Professional Documents
Culture Documents
P R O J E C T G U I D E B O O K
TABLE TABLE OF OF CONTENTS
C ONTENTS
MODIFY
REDESIGN Redesign — Revise current measurement
DESIGN?
systems if incorrect
σ
2
Defects per
million
opportunities
2 308,537
σ % Non-Defective
3
4
66,807
6,210
} American
Companies
2
3
4
69.1%
93.32%
99.379%
5 99.9767%
}
Texas
5 233 Instruments 6 99.99966%
& Motorola
6 3.4 are here
PROJECT SELECTION
P ROJECT S ELECTION
PROJECT SELECTION
Project Selection Project Selection Checklist
Proper Project Selection is one of the most To maximize the success of a project, you
critical success factors influencing the outcome should be able to answer “yes” to the following
of Consumer Driven 6-Sigma projects. questions:
Selecting a project that is too large will cause • Recurring events . . . . . . . . . . . . . . . . . . . .❏
✓
valuable time to be lost during the Define • Narrow scope . . . . . . . . . . . . . . . . . . . . . .❏
✓
Phase as Project Teams struggle to scope their • Available metrics, or measurements that
projects and develop “Problem Statements” can be developed quickly . . . . . . . . . . . . .❏ ✓
that can be addressed using 6-Sigma tools. • Control of the process . . . . . . . . . . . . . . . .❏ ✓
• Customer Satisfaction Improvement . . . . .❏ ✓
• Cost Reduction – $250K . . . . . . . . . . . . . .❏ ✓
3
PROJECT SELECTION
The 'Y' and 'X' Relationship Building on this concept, it can be said that
every “X” is correlated to a “Y” at some deeper
A fundamental formula used in 6-Sigma training level in the process. The diagram for this
is that “Y” is a function of “X.” cascading relationship is as follows:
Y = (ƒ) X
Y = X1 X2 X3 X4
In mathematical terminology, “Y” is the
dependent variable while “X” is the independent Y = X1 X2 X3 X4
variable. Expressed in another way, “Y” is an
output resulting from a number of inputs or Y = X1 X2 X3 X4
“Xs.”
This method of “drilling down” to the specific
The typical diagram used to communicate this issues is the fundamental technique for both
concept in Consumer Driven 6-Sigma training Project Selection and Project Scoping. Those
is as follows: trained in 6-Sigma techniques are often heard
describing issues in terms of the “Xs” and “Ys.”
Y = X1 X2 X3 X4
Example: Lemonade
The taste of a glass of lemonade 4
(the “Y”) is influenced by a number
of factors including:
• Type of lemon ingredient (an “X”)
• Amount of sugar added (an “X”)
Y = X1 X2 X3 X4
• Type of water used (an “X”)
• Amount of ice added (an “X”)
Y = X1 X2 X3 Y = The top line issue is what can be done to improve satisfaction, which can be broken down into subgroups:
X1 = Consumers, X2 = Dealers, X3 = Internal Customers
Y = X1 X2 Y = The decision is made to address Dealer Satisfaction, which is comprised of satisfaction with:
X1 = Processes, X2 = Product
High Level Outcomes
X1 = Condition, X2 = Timing
N o t e : D e l i v e r y C o n d i t i o n s a n d D e l i v e r y Ti m i n g w i l l b e u s e d t h r o u g h o u t t h i s G u i d e b o o k a s
e x a m p l e p r o j e c ts t o e x p l a i n 6 - S i g m a t o o l s a n d t e c h n i q u e s .
PROJECT SELECTION
Applying the Project becomes difficult to pinpoint what data to
collect and how to make use of it.
Selection Checklist
To ensure that projects are appropriately
Project Selection is most often performed at the selected, the Champion can apply the Project
Project Champion level, and it is critical that Selection Checklist appearing on Page 3.
Black Belts and Project Teams have specific
issues to address. The following shows how the example projects
used in this Guidebook compare to the Project
Without clear direction, projects frequently stall Selection criteria.
in the Define and Measure Phases as it
■ Recurring events? Yes, vehicles delivered ■ Control of process? Yes, Project Team is
each day. empowered to make improvements.
■ Narrow scope? Yes, the vehicle delivery ■ Customer Satisfaction? Yes,
7
DEFINE PHASE D EFINE P HASE
Problem Statement
Dealers are dissatisfied with the condition of vehicles when they are delivered to the dealership.
Identify Customer
To identify a customer for this 6-Sigma project, the Project Team reviews a random sample of
exception claims submitted by dealerships over a period of six months for repairs made to
vehicles delivered in a condition unacceptable for sale.
Based on this review, the Project Team identifies 5 large-volume dealerships in the Southwest
Region that have submitted a larger number of claims when compared to dealerships overall.
Targeting this smaller subset of dealerships increases the chances of success, as the scope of
the project is manageable and improvements will have a noticeable impact.
10
11
DEFINE PHASE D E F I N E P H A S E | Identify CTQs
Identify
CTQs
To determine CTQs, the Project Team interviews the dealership Inventory Manager and pre-
delivery personnel.
CTQ
New vehicles are delivered to the dealership free of mechanical and cosmetic defects.
OPERATIONAL DEFINITION
New vehicles arriving at the dealership will exhibit no condition that requires the expenditure
of materials and/or labor hours exceeding what is indicated as part of the standard
pre-delivery process for that model. This includes mechanical repairs and adjustments as
well as cosmetic preparation in excess of washing required to render the vehicle acceptable for
retail consumer sale.
13
Delivery Timing — Example “B”
To determine CTQs, the Project Team issues a survey to General Managers and Sales
Managers for dealerships in the Los Angeles Region.
CTQ
Vehicles will arrive at the dealership within 35 days of date of order.
OPERATIONAL DEFINITION
Vehicles ordered to fulfill consumer requests, or as part of standard stock replenishment, will
be received into dealership inventory on or earlier than the 35th day subsequent to the date
registered as part of online order submission. All days, except those recognized as holidays in
the union contract, will be included in the count.
DEFINE PHASE D E F I N E P H A S E | Map Process
Map
Process
14
DEFINE PHASE
Map Process
15
DEFINE PHASE D E F I N E P H A S E | Scope Project
Scope
Project
Scope Project
The Project Team extends the Y/X chart used during project selection to further define their
6-Sigma project.
•
VEHICLES
Y = X1 X2
Co Ti
nd m
iti in
on g
CONDITION
Y = X1 X2
M Co
ec sm
ha
etni
ic ca
l
COSMETIC
Y = X1 X2
In Ex
te te 17
rio rio
r r
18
19
1. Input the Minitab spreadsheet columns that contain information of possible causes of damage,
one column per branch
2. Input label information about where the causes of damage might be located
3. Input the effect that results from the causes
4. Input a title for the chart
5. Click on checkbox options as desired
6. Click on OK
DEFINE PHASE Scope Project
20
The resulting Cause-and-Effect Diagram offers a visual representation of potential sources and
causes of Exterior Vehicle Damage. Information contained in this chart is very useful in
developing a Data Collection Plan (Measure Phase) and can often be helpful in writing the
problem statement.
NOTE: Based on information gathered during the Define Phase, the Project Team can now refine
the Problem Statement.
Scope Project
The Project Team extends the Y/X chart used during Project Selection to further define their
6-Sigma project.
•
VEHICLES
Y = X X
C 1 T 2
on im
di in
tio g
n
TIMING
Y = X X
C 1 T 2
ar ru
s ck
s
TRUCKS
Y = X1 X2
Pi SU
ck
up V 21
SUV
Y = EX1 X2 X3
xc Exp Exp
ur ed lo
si
on itio rer
n
EXPLORER
Y = X1 X
L 2
St ou
. Lo is
ui vi
s lle
22
MEASURE PHASE
Develop
M EASURE P HASE
Perform
Conduct Perform Conduct
Data Measurement
Data Graphical Baseline
Collection System
Collection Analysis Analysis
Plan Analysis
MEASURE PHASE M EASURE P HASE
23
MEASURE PHASE
M EASURE P HASE | Develop Data Collection Plan
Develop
Data
Collection
Plan
In developing the Data Collection Plan, the • What model of vehicle? (check box
Project Team: for vehicle models)
1. Reviews and confirms validity of the Refined • Where was the damage located?
Problem Statement and CTQ Operational (schematic)
Definition. • When was damage noted? (match
process map)
Problem Statement
• Vehicle Identification Number? (VIN)
Los Angeles Region dealerships’ dissatisfaction
with delivery times exceeding 35 days in • Receiving dealership
connection with Explorer vehicles shipped from the • Why data is needed
Louisville Assembly Plant. (See page 21.)
– To determine whether any trends,
consistencies or correlations exist
OPERATIONAL DEFINITION
Vehicles ordered to fulfill consumer requests, or • How data will be collected
as part of standard stock replenishment, will be – A data collection form will accompany
received into dealership inventory on or earlier 50-100 vehicles destined for any of the
than the day subsequent to the date registered
as part of online order submission. All days,
five large-volume dealerships in the
25 except those recognized as holidays in the union Southwest
contract, will be included in the count. • Where data will be collected
– Data will be collected at each of the
2. Generates responses to the Data Collection three inspection points indicated on
Plan questions: the Process Map (Holding Yard
receiving inspection, Holding Yard
• What data will be collected (including type)
dispatching inspection, dealership
– Attribute Data indicating: receiving inspection)
• Was damage found? (yes/no) • Who will collect data
• What type of damage was found? – Inspectors as indicated on the
(check boxes for type) Process Map
MEASURE PHASE Develop Data Collection Plan
26
3. Produce a data collection tool and develop inclusion of data collection forms with
a process for using it to gather information. transportation paperwork of vehicles being
shipped to the targeted dealerships (5) in the
NOTE: The Project Team establishes a Southwest Region.
process based on Dealer Codes that ensures
MEASURE PHASE Develop Data Collection Plan
OPERATIONAL DEFINITION
Vehicles ordered to fulfill consumer requests, or as part of standard stock
replenishment, will be received into dealership inventory on or earlier than the
35th day subsequent to the date registered as part of online order submission.
All days, except those recognized as holidays in the union contract, will be
included in the count.
28
3. Produce a data collection tool and develop a NOTE: The Project Team establishes a
process for using it to gather information. process based on Dealer Codes that ensures
inclusion of data collection forms with
transportation paperwork of vehicles being
shipped to the Los Angeles Region.
MEASURE PHASE Develop Data Collection Plan
32
Project Teams can use a worksheet similar to this when conducting Gage R&R Studies.
The Gage is acceptable only if all measurements agree. If measurements do not agree,
the measurement system must be improved or replaced.
MEASURE PHASE Perform Measurement System Analysis
Project Team Members use a “Walking the Process” methodology to construct Operational
Definitions and perform the Measurement System Analysis.
Prior to beginning the process, Project Team Members will educate measurement takers on
use of the Data Collection Tool and the format to use when logging time and date information.
Measurement takers will also receive instruction on how to annotate process exceptions.
Measurement takers receiving instruction include:
• Assembly Plant final inspection personnel (data collection initiators)
• Drivers who load vehicles on railcars
• Drivers who unload vehicles at Mix Center
• Haulaway Carrier drivers who transport vehicles from Mix Center to dealerships
• Dealership inventory management personnel who receive the vehicles
NOTE: The starting date for the order will be established using data from the vehicle
ordering system.
Project Team Members will follow the delivery process beginning with the Assembly
33 Plant inspector, continuing through transfer to railhead, the Mix Center and finishing with
vehicle receipt at the dealership. “Walking the Process” will be repeated until Project Team
Members are confident that data compiled during the collection period accurately reflects the
existing process.
MEASURE PHASEPerform Measurement System Analysis
34
MEASURE PHASE
M EASURE P HASE | Conduct Data Collection
Conduct
Data
Collection
35
MEASURE PHASE Conduct Data Collection
36
Based on the strategy established in the Data Collection Plan, the Project Team compiles the
following data:
1. Vehicle/Model
2. Number of vehicles delivered to the target dealerships without damage
3. Number of vehicles with observable exterior damage delivered to the target dealerships
4. Total number of vehicles delivered
5. Type of damage
6. Who first detected the damage (to help determine where in the process damage occurred)
7. Location of damage on the vehicle
MEASURE PHASE Conduct Data Collection
37
To facilitate graphical analysis, the Project Team subgroups vehicle data by Brand category:
1. Vehicle data subgrouped by brand category
NOTE: The Project Team will conduct Graphical Analysis on the data collected. Results of
this analysis will help prioritize improvement opportunities and provide the Project Team
further direction.
MEASURE PHASE Conduct Data Collection
38
Based on the established Data Collection Plan, the Project Team investigating Delivery Timing
compiles the following data:
1. Data compiled from the Data Collection Tool provides:
– Vehicle information
– Time and date information for each checkpoint
MEASURE PHASE Conduct Data Collection
39
To make the data useful for conducting Graphical Analysis, the Project Team will need to
convert date and time information to numerical values. Minitab offers an easy utility to make
the conversion.
1. After pasting data into Minitab, click on “Manip” from the Minitab navigation toolbar
2. Click on “Change Data Type” from the pull-down menu
3. Click on “Date/Time to Numeric”
MEASURE PHASE Conduct Data Collection
40
41
42
1. To determine where delays occur in the process, the Project Team needs to consider the
interval of time between checkpoints. These intervals of time will be referred to as legs and
are calculated by subtracting the preceding checkpoint value from the one that follows.
For example: C9 – C8. The new value will be stored in a column labeled LEG 1.
MEASURE PHASE Conduct Data Collection
43
44
45
1. The resulting value represents the number of days between when a vehicle order is submitted
into the order system and when the vehicle passes inspection and leaves the assembly plant.
MEASURE PHASE Conduct Data Collection
46
Pareto
48
Pareto
49
Pareto
50
Pareto
51
Pareto Chart Analysis - continued
1. A Pareto Chart of Ford Car brand vehicles shows that the largest amount of damage is
occurring on Taurus vehicles.
2. The Project Team decides to check the Mercury brand data to see if a similar amount of
damage is occurring on the Sable. The data verifies the Team’s intuition.
NEXT STEPS
Based on the outcome of the Graphical Analysis, the Vehicle Condition at Delivery Project Team
will focus on damage occurring to Taurus and Sable vehicles delivered to five large-volume
Southwest Region dealerships.
In later steps, the team will try to determine:
• If one type of damage is occurring more frequently than another
• If damage is occurring during a specific step of the vehicle delivery process
MEASURE PHASE Perform Graphical Analysis
Histogram
52
Histogram Analysis
The Project Team begins by calculating the total of each “LEG” of the delivery process using the
Minitab “Calculator” function (reference page 43 for details on accessing the calculator function).
1. Input the column label where the “total” values will be stored
2. Input column labels reflecting the values to be added (e.g., LEGs of the delivery process)
Note: Click on the “plus” key between each column label. Also, a sum of these columns can
be derived using the “Sum” function (see Minitab materials for further details)
3. Click “OK”
MEASURE PHASE Perform Graphical Analysis
Histogram
53
Histogram
54
Histogram
55
Boxplot
56
Boxplot Analysis
To perform a Boxplot analysis using Minitab:
1. Click “Graph” from the Minitab navigation toolbar
2. Click “Boxplot” from the pull-down menu
MEASURE PHASE Perform Graphical Analysis
Boxplot
57
Boxplot
58
Boxplot
59
Boxplot
60
Run Chart
61
Run Chart
62
Run Chart
63
Run Chart
64
Run Chart
65
Run Chart
66
67
68
MEASURE PHASE
M EASURE P HASE | Conduct Baseline Analysis
Conduct
Baseline
Analysis
The purpose of the Baseline Capability Analysis In handling variable data, Baseline Capability
is to characterize the current process in relation analysis involves the calculation:
to customer requirements (CTQs). The goal of • Cp — an index of capability potential
Baseline Capability is to determine: that indicates the amount of variation in
• If the current process, given its natural the process
variation, is capable of meeting • Cpk — an index that compares the natural
customer requirements tolerance of a process against its specification
• The percentage of defect within the limits to reveal if there is a problem with
current process specification centering.
As with other tools, Baseline Capability analysis Baseline capability of variable data also
is handled differently depending on whether the includes calculation of Pp and Ppk indices that
data is attribute or variable. reveal problems with variation and centering
over the long term.
70
From the Graphical Analysis, the Project Team concludes that the biggest opportunity for
improvement exists in the Taurus/Sable vehicle line. Therefore, the Project Team needs to
conduct a Baseline Capability Analysis on data relating to those models.
1. To begin, the Project Team isolates Taurus/Sable data in a new Minitab Worksheet.
2. The Project Team adds a column to indicate the number of defect opportunities available per
unit (each Taurus/Sable delivered). Because the Project Team is addressing whether the
vehicles arrived in good condition versus damaged (Attribute data) the number of opportunities
equals 1 (e.g., the vehicle was either damaged or good).
MEASURE PHASE Conduct Baseline Analysis
71
72
1. Input the label for the column containing data on the number of vehicles damaged
(in this example, C4).
2. Input the label for the column containing data on the total number of vehicles delivered
(in this example, C5).
3. Input the label for the column containing data number of opportunities for defect per unit
(in this example, C6).
4. Click “OK”
MEASURE PHASE Conduct Baseline Analysis
73
1. The results show individual Sigma values by model as well as a overall Sigma value of
2.36 for both Taurus and Sable models. In addition to the numeric calculations, this Minitab
function also presents the data in three graphical views as seen on the following page.
MEASURE PHASE Conduct Baseline Analysis
74
1. Graphical views
MEASURE PHASE Conduct Baseline Analysis
75
To conduct a Baseline Capability Analysis on Delivery Timing (Variable Data), the Project Team
takes a look at data expressing the total number of delivery days (Total B).
Note: Totals (Total B) taken from data collected after the “Quick Win” process improvement was
made to Explorer Sport Trac models featuring the cargo cage option.
1. Click “Six Sigma” from the Minitab navigation toolbar
2. Click “Progress Report” from the pull-down menu
MEASURE PHASE Conduct Baseline Analysis
76
1. Input the column label representing the total data (C19 = TOTAL B)
2. Input subgroup size (in this example, the subgroup size is 1)
3. Input “Spec” limits (in this example, there is only an upper Spec limit of 35 days)
4. Click “OK”
MEASURE PHASE Conduct Baseline Analysis
77
Results provide two graphic summaries of the Capability Analysis. The first is an Executive
Summary showing:
• Process Centering relative to Spec limits
• Sigma Values
• Defects per million opportunities
MEASURE PHASE Conduct Baseline Analysis
78
1- 2- ANOVA
F-Test ANOVA Sample Sample for
T-Test T-Test Means
ANALYZE PHASE A N A LY Z E P H A S E
80
ANALYZE PHASEChi-Square 2 Proportion
Confirm Data Type
1 Proportion
Attribute
Data
Confirm
Data
Type
Variable
Data
Analysis of
Variation
Analysis of
Means
Correlation
Regression
1- 2- ANOVA
F-Test ANOVA Sample Sample For
T-Test T-Test Means
Confirm Data Type The pages that follow offer details on these
analysis tools as well as an overview of
To begin the Analyze Phase, the Project Team Minitab operations required to apply these
needs to confirm the type of data that has been tools to project data. Tools include:
collected, either Attribute or Variable. Based on
the data type, the Project Team will use one of
two sets of analysis tools.
81
ANALYZE PHASE Attribute Data Tools
1 Proportion
• Tests for independence between data
samples (shows no relationship)
• Null Hypothesis (Ho) = there is statistical
independence between the two data
samples (no relationship exists).
• P-Value greater than .05 = accept the null
hypothesis
• Tests a data sample against a test proportion, • P-Value less than .05 = reject null hypothesis
target or defined standard
• Determines if there is statistically
significant difference 82
• Null Hypothesis (Ho) = there is no significant
difference between the sample data and
the target
• P-Value greater than .05 = accept
null hypothesis
• P-Value less than .05 = reject null hypothesis
2 Proportions
Analysis of Variance — tests for level • Bartlett’s requires normal data, Levene’s does
of variation within a data sample not require normal data
• Null Hypothesis (Ho) = all data samples have
• Variable — data reflecting a range of conditions such as: the same degree of variance
– Processing Time
– Items Processed • P-Value greater than .05 = accept the null
– Number of Defects hypothesis
(requires a minimum sample size of 30 per subgroup)
• P-Value less than .05 = reject the null
hypothesis
F-Test Analysis of Means — tests and
compares the means of data samples
1-Sample T-Test
1- 2- ANOVA
F-Test ANOVA Sample Sample For
T-Test T-Test Means
1 Proportion
• Tests a data sample against a test proportion, target or
defined standard
• Determines if there is statistically significant difference
• Null Hypothesis (Ho) = there is no significant difference between the
sample data and the target
• P-Value greater than .05 = accept null hypothesis
• P-Value less than .05 = reject null hypothesis
85
1 Proportion
86
1. Select “Summarized data” by clicking the NOTE: The Project Team is testing their data
adjacent checkbox sample against a national average (3%) for
2. Input the total number of Taurus/Sable observable defects; therefore, defects are
vehicles delivered (155) in the field adjacent input as successes. Conversely, the equation
to “Number of trials” (In this example, total could be turned around to test against an
number of vehicles is the “Number of trials”) average percentage of “Good” units (97%).
(In this scenario, “Number of successes” would
3. Input the number of defects observed in the be 125.)
total number of vehicles delivered (30) in
the field adjacent to “Number of successes” 4. Click on “Options”
ANALYZE PHASE Attribute Data Tools
1 Proportion
87
1. In the field adjacent to “Confidence level,” wishes to test their data sample against a
input the confidence level for the data target. The null hypothesis is that the two
sample (in this example, 95%) are equal; the alternative hypothesis is that
the two are “not equal”)
2. In the field adjacent to “Test Proportion”
input the target against which the data 4. Click on the check box “Use test and
sample will be tested (In this example, the interval based on normal data” (as
average percentage for observable defect is appropriate for sample size)
3%; therefore, input .03)
5. Click “OK”
3. Select “not equal” in the “Alternative” pick
box (In this example, the Project Team
ANALYZE PHASE Attribute Data Tools
1 Proportion
88
1. Click “OK”
ANALYZE PHASE Attribute Data Tools
1 Proportion
89
1. Results of the 1 Proportion test provide a the Taurus/Sable data sample is statistically
P-Value of 0.000 different from the national average percentage
of observable defects.
A P-Value less than .05 means the Project
Team must reject the Null Hypothesis (If the Because the P-Value is less than .05, the
P is low, the null must go!). The Null Project Team is directed to the Alternative
Hypothesis is that Taurus/Sable data sample is Hypothesis. The data sample shows that
statistically the same as the national average Taurus/Sable models are exhibiting more
percentage of observable in-transit defects. in-transit damages than other models.
By contrast, the Alternative Hypothesis is that
ANALYZE PHASE Chi-Square 2 Proportion
Attribute Data Tools
1 Proportion
Attribute
Data
Confirm
Data
Type
Variable
Data
Analysis of
Variation
Analysis of
Means
Correlation
Regression
1- 2- ANOVA
F-Test ANOVA Sample Sample For
T-Test T-Test Means
2 Proportion
• Tests two data samples against each other
• Determines if there is statistically significant difference
between the two data samples
• Null Hypothesis (Ho) = there is no significant difference
between the two data samples (A = B)
• P-Value greater than .05 = accept null hypothesis
• P-Value less than .05 = reject null hypothesis
90
The Project Team wants to see if there is a To perform the 2 Proportion test:
statistical difference in the amount of in-transit 1. Click on “Stat” from the Minitab navigation
damage observed in Taurus models compared toolbar
to Sable models. Statistical difference will be
tested using a 2 Proportion test. 2. Click “Basic Statistics” from the pull-down
menu
3. Click “2 Proportion” from the secondary
pull-down menu
ANALYZE PHASE Attribute Data Tools
2 Proportion
91
2 Proportion
92
1. In the “Confidence level” field, enter the 3. In the “Alternative” pick box, select “not
appropriate confidence level (In this equal” (In this example, the Null Hypothesis
example, the confidence level is 95%) states that the two data samples are equal.
The “Alternate Hypothesis” states that the
2. In the “Test difference” field, enter the two data samples are not equal)
appropriate parameter (In this example, the
Project Team is testing to see if the degree 4. Click “OK”
of observable defective units is the same
between Taurus and Sable models;
therefore, the difference is set to zero)
ANALYZE PHASE Attribute Data Tools
2 Proportion
93
1. Click “OK”
ANALYZE PHASE Attribute Data Tools
2 Proportion
94
1. Results indicate a P-Value greater than .05; indicated that both Taurus and Sable
therefore the Project Team fails to reject the models are experiencing the same degree
Null Hypothesis that the two data samples of in-transit defects)
are the same (In this example, the data
ANALYZE PHASE Chi-Square 2 Proportion
Attribute Data Tools
1 Proportion
Attribute
Data
Confirm
Data
Type
Variable
Data
Analysis of
Variation
Analysis of
Means
Correlation
Regression
1- 2- ANOVA
F-Test ANOVA Sample Sample For
T-Test T-Test Means
Chi-Square
• Tests for independence between data samples (shows no relationship)
• Null Hypothesis (Ho) = there is statistical independence between
the two data samples (no relationship exists)
• P-Value greater than .05 = accept null hypothesis
• P-Value less than .05 = reject null hypothesis
95
The Project Team notices that damage related To prepare for this test the Project Team
to “Paint (scratches)” occurs more frequently structures the data sample such that
than other types of damage. The Project Taurus and Sable counts are combined and
Team next wants to see if there is a all other vehicles are totaled and placed into
relationship between paint damage and another row.
location on the vehicle. To take a look at these 1. Row 1 contains data on all vehicles other
possible relationships, the Project Team than Taurus and Sable
conducts a Chi-Square test.
2. Row 2 contains Taurus and Sable data
ANALYZE PHASE Attribute Data Tools
Chi-Square
96
1. Click “Stat” on the Minitab navigation toolbar 3. Click “Chi-Square Test” on the secondary
2. Click “Tables” on the pull-down menu pull-down menu
ANALYZE PHASE Attribute Data Tools
Chi-Square
97
1. In the field adjacent to “Columns containing location of defect including: “Doors,” “Hood,”
the table,” input labels for those columns to “Truck/Tailgate,” “Front Qtr,” “Rear Qtr.”)
be analyzed (In this example, the Project
Team analyzes data for “Paint (Scratch)” 2. Click “OK”
and “Dent” as well as data relating to
ANALYZE PHASE Attribute Data Tools
Chi-Square
98
1- 2- ANOVA
F-Test ANOVA Sample Sample For
T-Test T-Test Means
F-Test
99
After making the “Easy Win” improvement relating to Sport Trac models, the Project Team
collects a second data sample:
The second data sample will be used throughout the Analyze Phase.
ANALYZE PHASE Variable Data Tools | Analysis of Variation
F-Test
• Tests and compares two data sets for variance
• Assumes a random sample of normal data (see “Testing for
Normal Data” in Appendix)
• Null Hypothesis (Ho) = both data samples have the same
degree of variance
• P-Value greater than .05 = accept the null hypothesis
• P-Value less than .05 = reject the null hypothesis
100
F-Test
101
F-Test
102
F-Test
103
1. Data from column C15 has been copied into the first 40 rows of column C20 and automatically
assigned the subscript 1
2. Data from column C19 has been copied and stacked underneath in rows 41 through 80 and
automatically assigned the subscript 2
NOTE: Minitab software requires stacked columns data to perform many of the tests used in
the Analyze Phase. From this point forward in the Guidebook, instructions will refer to
Stack Columns as the first step in performing the test. Please refer back to these pages for
referencing the steps involved in stacking data.
ANALYZE PHASE
Variable Data Tools | Analysis of Variation
F-Test
104
After the Stack Columns function has been performed to arrange data for analysis,
the Project Team will:
1. Click “Stat” from the Minitab navigation toolbar
2. Click “ANOVA” from the pull-down menu
3. Click “Homogeneity of Variance” from the secondary pull-down menu
ANALYZE PHASE
Variable Data Tools | Analysis of Variation
F-Test
105
1. In the field next to “Response,” input the label for the column containing the stacked data
(C20)
2. In the field next to “Factors,” input the label for the column containing the subscripts for the
stacked data (C21)
3. In the field next to “Confidence level,” enter the appropriate number (In this example, the
Project Team has a Confidence level of 95% based on the generally accepted 5% error
rate in 6-Sigma)
4. Click “OK”
ANALYZE PHASE
Variable Data Tools | Analysis of Variation
F-Test
106
1. The Null Hypothesis for the F-Test is that the two data samples will have the same degree of
variance. Since results of the F-Test have produced a P-Value greater than .05, the Project
Team must accept that the two data samples have the same degree of variance.
In this example, data indicates that LEG2b (the amount of time Explorer vehicles wait on the
plant lot prior to loading on a railcar) has the same degree of variance as TOTAL B (the
amount of time it takes from order submission until the Explorer arrives at the dealership).
Next, the Project Team will look at variation in the other LEGs of the process to determine
impact on the overall total.
ANALYZE PHASE Variable Data Tools | Analysis of Variation
Chi-Square 2 Proportion 1 Proportion
Attribute
Data
Confirm
Data
Type
Variable
Data
Analysis of
Variation
Analysis of
Means
Correlation
Regression
1- 2- ANOVA
F-Test ANOVA Sample Sample For
T-Test T-Test Means
107
Results of the F-Test indicate a similarity between the degree of variation seen in LEG2 and the
degree of variation seen in the overall delivery timing. However, the Project Team wants to look
at all the LEGs of the process to see the degree of variation in each step. To look at all LEGs
compared to one another, the Project Team will perform an ANOVA test.
1. To perform the ANOVA test, the Project Team must first stack data using the Minitab Stack
Columns function. (In this example, stacked data has been placed in columns C22 & C23.)
ANALYZE PHASE
Variable Data Tools | Analysis of Variation
ANOVA
108
ANOVA
109
1. In the field next to “Response,” input the label for the column containing stacked data (In this
example, data for LEG1b, LEG2b, LEG3b, LEG4b, LEG5b is contained in C22)
2. In the field next to “Factors,” input the label for the column containing subscripts for the
stacked data (C23)
3. In the field next to “Confidence level,” input the appropriate value (In this example, the Project
Team has determined a 95% Confidence level based on the generally accepted % error rate
in 6-Sigma)
4. Click “OK”
ANALYZE PHASE
Variable Data Tools | Analysis of Variation
ANOVA
110
1. Results of the ANOVA test on all LEGs of Delivery Timing have produced a P-Value less
than .05. (In this example the data is normal; therefore, the Project Team refers to results of
the Bartlett’s Test.) The Null Hypothesis for an ANOVA test states that all data samples will
have the same degree of variance. Since the P-Value is less than .05, the Project Team
must reject the Null Hypothesis and accepts that all data samples do not have the same
degree of variance.
2. The graphic also indicates that all data samples do not have the same degree of variance.
Furthermore, the graphic shows that LEG2 has the greatest degree of variation when
compared with the other LEGs.
ANALYZE PHASE Variable Data Tools | Analysis of Means
Chi-Square 2 Proportion 1 Proportion
Attribute
Data
Confirm
Data
Type
Variable
Data
Analysis of
Variation
Analysis of
Means
Correlation
Regression
1- 2- ANOVA
F-Test ANOVA Sample Sample For
T-Test T-Test Means
1-Sample T-Test
• Tests the data sample against a test proportion, target or
defined standard
• Determines if there is statistically significant difference
• Null Hypothesis (Ho) = there is no significant difference
between the sample data and the target
• P-Value greater than .05 = accept the null hypothesis
• P-Value less than .05 = reject the null hypothesis
111
To determine how overall Delivery Timing compares to the target of 35 days, the Project Team
will conduct a 1-Sample T-Test.
To perform a 1-Sample T-Test, the Project Team will:
1. Click on “Stat” on the Minitab navigation toolbar
2. Click on “Basic Statistics” from the pull-down menu
3. Click on “1-Sample T-Test” from the secondary pull-down menu
ANALYZE PHASEVariable Data Tools | Analysis of Means
1-Sample T-Test
112
1. In the “Variables” field, input the label for the column of data to be tested
2. Click on the selection box adjacent to “Test mean”
3. In the field next to “Test mean,” input the appropriate value. (In this example, the Project Team
tests current Delivery Timing (TOTAL B) against the goal of 35 days.)
4. In the field next to “Alternative,” select the default of “not equal”
5. Click on “Graphs”
ANALYZE PHASEVariable Data Tools | Analysis of Means
1-Sample T-Test
113
1-Sample T-Test
114
1. Click “OK”
ANALYZE PHASE
Variable Data Tools | Analysis of Means
1-Sample T-Test
115
1. Results of the 1-Sample T-Test shows the mean of TOTAL B (40.562 days) as well as the
confidence interval
2. The target mean of 35 days is shown as point “Ho”
ANALYZE PHASE
Variable Data Tools | Analysis of Means
1-Sample T-Test
The 1-Sample T-Test also provides a Dotplot and Histogram view of mean of the data sample 116
compared to the target. Based on the results of the 1-Sample T-Test, the Project Team looks
for an improvement opportunity or combination of opportunities that will lessen Delivery Timing
by approximately 5 days.
ANALYZE PHASE Variable Data Tools | Analysis of Means
Chi-Square 2 Proportion 1 Proportion
Attribute
Data
Confirm
Data
Type
Variable
Data
Analysis of
Variation
Analysis of
Means
Correlation
Regression
1- 2- ANOVA
F-Test ANOVA Sample Sample For
T-Test T-Test Means
2-Sample T-Test
• Tests and compares the means of two data samples
• Assumes a random sample of normal data (see “Testing
for Normal Data” in Appendix)
• Null Hypothesis (Ho) = both data samples have the
same means
• P-Value greater than .05 = accept the null hypothesis
• P-Value less than .05 = reject the null hypothesis
117
Because assembly plant production capacity is not within the Project Team’s area of influence,
the team will focus its efforts on downstream activities. The Project Team will be looking for
improvement opportunities that can shorten Delivery Timing by five days.
The Project Team believes LEG2 provides improvement opportunities and suspects there might
be additional opportunities in LEG3. To see a comparison of the means of these LEGs, the
Project Team will perform a 2-Sample T-Test.
To perform the 2-Sample T-Test, the Project Team will:
1. Click on “Stat” from the Minitab navigation toolbar
2. Click on “Basic Statistics” from the pull-down menu
3. Click on “2-Sample T-Test” from the secondary pull-down menu
ANALYZE PHASEVariable Data Tools | Analysis of Means
2-Sample T-Test
118
2-Sample T-Test
119
2-Sample T-Test
120
1. Click “OK”
ANALYZE PHASEVariable Data Tools | Analysis of Means
2-Sample T-Test
121
1- 2- ANOVA
F-Test ANOVA Sample Sample For
T-Test T-Test Means
122
To better understand the overall Delivery Process, Project Team Members will conduct an
ANOVA for Means. This analysis will show whether there is significant statistical difference
between the means of each LEG of the process. The Null Hypothesis is that there will be no
statistically significant difference between one LEG and any other LEG.
To perform an ANOVA for Means the Project Team will:
1. Click on “Stat” from the Minitab navigation toolbar
2. Click on “ANOVA” from the pull-down menu
3. Click on “One-way” from the secondary pull-down menu
ANALYZE PHASEVariable Data Tools | Analysis of Means
Chi-Square 2 Proportion 1 Proportion
Attribute
Data
Confirm
Data
Type
Variable
Data
Analysis of
Variation
Analysis of
Means
Correlation
Regression
1- 2- ANOVA
F-Test ANOVA Sample Sample For
T-Test T-Test Means
123
1. In the field next to “Response,” input the label for the column for the data to be tested.
(In this example, the Project Team uses Stack Columns data compiled earlier for
ANOVA-Homogeneity of Variance Analysis (C22).)
2. In the field next to “Factor,” input the label for the column containing subscript information
(in this example, C23).
3. Click “OK”
ANALYZE PHASEVariable Data Tools | Analysis of Means
124
1. Results of the ANOVA for Means analysis indicate a P-Value below .05 (in this example,
0.000). Based on this P-Value, the Project Team cannot accept the Null Hypothesis and instead
looks to the Alternative Hypothesis, which states that there is a statistically significant difference
between at least one LEG and one other LEG.
ANALYZE PHASEVariable Data Tools | Analysis of Means
125
1. Scrolling down in the Session window reveals a summary table that includes the Mean and
Standard Deviation for each LEG of the Delivery Timing process.
2. A graphical representation (Boxplot) of the ANOVA for Means shows a mean of each LEG
compared to means for the other LEGs.
Based on the results of the ANOVA for Means test, Project Team Members direct their attention
to LEG1, LEG2 and LEG3, as they represent the largest means.
The Project Team notes that LEG1 (the Factory) is beyond the scope of their project and that
LEG3 (Rail Transit) does not exhibit the level of variation seen in LEG2. Therefore, the Project
Team will target LEG2 for improvement opportunities.
ANALYZE PHASE Chi-Square 2 Proportion
Correlation Regression
1 Proportion
Attribute
Data
Confirm
Data
Type
Variable
Data
Analysis of
Variation
Analysis of
Means
Correlation
Regression
1- 2- ANOVA
F-Test ANOVA Sample Sample For
T-Test T-Test Means
Correlation Regression
• Tests and compares data samples for correlation
• Requires normal data
• R-Squared > .80 correlation may be significant
• R-Squared > .50 and < .80 weak to moderate correlation
(use sound judgment)
• R-Squared < .50 correlation may not be significant
126
To better understand the relationship LEG2 variation has on the degree of variation seen in the
overall process, the Project Team will perform a Correlation Regression test.
To perform a Correlation Regression test, the Project Team will:
1. Click on “Stat” from the Minitab navigation toolbar
2. Click on “Regression” from the pull-down menu
3. Click on “Fitted Line Plot”
ANALYZE PHASE Correlation Regression
Correlation Regression
127
1. In the field next to “Response (Y),” input the label for the column that contains data which will
be impacted by a correlation. (In this example, the Project Team wants to see if LEG2
correlates to the TOTAL; therefore, the “Response” field should contain the label for TOTAL
data (C19).)
2. In the field next to “Predictor (X),” input the label for the column that contains data which will
impact “Response” data. (In this example, Predictor data is LEG2 (C15).)
3. Click “Linear” from the selection boxes beneath Type of Regression Model.
4. Click “Options”
ANALYZE PHASE Correlation Regression
Correlation Regression
128
Correlation Regression
129
1. Click “OK”
ANALYZE PHASE Correlation Regression
Correlation Regression
130
1. Results of the Correlation Regression test produces an R-Square value of .853, which
indicates a correlation that may be statistically significant
2. The Regression Line is upward sloping, which indicates a positive correlation (In this example,
as the duration of LEG2 increases there is a corresponding increase in the overall duration of
the Delivery Timing process.)
3. Confidence Interval lines (red) show the confidence interval on the coefficients
4. Prediction Interval lines (blue) show the area (between the prediction bands) where for any
given “X” (Response) there is a 95% chance that the corresponding “Y” (Predictor) will lie
Based on results from the Correlation Regression test, the Project Team will look more closely
at improvement opportunities that exist in LEG2.
IMPROVE PHASE Create
I M P ROV E P H A S E
Perform
Brainstorm “Should Be” Conduct Validate
Cost/Benefit Pilot
Alternatives Process FMEA Improvement
Analysis
Map
IMPROVE PHASE I M P ROV E P H A S E
Alternatives
132
IMPROVE PHASE Brainstorm Alternative
At the conclusion of the Analyze Phase, the Project Team is directed toward paint scratch/hood
damage observed on Taurus and Sable vehicles being shipped to high-volume dealerships in the
Southwest Region.
Further investigation by the Project Team indicates that damage is first observed upon the
vehicle arrival at the dealership. Therefore, Project Team Members decide to “Walk-the-Process”
between the outbound Holding Yard inspection and inspection at the dealership to determine
possible causes for damage.
Further Findings
Project Team Members monitor the loading of vehicles on haulaway trucks and then inspect
vehicles upon arrival at the destination dealership. The Project Team notes damage is
observable only on those units located at the most rearward position on the upper level of the
car carrier with the front of the vehicle facing downward.
The Project Team repeats the monitoring process placing special attention to the most rearward
vehicle on the upper level of the car carrier. Team Members note that damage was not observed
133 when the vehicle was loaded with the rear of the vehicle facing downward.
At the conclusion of the monitoring process, the Project Team determines that the technique
used to chain the vehicles to the carrier compresses the vehicle’s suspension such that front
quarter panels of the vehicle rub against the side of the hood. The rubbing action that occurred
during transit removed paint from the hood panel and is noted as Paint Scratch damage during
dealership inspection.
Brainstormed Solutions
1) Training for carrier drivers regarding loading and chaining procedures
2) Panel Spacers for use during transit
IMPROVE PHASE Brainstorm Alternative
At the conclusion of the Analyze Phase, Project Team Members are directed toward LEG2
(waiting period between assembly and loading onto railcars for transit) and LEG3 (railcar
in-transit time). Therefore, the Project Team conducts a deeper investigation into LEG2 and
LEG3 process steps.
Further Findings
The Project Team Members find that notification to the rail transit provider of vehicles ready for
transport occurs once per day as a batch data file. Once the information is received, the rail
transit provider begins a process of sorting vehicles by destination, then sequencing loading at
the assembly plant.
Brainstormed Solutions
• More frequent data transfers
• Pre-notification of vehicle order and build status
• Rate incentives for target date delivery
134
IMPROVE PHASE Create “Should Be” Process Map
Create
“Should Be”
Process
Map
Create “Should Be” Process Map solutions must be determined through the
experience and expertise of the Project Team
After alternatives have been brainstormed, the Members.
Project Team determines which alternative
offers the “best” solution or “best set” of Once a “best” solution has been identified, the
solutions for implementation. Project Team will Create [a] “Should Be”
Process Map." This revised process map will
Design of Experiment (DOE) techniques can be include all steps necessary to implement the
helpful in determining a “best” solution if the improvement. The “Should Be” Process Map
project allows for easy implementation and will serve as the reference document for the
monitoring of incremental changes. However, new process.
on many non-manufacturing projects, “best”
135
The Project Team determines that “Panel Spacers” offers the best solution for reducing the
amount of observable Paint Scratch damage on Taurus and Sable vehicles.
The Panel Spacer solution includes the placement of four small pieces of high density Styrofoam
near the corners of the hood panel. After placement, these spacers maintain a proper gap
between the hood panel and front quarter panels during transit.
The “Should Be” Process Map includes installation of Panel Spacers at the Assembly Plant
as part of the body assembly process.
IMPROVE PHASE Create “Should Be” Process Map
The Project Team determines that the “best” solution alternative is to submit “vehicles ready for 136
transport” information to the rail transit provider twice daily rather than once per day.
The “Should Be” Process Map will include the transmission of data at two intervals each day:
• Late morning between 11:00 A.M. and Noon
• Close of business 5:00 P.M.
IMPROVE PHASE Conduct FMEA
Conduct
FMEA
Conduct FMEA
FMEA is an effective tool that helps ensure
potential process/product failure modes; the
causes and customer impacts of those failures
have been considered and addressed.
137
138
IMPROVE PHASE Perform Cost/Benefit Analysis
Perform
Cost/
Benefit
Analysis
139
The Project Team determines the following costs and benefits from implementing the “Panel
Spacer” solution:
Based on the Cost/Benefit Analysis, the Project Team concludes that they should conduct a pilot
of the Panel Spacer solution in preparation for full implementation.
IMPROVE PHASE Perform Cost/Benefit Analysis
140
The Project Team determines the following costs and benefits from implementing the “Twice
Daily Data Submission” solution:
Based on the Cost/Benefit Analysis, the Project Team concludes that they should conduct a pilot
of the “Twice Daily Data Submission” solution in preparation for full implementation.
IMPROVE PHASE Pilot
Pilot
The Project Team secures enough Panel Spacers to cover one week’s production of Taurus and
Sable vehicles. Using the same data collection methodology specified in the Measurement
System Analysis, the Project Team collects a new data sample to verify the improvement.
142
The Pilot of the Panel Spacers solution is producing defect levels at slightly less than 3.5 Sigma
(3.334). While this performance is not World Class, it does bring the amount of observable
Paint Scratch damage on Taurus and Sable vehicles near the target of 3%. Therefore, the Panel
Spacer solution will most likely produce the projected cost savings and increases in
customer satisfaction.
IMPROVE PHASE Validate Improvement
143
Results of the “Twice Daily Data Submission” have produced a Sigma shift of 1.76 Sigma (.85 to
2.61). While the projection for long-term performance is slightly above the 35-day target, the
improvement is significant enough to warrant full-scale implementation.
CONTROL PHASE CONTROL PHASE
Long-term Control
Mistake Reaction Update
MSA Chart
Proofing Plan S.O.P.
Plan (SPC)
144
CONTROL PHASE
Outcomes
Upon completion of the Control Phase, the
Process Owner will understand performance
expectations, how to measure and monitor Xs
to ensure performance of the Y, and what
corrective actions should be executed if
measurements drop below desired levels.
Likewise, after completion of the Control
Phase, the Project Team members disband
while the Black Belt begins the next Consumer
Driven 6-Sigma Project with a new team.
145
CONTROL PHASE Mistake Proofing
Mistake
Proofing
To mistake proof the process of installing panel spacers, Project Team members will need to
investigate how mistakes occur and what means can be used to prevent missed installation.
Possible areas for review included:
146 • How installation of Panel Spacers is embedded in the assembly process
• Determine whether spacers should be installed on the hood or quarter panels
Long-term MSA Plan Xs, the Process Owner can ensure that the Y
output continues to meet the CTQs.
After mistake proofing the process, the Project
Team will develop a Long-term Measurement A critical consideration in the development of a
System Analysis (MSA) Plan. Similar to the Long-term MSA Plan is how changes to the
original Measurement System Analysis, the process will impact what is measured and how
Long-term MSA Plan looks at all aspects of that measurement takes place. Addressing this
data collection relating to ongoing situation avoids measuring process steps that
measurement of the Xs and high-level are no longer relevant to the CTQ.
monitoring of the Y output. By controlling the
The Project Team documents steps involved in validating the measurement system should the
vehicle design change. For example, subsequent product design changes (such as improved
vehicle rigidity) may eliminate observable damage relating to rubbing of the hood and quarter
panels during transit. Should such a design change take place, the measurement system needs
to reflect the change and the use of panel spacer may be unnecessary.
147
Project Team members develop a Long-term MSA Plan that covers measurement system
changes required should steps in the Delivery Process be adjusted or eliminated. For example,
if the Mix Center was eliminated from the process, LEG 4 would no longer be valid and
specifications for other LEGs may need subsequent adjustments.
CONTROL PHASE Control Chart (SPC)
Control
Chart
149
CONTROL PHASE Control Chart (SPC) | Attribute Data
P-Chart
150
To monitor the number of damaged units (against a target of 3%), the Project Team and Process
Owners will use a P-Chart.
1. To implement the SPC P-Chart, data is collected and input into a Minitab worksheet over
several weeks including:
• Number of Taurus/Sable units delivered
• Number of damaged units (defective)
• Number of defects observed
CONTROL PHASE Control Chart (SPC) | Attribute Data
P-Chart
151
P-Chart
152
1. In the field next to “Variable,” input the label for the column containing data on damaged units
(in this example, C28).
2. Click on the selection box adjacent to “Subgroups in”
3. In the field next to “Subgroups in,” input the label for the column containing sample size data
(in this example, C27)
4. Click on “Tests”
CONTROL PHASE Control Chart (SPC) | Attribute Data
P-Chart
153
P-Chart
154
1. Click “OK”
CONTROL PHASE Control Chart (SPC) | Attribute Data
P-Chart
155
1. The P-Chart shows the data line to be within the upper and lower control limits; therefore, the
process can be considered in control.
2. The data line trends downward indicating improvement over time as Process Owners and
participants become familiar with new procedures and requirements.
3. The average proportion is below the .03 target.
CONTROL PHASE Control Chart (SPC) | Attribute Data
U-Chart
156
As an additional process control, the Project Team will monitor the number of defects observable
on each damaged unit using a U-Chart.
To generate a U-Chart, the Project Team will:
1. Click on “Stat” from the Minitab navigation toolbar
2. Click on “Control Charts” from the pull-down menu
3. Click on “U” from the secondary pull-down menu
CONTROL PHASE Control Chart (SPC) | Attribute Data
U-Chart
157
1. In the field next to “Variable,” enter the label for the column containing defect data (in this
example, C29)
2. Click the selection box adjacent to “Subgroup in”
3. In the field next to “Subgroup in,” input the label for the column containing data relating to the
number of damaged units (in this example, C28)
4. Click “Tests”
CONTROL PHASE Control Chart (SPC) | Attribute Data
U-Chart
158
U-Chart
159
1. Click “OK”
CONTROL PHASE Control Chart (SPC) | Attribute Data
U-Chart
160
1. Results of the U-Chart indicate the process is within default control limits.
2. U-Chart results also indicate the average number of defects per damaged unit is less
than two.
CONTROL PHASE Control Chart (SPC) | Variable Data
I-MR
I-MR Chart
161
To monitor process variation, the Project Team will use an I-MR (Individual – Moving Range)
control chart.
To generate an I-MR chart, the Project Team will:
1. Click on “Stat” from the Minitab navigation toolbar
2. Click on “Control Charts” from the pull-down menu
3. Click on I-MR from the secondary pull-down menu
CONTROL PHASE Control Chart (SPC) | Variable Data
I-MR Chart
162
1. In the field next to “Variable,” input the label for the column containing data to be monitored
(in this example, the Project Team wants to monitor LEG2 - C4 for variation)
2. Click on “Test”
CONTROL PHASE Control Chart (SPC) | Variable Data
I-MR Chart
163
I-MR Chart
164
1. Click “OK”
CONTROL PHASE Control Chart (SPC) | Variable Data
I-MR Chart
165
1. Results of the I-MR Chart show individual values in the top chart. (In this example, the
process is producing fairly stable results within the control limits.)
2. Average number of days for LEG2 is 4.881
3. The bottom chart indicates the Moving Range. The Moving Range chart shows the degree of
variation in the data sample. (In this example, the variation is within control limits, but there
are some “spikes” that might be investigated for special causes.)
4. The average degree of variation is 0.6552
CONTROL PHASE Control Chart (SPC) | Variable Data
Xbar-R
Xbar-R Chart
166
To monitor the means for the entire delivery process, the Project Team uses an Xbar-R
control chart.
To generate the Xbar-R chart, the Project Team will:
1. Click on “Stat” from the Minitab navigation toolbar
2. Click on “Control Charts” from the pull-down menu
3. Click on Xbar-R from the secondary pull-down menu
CONTROL PHASE Control Chart (SPC) | Variable Data
Xbar-R Chart
167
Xbar-R Chart
168
Xbar-R Chart
169
1. Click “OK”
CONTROL PHASE Control Chart (SPC) | Variable Data
Xbar-R Chart
170
Reaction Plan
One of the final steps in the Control Phase is
the development of a detailed Reaction Plan.
This document will be used by the Process
Owners as a guide to adjusting the process
should one of the Control Charts begin showing
data that plots outside of the control limits.
The Reaction Plan should include:
• How to differentiate between a Special Cause
and a Process Trend
• When to react
• What actions should be taken
• Who is responsible for implementing
those actions
171
CONTROL PHASE Update S.O.P.
Update
S.O.P.
172
APPENDIX
APPENDIX
APPENDIX APPENDIX
173
APPENDIX
Design of Experiments constant during subsequent process
implementation. Ultimately, these variables
The Design of Experiment (DOE) is an need to be desensitized through a more robust
extremely valuable tool not only in identifying process design that will render them
key input variables causing variation but also in controllable.
setting the operating tolerances (precise levels) • Key Process Variables — those variables that
of the key input variables in order to optimize will be varied during experimentation.
the output. DOE works well in manufacturing
processes where equipment can be adjusted Experimentation Types
incrementally and output can be analyzed
over time. • Full Factorial — experimentation that includes
all possible combinations of factors and
In conducting DOE, Project Team Members will levels. If there are k factors and 2 levels, the
make changes to the inputs (factors) for the number of runs will be 2k. Likewise, if there
purpose of observing corresponding changes in are 3 levels, the number of tests would be 3k.
the outputs (responses). In complex processes with many levels, Full
Factorial experimentation can be time and
Key Concepts resource extensive.
175
176
1. Input the column label corresponding with the data to be tested (in this example, LEG1 = C14)
2. Click on “Graphics”
APPENDIX
Testing for Normal Data
177
1. Clicking the “Graphical Summary” checkbox gives the Project Team a complete set of
data graphs.
2. The default value of 95.0 is appropriate, as the data represents the appropriate number of
random samples.
3. Click “OK"
APPENDIX
Testing for Normal Data
178
1. The histogram shows the distribution of data for LEG1 following the normal curve.
2. Likewise the “P-Value” is above .05, thus indicating that the data has a normal distribution.
NOTE: The “P-Value” is an indicator as to whether the Null Hypothesis (H0) can be
accepted or rejected. In cases where the Null Hypothesis is rejected, the Alternative Hypothesis
(Ha) is accepted.
As a rule, If the P-Value is low, the Null has to go. In other words, if the P-Value is less than .05,
the Null Hypothesis must be rejected.
In the case of Normality testing, the Null Hypothesis (H0) holds that the data tested will have a
normal distribution. The Alternative Hypothesis (Ha) is that the data is not normally distributed.
APPENDIX
Testing for Normal Data
179
1. When LEG2 is tested for Normality, the Project Team notes the same bimodal data.
2. A low P-Value (0.001) confirms that the data is not normally distributed.
APPENDIX
Correcting Non-normal Data • Batch Processing Cycles — where items
missing the current processing cycle are
To correct non-normal data, the Project Team delayed until the next cycle
should investigate the existing process to
determine what conditions are causing data to • Shift Work — work in process at the
exhibit a non-normal distribution. conclusion of the first shift is not finished by
the second, but instead is delayed until the
Often in cases of bimodal distribution, there is a first shift returns
process step which, depending on outcome,
causes measurements to be either high or low. Once identified, conditions such as these can
Examples of this type of condition include: be corrected by either changing the process
step or by adjusting the measurement system
• Parts Availability — where products ship on to accommodate for the condition.
time, when parts are available, and much later
than normal when parts are not available
180
APPENDIX
Gage R&R for Variable Data
181
Project Teams working with Variable Data may want to conduct a Gage R&R to determine the
level of Repeatability and Reproducibility in their measurement system.
1. To conduct a Gage R&R, the Project Team will collect an appropriate data sample. (In this
example, the Project Team asks 3 dealership Technicians (labeled Operators) to measure 10
different vehicle paint scratches (labeled Parts). Measurements (labeled Response) are taken
in inches and the Technicians are asked to measure each scratch twice (in random order).
Results are logged in a Minitab Worksheet.)
APPENDIX
Gage R&R for Variable Data
182
183
1. In the field next to “Part Numbers,” input the column label for part data (in this example, C1)
2. In the field next to “Operators,” input the column label for operator data (in this example, C2)
3. In the field next to “Measurement data,” input the column label for response data (in this
example, C3)
4. Click on the selection box adjacent to “ANOVA”
5. Click “OK”
APPENDIX
Gage R&R for Variable Data
184
ANALYZE PHASE A N A LY Z E P H A S E
Attribute Confirm Variable Analysis of Analysis of Correlation
Chi-Square 2 Proportion 1 Proportion Data Variation Means Regression
Data Type Data
1- 2- ANOVA
F-Test ANOVA Sample Sample For
T-Test T-Test Means
IMPROVE PHASEBrainstorm
I M P ROV E P H A S E
Create
“Should Be”
Conduct
Perform
Cost/Benefit Pilot
Validate
Alternatives FMEA Improvement
Process Map Analysis