewhoareperformingoutsidetheacceptablerange.

www.it-ebooks.info

Figure812.Asampletelesalesdashboard.

Asadashboardformonitoringrealtimeoperations,thedatawouldprobablychangewithupdatesevery fewseconds.Thiscanbedistractingwhenyou'retryingtofocusonaproblem,however,soa"Freeze Data/UnfreezeData"buttonhasbeenprovidedtotemporarilyputahalttoupdates.Whenupdatesare frozen,thebuttonshinesyellowtoremindyouofthisfact.Ifthedisplayremainsfrozenfortoolong,the buttonbeginstoblinkwithabrighteryellowuntilclickedtoonceagainallowupdates.Whenalertsfirst appear(theredcircles),theyblinktoattractattentionandperhapsevenemitanaudiosignaltoalertyouif youaren'twatchingthescreen.Tostopthesesignals,youclicktheredalert.Toremindyouthatyou've blockedthealertsfromprovidingurgentsignals,the"ResetAlerts"buttonturnsyellow,andafterawhile beginstoblink.Onceclicked,allalertscanonceagainsignalurgentconditionsifnecessary.

8.4.SampleMarketingAnalysisDashboard
Thelastsampledashboardwe'lllookatisanexampleofonethatsupportsanalysis(Figure813).Likeall dashboards,itisusedtomonitortheinformationneededtodoajob,butinthiscasethatjobhappensto primarilyinvolveanalysis.Dashboardscanprovideausefulmeansforanalyststowatchovertheirdomains andspotconditionsthatwarrantexamination.Ideally,theycanalsoserveasdirectlaunchpadstothe additionaldataandtoolsnecessarytoperformcomprehensiveanalyses. Thisparticularscenarioinvolvesananalystwhoseworksupportsthemarketingeffortsofthecompany's website.Shemonitorscustomerbehavioronthesitetoidentifybothproblemsthatpreventcustomers fromfindingandpurchasingwhattheywantandopportunitiestointerestcustomersinadditional

www.it-ebooks.info

products.Toexposeactivitiesonthewebsitethatcouldleadtoinsightifstudiedandunderstood,the followingdataappearsonthedashboard:

Numberofvisitors(daily,monthly,andyearly) Numberoforders Numberofregisteredvisitors Numberoftimesindividualproductswereviewedonthesite Occasionswhenproductsthatweredisplayedonthesamepagewererarelypurchasedtogether Occasionswhenproductsthatwerenotdisplayedonthesamepagewerepurchasedtogether Referralsfromotherwebsitesthathaveresultedinthemostvisits

Theinformationthatappearsatthetopofthisdashboardprovidesanoverviewofthewebsite's performancethroughtimeandlistsmissedopportunitiesandineffectivemarketingefforts.Noticethatthe timeseriesinformationregardingvisitorstothesiteissegmentedintothreesections,eachfeaturinga differentintervaloftime.Theintervalshavebeentailoredtorevealgreaterdetailfortherecentpastand increasinglylessdetailthefartherbackthedatagoes. Muchoftheinformationonthisdashboardhasbeenselectedandarrangedtodisplayaranking relationship.Thisiscommonwhenadashboardisusedtofeatureexceptionalconditions,bothgoodand bad.Muchofthisrankedinformationiscommunicatedintheformoftext,withlittlegraphicalcontent. Giventhepurposetoinformtheanalystofpotentialareasofinterestwithabriefexplanationofwhy,text doesthejobnicely.Theanalystmustreadeachentrytodecideifshe'llinvestigatethematter,butgraphical displays,whichcouldbescannedfaster,wouldnotdothejobaswell.Thefactthatanitemappearsonone oftheselistsalreadyimpliesitsimportance,sographicaldevicessuchasalertswouldaddnothing.

www.it-ebooks.info

Figure813.Asamplewebmarketinganalysisdashboard.

8.5.AFinalWord
Todesigndashboardsthatreallywork,youmustalwaysfocusonthefundamentalgoal:communication. Morethananythingelse,youmustcarethatthepeoplewhouseyourdashboardscanlookatthemand understandthemsimply,clearly,andquickly.Dashboardsdesignedforanyotherreason,nomatterhow impressiveorentertaining,willbecometiresomeinafewdaysandwillbediscardedinafewweeksandfew thingsaremorediscouragingthanhavingyourhardworktossedasideasuseless. WhenIdesignsomethingthatmakespeople'slivesbetter,helpsthemworksmarter,orgivesthemwhat theyneedtosucceedinsomethingthatisimportanttothem,Iamremindedthatoneofthegreat cornerstonesofalifeworthlivingisthejoyofdoinggoodwork.Thisdoesn'tjusthappen;itistheresultof effortthatyoumakebecauseyoucare.Yourdashboardsmaynotchangetheworldinanybigway,but anythingyoudowellwillchangeyoutosomedegreeforthebetter.Evenifthebusinessgoalsthatyou're helpingsomeoneachievethroughawelldesigneddashboarddon'tultimatelymattertoyouorarenot intrinsicallyworthyofgreateffort,you'reworththeeffort,andthat'senough.Infact,that'splenty.

www.it-ebooks.info

AppendixA.RecommendedReading
BooksbythreeauthorsinparticularstandoutascomplementarytotheinformationthatI'vepresented aboutdashboarddesign,andeachdeservesaplaceinyourlibrary: WayneW.Eckerson,DirectorofResearch,TheDataWarehousingInstitute(TDWI). PerformanceDashboards:Measuring,Monitoring,andManagingYourBusiness (Indianapolis,IN:WileyPublishing,Inc.,2005) Wayneisoneofthetopindustryanalystsfocusedonbusinessintelligenceanddatawarehousing.Inhis book,hecoversseveralaspectsofdashboardsthatfalloutsideofmyexclusiveconcentrationonvisual design,includinghowtheycanbeusedtoimprovebusinessperformance. EdwardR.Tufte,ProfessorEmeritusatYaleUniversity TheVisualDisplayofQuantitativeInformation(Cheshire,CT:GraphicsPress,1983) VisualExplanations(Cheshire,CT:GraphicsPress,1990) EnvisioningInformation(Cheshire,CT:GraphicsPress,1997) BeautifulEvidence(Cheshire,CT:GraphicsPress,2006) NooneinrecenthistoryhascontributedmoretoourunderstandingofvisualinformationdisplaythanDr. Tufte.Allofhisbooksarebeautifullydesigned,eloquentlywritten,andoverflowingwithinsights. ColinWare,DirectoroftheDataVisualizationResearchLaboratory,UniversityofNewHampshire InformationVisualization:PerceptionforDesign,SecondEdition(SanFrancisco,CA: MorganKaufmannPublishers,2004) Whatweknowtodayaboutvisualperceptioncomesfromtheworkofmanyresearchersfrommany scientificdisciplines,butDr.Wareappliesthisknowledgetothevisualpresentationofinformationbetter thananyoneelse.

www.it-ebooks.info

Colophon
Genevieved'EntremontwastheproductioneditorforInformationDashboardDesign.RachelWheelerwas thecopyeditor.ClaireCloutierprovidedqualitycontrol.SpecializedComposition,Inc.providedproduction services. StephenFewdesignedthecoverofthisbook.KarenMontgomeryproducedthecoverlayoutinAdobe InDesignCS,usingSabonandNewsGothicCondensedfonts. MikeKohnkeandTerriDriscolldesignedtheinteriorlayout.ThetextfontisSabon,andtheheadingfontis NewsGothicCondensed.Theoriginalillustrationsthatappearinthisbookwereproducedbytheauthor, StephenFew,usingMicrosoftExcelandAdobeIllustratorCS.

www.it-ebooks.info

You might also like