Decoding the DNA of the

Toyota Production System

by Steven Spear and H. Kent Bowen
Repri nt 99509
96 ar t wo r k by amy n i n g
T
h e To yo t a Pr o d u c t i o n Sy s t e m h a s
l ong been hai l ed as the source of Toyotas
outstandi ng performance as a manufacturer.
The systems di sti ncti ve practi ces i ts kanban cards
and qual i ty ci rcl es, for i nstance have been wi del y
i ntroduced el sewhere. Indeed, fol l owi ng thei r own
i nternal efforts to benchmark the worl ds best man-
ufacturi ng compani es, GM, Ford, and Chrysl er have
i ndependentl y created major i ni ti ati ves to devel op
Toyota-l i ke producti on systems. Compani es that
have tri ed to adopt the system can be found i n el ds
as di verse as aerospace, consumer products, metal s
processi ng, and i ndustri al products.
Whats curi ous i s that few manufacturers have
managed to i mi tate Toyota successful l y even
though the company has been extraordi nari l y open
about i ts practi ces. Hundreds of thousands of exec-
uti ves from thousands of busi nesses have toured
Toyotas pl ants i n Japan and the Uni ted States.
Frustrated by thei r i nabi l i ty to repl i cate Toyotas
performance, many vi si tors assume that the secret
of Toyotas success must l i e i n i ts cul tural roots.
But thats just not the case. Other Japanese compa-
ni es, such as Ni ssan and Honda, have fal l en short
of Toyotas standards, and Toyota has successful l y
i ntroduced i ts producti on system al l around the
worl d, i ncl udi ng i n North Ameri ca, where the com-
pany i s thi s year bui l di ng over a mi l l i on cars, mi ni -
vans, and l i ght trucks.
So why has i t been so di ffi cul t to decode the Toy-
ota Producti on System? The answer, we bel i eve, i s
that observers confuse the tool s and practi ces they
see on thei r pl ant vi si ts wi th the system i tsel f. That
makes i t i mpossi bl e for them to resol ve an apparent
paradox of the system namel y, that acti vi ti es, con-
necti ons, and producti on ows i n a Toyota factory
are ri gi dl y scri pted, yet at the same ti me Toyotas
operati ons are enormousl y exi bl e and adaptabl e.
Acti vi ti es and processes are constantl y bei ng chal -
l enged and pushed to a hi gher l evel of performance,
enabl i ng the company to conti nual l y i nnovate and
i mprove.
To understand Toyotas success, you have to un-
ravel the paradox you have to see that the ri gi d
speci cati on i s the very thi ng that makes the exi -
bi l i ty and creati vi ty possi bl e. Thats what we came
to real i ze after an extensi ve, four-year study of the
Toyota Producti on System i n whi ch we exami ned
the i nner worki ngs of more than 40 pl ants i n the
Uni ted States, Europe, and Japan, some operati ng
accordi ng to the system, some not. We studi ed both
process and di screte manufacturers whose products
ranged from prefabri cated housi ng, auto parts and
nal auto assembl y, cel l phones, and computer
pri nters to i njecti on-mol ded pl asti cs and al umi num
Copyri ght 1999 by the Presi dent and Fel l ows of Harvard Col l ege. Al l ri ghts reserved. 97
by Steven Spear and H.Kent Bowen
The
Toyot a
st or y has
been int ensively
researched and
painst akingly
document ed,
yet what really
happens
inside t he
company
remains a
myst er y.
Heres new
insight int o
t he unspoken
rules t hat
give Toyot a
it s compet it ive
edge.
Decoding the
DNA of the
Toyota
Production
System
extrusi ons. We studi ed not onl y routi ne producti on
work but al so servi ce functi ons l i k e equi pment
mai ntenance, workers trai ni ng and supervi si on,
l ogi sti cs and materi al s handl i ng, and process desi gn
and redesi gn.
We found that, for outsi ders, the key i s to under-
stand that the Toyota Producti on System creates a
communi ty of sci enti sts. Whenever Toyota denes
a speci cati on, i t i s establ i shi ng sets of hypotheses
that can then be tested. In other words, i t i s fol l ow-
i ng the sci enti c method. To make any changes,
Toyota uses a ri gorous probl em-sol vi ng process
that requi res a detai l ed assessment of the current
state of affai rs and a pl an for i mprovement that i s,
i n effect, an experi mental test of the proposed
changes. Wi th anythi ng l ess than such sci enti c
ri gor, change at Toyota woul d amount to l i ttl e
more than random tri al and error a bl i ndfol ded
wal k through l i fe.
The fact that the sci enti fi c method i s so i n-
grai ned at Toyota expl ai ns why the hi gh degree of
speci cati on and structure at the company does not
promote the command and control envi ronment
one mi ght expect. Indeed, i n watchi ng peopl e doi ng
thei r jobs and i n hel pi ng to desi gn producti on pro-
cesses, we l earned that the system actual l y sti mu-
l ates workers and managers to engage i n the ki nd of
experi mentati on that i s wi del y recogni zed as the
cornerstone of a l earni ng organi zati on. That i s
what di sti ngui shes Toyota from al l the other com-
pani es we studi ed.
The Toyota Producti on System and the sci enti c
method that underpi ns i t were not i mposed on
Toyota they were not even chosen consci ousl y.
The system grew natural l y out of the worki ngs of
the company over ve decades. As a resul t, i t has
never been wri tten down, and Toyotas workers
often are not abl e to arti cul ate i t. Thats why i ts so
hard for outsi ders to grasp. In thi s arti cl e, we attempt
to l ay out how Toyotas system works. We try to
make expl i ci t what i s i mpl i ci t. We descri be four
pri nci pl es three rul es of desi gn, whi ch show how
Toyota sets up al l i ts operati ons as experi ments,
and one rul e of i mprovement, whi ch descri bes how
Toyota teaches the sci enti c method to workers at
every l evel of the organi zati on. It i s these rul es and
not the speci c practi ces and tool s that peopl e ob-
serve duri ng thei r pl ant vi si ts that i n our opi ni on
form the essence of Toyotas system. That i s why we
thi nk of the rul es as the DNA of the Toyota Produc-
ti on System. Lets take a cl oser l ook at those rul es
(for a summary, see the si debar The Four Rul es ).
Rule 1:How People Work
Toyotas managers recogni ze that the devi l i s i n the
detai l s; thats why they ensure that al l work i s hi ghl y
speci ed as to content, sequence, ti mi ng, and out-
come. When a cars seat i s i nstal l ed, for i nstance,
the bol ts are al ways ti ghtened i n the same order, the
ti me i t takes to turn each bol t i s speci ed, and so
i s the torque to whi ch the bol t shoul d be ti ghtened.
Such exactness i s appl i ed not onl y to the repeti ti ve
moti ons of producti on workers but al so to the acti v-
i ti es of al l peopl e regardl ess of thei r functi onal spe-
ci al ty or hi erarchi cal rol e. The requi rement that
every acti vi ty be speci ed i s the rst unstated rul e
of the system. Put thi s bal dl y, the rul e seems si m-
pl e, somethi ng youd expect everyone to under-
stand and be abl e to fol l ow easi l y. But i n real i ty,
most managers outsi de Toyota and i ts partners
dont take thi s approach to work desi gn and execu-
ti on even when they thi nk they do.
98 h a r va r d bu si n ess r evi ew SeptemberOctober 1999
d e c o d i n g t h e d n a o f t h e t o yo t a pr o d u c t i o n s y s t e m
Steven Spear is an assistant professor of business admin-
istration at Harvard Business School in Boston.
H. Kent Bowenis the Bruce Rauner Professor of Business
Administration, also at Harvard Business School. Pro-
fessor Bowen is the coauthor of Regaining the Lead in
Manufacturing (HBR SeptemberOctober 1994).
The tacit knowledge that underlies the Toyota
Production System can be captured in four basic
rules.These rules guide the design,operation,and
improvement of every activity,connection,and
pathway for every product and service.The rules are
as follows:
Rule 1: All work shall be highly specied as to
content,sequence,timing,and outcome.
Rule 2: Every customer-supplier connection must be
direct,and there must be an unambiguous yes-or-no
way to send requests and receive responses.
Rule 3: The pathway for every product and service
must be simple and direct.
Rule 4: Any improvement must be made in
accordance with the scientic method,under the
guidance of a teacher,at the lowest possible level in
the organization.
All the rules require that activities,connections,and
ow paths have built-in tests to signal problems
automatically.It is the continual response to problems
that makes this seemingly rigid system so exible and
adaptable to changing circumstances.
The Fo ur Rules
Lets l ook at how operators at a typi cal U.S. auto
pl ant i nstal l the front passenger seat i nto a car.
They are supposed to take four bol ts from a card-
board box, carry them and a torque wrench to the
car, ti ghten the four bol ts, and enter a code i nto a
computer to i ndi cate that the work has been done
wi thout probl ems. Then they wai t for the next car
to arri ve. New operators are usual l y trai ned by ex-
peri enced workers, who teach by demonstrati ng
what to do. A seasoned col l eague mi ght be avai l abl e
to hel p a new operator wi th any di ffi cul ti es, such as
fai l i ng to ti ghten a bol t enough or forgetti ng to enter
the computer code.
Thi s sounds strai ghtforward, so whats wrong
wi th i t? The probl em i s that those speci cati ons ac-
tual l y al l ow and even assume consi derabl e vari a-
ti on i n the way empl oyees do thei r work. Wi thout
anyone real i zi ng i t, there i s pl enty of scope for a
new operator to put the seat i nto the vehi cl e di ffer-
entl y than an experi enced empl oyee woul d. Some
operators mi ght put the front bol ts i n after the rear
bol ts; some mi ght do i t the other way around. Some
operators mi ght put each bol t i n and then ti ghten
them al l ; others mi ght ti ghten as they go al ong. Al l
thi s vari ati on transl ates i nto poorer qual i ty, l ower
producti vi ty, and hi gher costs. More i mportant, i t
hi nders l earni ng and i mprovement i n the organi za-
ti on because the vari ati ons hi de the l i nk between
how the work i s done and the resul ts.
At Toyotas pl ants, because operators (new and
ol d, juni or and supervi sory) fol l ow a wel l -dened
sequence of steps for a parti cul ar job, i t i s i nstantl y
cl ear when they devi ate from the speci cati ons.
Consi der how work ers at Toyotas Georgetown,
Kentucky, pl ant i nstal l the ri ght-front seat i nto a
Camry. The work i s desi gned as a sequence of seven
tasks, al l of whi ch are expected to be compl eted
i n 55 seconds as the car moves at a xed speed
through a workers zone. If the producti on worker
nds hi msel f doi ng task 6 (i nstal l i ng the rear seat-
bol ts) before task 4 (i nstal l i ng the front seat-bol ts),
then the job i s actual l y bei ng done di fferentl y than
i t was desi gned to be done, i ndi cati ng that some-
thi ng must be wrong. Si mi l arl y, i f after 40 seconds
the worker i s sti l l on task 4, whi ch shoul d have
been compl eted after 31 seconds, then somethi ng,
too, i s ami ss. To make probl em detecti on even si m-
pl er, the l ength of the oor for each work area i s
marked i n tenths. So i f the worker i s passi ng the
si xth of the ten oor marks (that i s, i f he i s 33 sec-
onds i nto the cycl e) and i s sti l l on task 4, then he
and hi s team l eader know that he has fal l en behi nd.
Si nce the devi ati on i s i mmedi atel y apparent, worker
and supervi sor can move to correct the probl em
ri ght away and then determi ne how to change the
speci cati ons or retrai n the worker to prevent a re-
currence. (See the si debar How Toyotas Workers
Learn the Rul es for a short descri pti on of the
process by whi ch workers l earn how to desi gn work
i n thi s way.)
Even compl ex and i nfrequent acti vi ti es, such as
trai ni ng an i nexperi enced workforce at a new pl ant,
l aunchi ng a new model , changi ng over a producti on
l i ne, or shi fti ng equi pment from one part of a pl ant
to another, are desi gned accordi ng to thi s rul e. At
one of Toyotas suppl i ers i n Japan, for exampl e,
equi pment from one area of the pl ant was moved to
create a new producti on l i ne i n response to changes
i n demand for certai n products. Movi ng the machi n-
ery was broken i nto 14 separate acti vi ti es. Each ac-
ti vi ty was then further subdi vi ded and desi gned as
a seri es of tasks. A speci c person was assi gned to do
each task i n a speci ed sequence. As each of the ma-
chi nes was moved, the way the tasks were actual l y
h a r va r d bu si n ess r evi ew SeptemberOctober 1999
99
d e c o d i n g t h e d n a o f t h e t o yo t a pr o d u c t i o n s y s t e m
If the rules of the Toyota Production System arent
explicit,how are they transmitted? Toyotas managers
dont tell workers and supervisors specically how to
do their work.Rather,they use a teaching and learning
approach that allows their workers to discover the
rules as a consequence of solving problems.For
example,the supervisor teaching a person the
principles of the rst rule will come to the work site
and,while the person is doing his or her job,ask a
series of questions:
"
How do you do this work?
" How do you know you are doing
this work correctly?
" How do you know that the outcome
is free of defects?
" What do you do if you have
a problem?
This continuing process gives the person increasingly
deeper insights into his or her own specic work.From
many experiences of this sort,the person gradually
learns to generalize how to design all activities
according to the principles embodied in rule 1.
All the rules are taught in a similar Socratic fashion
of iterative questioning and problem solving.
Although this method is particularly effective for
teaching,it leads to knowledge that is implicit.
Consequently,the Toyota Production System has so
far been transferred successfully only when managers
have been able and willing to engage in a similar
process of questioning to facilitate learning by doing.
How Toyot as Workers
Learn t he Rules
done was compared wi th what was expected ac-
cordi ng to the ori gi nal desi gn, and di screpanci es
were i mmedi atel y si gnal ed.
In cal l i ng for peopl e to do thei r work as a hi ghl y
speci ed sequence of steps, rul e 1 forces them to
test hypotheses through acti on. Performi ng the
acti vi ty tests the two hypotheses i mpl i ci t i n i ts de-
si gn: rst, that the person doi ng the acti vi ty i s capa-
bl e of performi ng i t correctl y and, second, that per-
formi ng the acti vi ty actual l y creates the expected
outcome. Remember the seat i nstal l er? If he cant
i nsert the seat i n the speci ed way wi thi n the spec-
i ed amount of ti me, then he i s cl earl y refuti ng at
l east one of these two hypotheses, thereby i ndi cat-
i ng that the acti vi ty needs to be redesi gned or the
worker needs to be trai ned.
Rule 2:How People Connect
Where the rst rul e expl ai ns how peopl e perform
thei r i ndi vi dual work acti vi ti es, the second rul e
expl ai ns how they connect wi th one another. We
express thi s rul e as fol l ows: every connecti on must
be standardi zed and di rect, unambi guousl y speci fy-
i ng the peopl e i nvol ved, the form and quanti ty of
the goods and servi ces to be provi ded, the way re-
quests are made by each customer, and the expected
ti me i n whi ch the requests wi l l be met. The rul e
creates a suppl i er-customer rel ati onshi p between
each person and the i ndi vi dual who i s responsi bl e
for provi di ng that person wi th each speci c good or
servi ce. As a resul t, there are no gray zones i n deci d-
i ng who provi des what to whom and when. When a
worker makes a request for parts, there i s no confu-
si on about the suppl i er, the number of uni ts re-
qui red, or the ti mi ng of the del i very. Si mi l arl y,
when a person needs assi stance, there i s no confu-
si on over who wi l l provi de i t, how the hel p wi l l be
tri ggered, and what servi ces wi l l be del i vered.
The real questi on that concerns us here i s whether
peopl e i nteract di fferentl y at Toyota than they do at
other compani es. Lets return to our seat i nstal l er.
When he needs a new contai ner of pl asti c bol t cov-
ers, he gi ves a request to a materi al s handl er, who i s
100 h a r va r d bu si n ess r evi ew SeptemberOctober 1999
d e c o d i n g t h e d n a o f t h e t o yo t a pr o d u c t i o n s y s t e m
When organizations are managed according to the four rules,
individuals are repeatedly conducting experiments,testing in
operation the hypotheses built into the designs of individual
work activities,customer-supplier connections,pathways,and
improvement efforts.The hypotheses,the way they are tested,
and the response if they are refuted are summarized below.
The Experi ment s o f t he Toyot a Pro duct i on Syst em
Rule
1
2
3
4
Hypotheses
The person or machine can do the
activity as specied.
If the activity is done as specied,the
good or service will be defect free.
Customersrequests will be for goods
and services in a specic mix and
volume.
The supplier can respond to
customersrequests.
Every supplier that is connected
to the ow path is required.
Any supplier not connected to
the ow path is not needed.
A specic change in an activity,
connection,or ow path will improve
cost,quality,lead time,batch size,or
safety by a specic amount.
Signs of a problem
The activity is not done as
specied.
The outcome is defective.
Responses dont keep pace
with requests.
The supplier is idle,waiting
for requests.
A person or machine is
not actually needed.
A nonspecied supplier
provides an intermediate
good or service.
The actual result is different
from the expected result.
Responses
Determine the true skill level of the person
or the true capability of the machine
and train or modify as appropriate.
Modify the design activity.
Determine the true mix and volume of
demand and the true capability of the
supplier;retrain,modify activities,or reassign
customer-supplier pairs as appropriate.
Determine why the supplier was unnecessary,
and redesign the ow path.
Learn why the nonspecied supplier was
actually required,and redesign the ow path.
Learn how the activity was actually performed
or the connection or ow path was actually
operated.Determine the true effects of the
change.Redesign the change.
den and are nei ther shared nor resol ved company-
wi de. The si tuati on i s made worse i f workers begi n
to sol ve probl ems themsel ves and then arbi trari l y
deci de when the probl em i s bi g enough to warrant a
cal l for hel p. Probl ems mount up and onl y get
sol ved much l ater, by whi ch ti me val uabl e i nfor-
mati on about the real causes of the probl em may
have been l ost.
Rule 3:How the Production Line
Is Constructed
Al l producti on l i nes at Toyota have to be set up so
that every product and servi ce ows al ong a si mpl e,
speci ed path. That path shoul d not change unl ess
the producti on l i ne i s expressl y redesi gned. In pri n-
ci pl e, then, there are no forks or l oops to convol ute
the ow i n any of Toyotas suppl y chai ns. Thats the
thi rd rul e.
To get a concrete i dea of what that means, l ets
return to our seat i nstal l er. If he needs more pl asti c
bol t covers, he orders them from the speci c mater-
i al handl er responsi bl e for provi di ng hi m wi th bol t
covers. That desi gnated suppl i er makes requests to
hi s own desi gnated suppl i er at the off-l i ne store i n
the factory who, i n turn, makes requests di rectl y to
hi s desi gnated suppl i er at the bol t cover factorys
shi ppi ng dock . I n thi s way, the producti on l i ne
l i nks each person who contri butes to the produc-
ti on and del i very of the product, from the Toyota
factory, through the mol di ng company, to even the
pl asti c pel l et manufacturer.
The poi nt i s that when producti on l i nes are de-
si gned i n accordance wi th rul e 3, goods and servi ces
do not ow to the next avai l abl e person or machi ne
but to a specic person or machi ne. If for some rea-
son that person or machi ne i s not avai l abl e, Toyota
wi l l see i t as a probl em that mi ght requi re the l i ne
to be redesi gned.
The sti pul ati on that every product fol l ow a si m-
pl e, prespeci ed path doesnt mean that each path
i s dedi cated to onl y one parti cul ar product, how-
ever. Qui te the contrary: each producti on l i ne at a
Toyota pl ant typi cal l y accommodates many more
types of products than i ts counterparts do at other
compani es.
The thi rd rul e doesnt appl y onl y to products i t
appl i es to servi ces, l i ke hel p requests, as wel l . If our
seat i nstal l er, for exampl e, needs hel p, that too
comes from a si ngl e, speci ed suppl i er. And i f that
suppl i er cant provi de the necessary assi stance, she,
i n turn, has a desi gnated hel per. In some of Toyotas
pl ants, thi s pathway for assi stance i s three, four, or
ve l i nks l ong, connecti ng the shop oor worker
to the pl ant manager.
the desi gnated bol t-cover suppl i er. Commonl y,
such a request i s made wi th a kanban, a l ami nated
card that speci es the parts i denti cati on number,
the quanti ty of parts i n the contai ner, and the l oca-
ti ons of the part suppl i er and of the worker (the cus-
tomer) who wi l l i nstal l i t. At Toyota, kanban cards
and other devi ces l i ke andon cords set up di rect
l i nks between the suppl i ers and the customers. The
connecti ons are as smooth as the passi ng of the baton
i n the best Ol ympi c rel ay teams because they are
just as careful l y thought out and executed. For ex-
ampl e, the number of parts i n a contai ner and the
number of contai ners i n ci rcul ati on for any gi ven
part are determi ned by the physi cal real i ti es of the
producti on system the di stances, the changeover
ti mes, and so on. Li kewi se, the number of workers
per team i s determi ned by the types of probl ems ex-
pected to occur, the l evel of assi stance the team
members need, and the ski l l s and capabi l i ti es of the
teams l eader.
Other compani es devote substanti al resources to
coordi nati ng peopl e, but thei r connecti ons general l y
arent so di rect and unambi guous. In most pl ants, re-
quests for materi al s or assi stance often take a con-
vol uted route from the l i ne worker to the suppl i er
vi a an i ntermedi ary. Any supervi sor can answer any
cal l for hel p because a speci c person has not been
assi gned. The di sadvantage of that approach, as Toy-
ota recogni zes, i s that when somethi ng i s everyones
probl em i t becomes no ones probl em.
The requi rement that peopl e respond to suppl y re-
quests wi thi n a speci c ti me frame further reduces
the possi bi l i ty of vari ance. That i s especi al l y true i n
servi ce requests. A worker encounteri ng a probl em
i s expected to ask for assi stance at once. The desi g-
nated assi stant i s then expected to respond i mmedi -
atel y and resol ve the probl em wi thi n the workers
cycl e ti me. If the worker i s i nstal l i ng a front seat
every 55 seconds, say, then a request for hel p must
be answered and deal t wi th i n l ess than the 55 sec-
onds. If the probl em cannot be resol ved i n l ess than
55 seconds, that fai l ure i mmedi atel y chal l enges
the hypotheses i n thi s customer-suppl i er connec-
ti on for assi stance. Perhaps the request si gnal i s
ambi guous. Perhaps the desi gnated assi stant has too
many other requests for hel p and i s busy or i s not
a capabl e probl em sol ver. Constantl y testi ng the hy-
potheses i n thi s way keeps the system exi bl e, mak-
i ng i t possi bl e to adjust the system conti nual l y and
constructi vel y.
The stri ki ng thi ng about the requi rement to ask
for hel p at once i s that i t i s often counteri ntui ti ve
to managers who are accustomed to encouragi ng
workers to try to resol ve probl ems on thei r own be-
fore cal l i ng for hel p. But then probl ems remai n hi d-
h a r va r d bu si n ess r evi ew SeptemberOctober 1999
101
d e c o d i n g t h e d n a o f t h e t o yo t a pr o d u c t i o n s y s t e m
The thi rd rul e runs contrary to conventi onal wi s-
dom about producti on l i nes and pool i ng resources
even contrary to how most peopl e thi nk the Toyota
Producti on System works. Accordi ng to recei ved
wi sdom, as a product or servi ce i s passed down the
l i ne, i t shoul d go to the next machi ne or person avai l -
abl e to process i t further. Si mi l arl y, most peopl e as-
sume that hel p shoul d come from the rst avai l abl e
person rather than from a speci c person. At one
auto parts suppl i er we studi ed, for exampl e, most of
the parts coul d be stamped on more than one press
machi ne and wel ded at more than one wel di ng sta-
ti on. Before the company adopted the Toyota sys-
tem, i ts practi ce was to pass each part on to the rst
avai l abl e press machi ne and to the rst avai l abl e
wel der. When the pl ant swi tched over, under Toy-
otas gui dance, each type of part fol l owed onl y one
producti on path through the pl ant.
By requi ri ng that every pathway be speci ed, the
rul e ensures that an experi ment wi l l occur each
ti me the path i s used. Here the hypotheses embed-
ded i n a pathway desi gned accordi ng to rul e 3 are
that every suppl i er connected to the pathway i s nec-
essary, and any suppl i er not connected i s not neces-
sary. I f workers at the auto parts suppl i er found
themsel ves wanti ng to di vert producti on to another
machi ne or wel di ng stati on, or i f they began turni ng
for hel p to someone other than thei r desi gnated
hel pers, theyd concl ude that thei r actual demand or
capaci ty di dnt match thei r expectati ons. And there
102 h a r va r d bu si n ess r evi ew SeptemberOctober 1999
d e c o d i n g t h e d n a o f t h e t o yo t a pr o d u c t i o n s y s t e m
woul d al so be no ambi gui ty about whi ch press
or wel der was i nvol ved. Agai n, the workers woul d
revi si t the desi gn of thei r producti on l i ne. Thus
rul e 3, l i ke rul es 1 and 2, enabl es Toyota to conduct
experi ments and remai n exi bl e and responsi ve.
Rule 4:How to Improve
Identi fyi ng probl ems i s just the rst step. For peo-
pl e to consi stentl y make effecti ve changes, they
must know how to change and who i s responsi bl e
for maki ng the changes. Toyota expl i ci tl y teaches
peopl e how to i mprove, not expecti ng them to l earn
stri ctl y from personal experi ence. Thats where the
rul e for i mprovement comes i n. Speci cal l y, rul e 4
sti pul ates that any i mprovement to producti on
acti vi ti es, to connecti ons between workers or ma-
chi nes, or to pathways must be made i n accordance
wi th the sci enti c method, under the gui dance of
a teacher, and at the l owest possi bl e organi zati onal
l evel . Lets l ook rst at how Toyotas peopl e l earn
the sci enti c method.
How People Learn to Improve.In 1986, Ai si n Sei ki ,
a Toyota Group company that made compl ex prod-
ucts such as power trai ns for the auto i ndustry, cre-
ated a l i ne to manufacture mattresses to absorb ex-
cess capaci ty i n one of i ts pl ants.
Si nce 1986, i ts range has grown
from 200 to 850 types of mat-
tresses, i ts vol ume has grown
from 160 mattresses per day to
550, and i ts producti v i ty has
doubl ed. Heres an exampl e of
how they di d i t.
On one of our vi si ts to thi s
pl ant, we studi ed a team of mat-
tr ess assembl y wor k er s wh o
were bei ng taught to i mprove
thei r probl em-sol vi ng ski l l s by
r edesi gn i n g th ei r own wor k .
Ini ti al l y, the workers had been
responsi bl e for doi ng onl y thei r
own standardi zed work ; they
had not been responsi bl e for
s ol v i n g pr obl ems . Th en th e
workers were assi gned a l eader
who trai ned them to frame prob-
l ems better and to for mul ate
and test hypotheses i n other
words, he taught them how to
use the sci enti fi c method to
desi gn thei r teams work i n accordance wi th the
rst three rul es. The resul ts were i mpressi ve. One
of the teams accompl i shments, for i nstance, was to
redesi gn the way edgi ng tape was attached to the
On- Demand Product i on at t he Ai si n Mat t ress Factory
1986 1988 1992 1996 1997
Styles 200 325 670 750 850
Units per day 160 230 360 530 550
Units per person 8 11 13 20 26
Productivity index 100 138 175 197 208
Finished-goods inventory (days) 30 2.5 1.8 1.5 1.5
Number of assembly lines 2 2 3 3 2
Aisin Seiki produces 850 varieties of mattresses,distinguished by size,rmness,
covering fabric,quilting pattern,and edge trim.Customers can order any one of
these in a retail store and have it delivered to their homes in three days,yet Aisin
maintains an inventory at the plant equal to just 1.5 days of demand.To be able to
do so,Aisin has made thousands of changes in individual work activities,in the
connections linking customers and suppliers of intermediate goods and services,
and to the overall production lines.This table captures how dramatic the results of
those changes have been.
mattresses, thereby reduci ng the defect rate by
90%. (See the exhi bi t On-Demand Producti on at
the Ai si n Mattress Factory. )
To make changes, peopl e are expected to pre-
sent the expl i ci t l ogi c of the hypotheses. Lets
l ook at what that can i nvol ve. Haji me Ohba, gen-
eral manager of the Toyota Suppl i er Support Cen-
ter, was vi si ti ng a factory i n whi ch one of TSSCs
consul tants was l eadi ng a trai ni ng and i mprove-
ment acti vi ty (for a descri pti on of the rol e of the
Toyota Producti on System promoti on centers,
see the si debar Toyotas Commi tment to Learn-
i ng ). The consul tant was hel pi ng factory em-
pl oyees and thei r supervi sor reduce the manufac-
turi ng l ead ti me of a parti cul ar l i ne, and Ohba
was there to eval uate the groups progress.
Group members began thei r presentati on by
descri bi ng the steps by whi ch thei r product was
created del i neati ng al l the probl ems they i denti -
ed when they had rst studi ed the process for
changi ng over a machi ne from maki ng one part
to maki ng another, and expl ai ni ng the speci c
changes they had made i n response to each of those
probl ems. They concl uded by sayi ng, When we
started, the changeover requi red 15 mi nutes. We
were hopi ng to reduce that by two-thi rds to
achi eve a ve-mi nute changeover so that we
coul d reduce batch si zes by two-thi rds. Because
of the modi fi cati ons we made, we achi ev ed
a changeover ti me of seven and a hal f mi nutes a
reducti on of one-hal f.
After thei r presentati on, Ohba asked why the
group members had not achi eved the ve-mi nute
goal they had ori gi nal l y establ i shed. They were a
bi t taken aback. After al l , they had reduced the
changeover ti me by 50%, yet Ohbas questi on
suggested he had seen opportuni ti es for even
greater i mprovement that they had mi ssed. They
offered expl anati ons havi ng to do wi th machi ne
compl exi ty, techni cal di ffi cul ty, and equi pment
upgrade costs. Ohba responded to these repl i es
wi th yet more questi ons, each one meant to push
the consul tant and the factory peopl e to arti cu-
l ate and chal l enge thei r most basi c assumpti ons
about what coul d and coul d not be changed as-
sumpti ons that both gui ded and constrai ned the
way they had sol ved thei r probl ems. Were they
su r e fou r bol ts w er e n ecessar y ? Mi gh t th e
changeover be accompl i shed wi th two? Were
they certai n that al l the steps they i ncl uded i n the
changeover were needed? Mi ght some be com-
bi ned or el i mi nated? In aski ng why they had not
achi eved the ve-mi nute goal , Ohba was not sug-
gesti ng that the team had fai l ed. Rather, he was
tryi ng to get them to real i ze that they had not
h a r va r d bu si n ess r evi ew SeptemberOctober 1999
103
d e c o d i n g t h e d n a o f t h e t o yo t a pr o d u c t i o n s y s t e m
All the organizations we studied that are managed
according to the Toyota Production System share an
overarching belief that people are the most signicant
corporate asset and that investments in their knowledge
and skills are necessary to build competitiveness.Thats
why at these organizations all managers are expected to
be able to do the jobs of everyone they supervise and also
to teach their workers how to solve problems according to
the scientic method.The leadership model applies as
much to the rst-level team leadersupervisors as it does
to those at the top of the organization.In that way,
everybody at Toyota shares in the development of human
resources.In effect,there is a cascading pathway for
teaching,which starts with the plant manager,that
delivers training to each employee.
To reinforce the learning and improvement process,
each plant and major business unit in the Toyota Group
employs a number of Toyota Production System
consultants whose primary responsibility is to help senior
managers move their organizations toward the ideal.
These learner-leader-teachersdo so by identifying ever
more subtle and difficult problems and by teaching
people how to solve problems scientically.
Many of these individuals have received intensive
training at Toyotas Operations Management Consulting
Division.OMCD was established in Japan as an outgrowth
of efforts by Taiichi Ohno one of the original architects of
the Toyota Production System to develop and diffuse the
system throughout Toyota and its suppliers.Many of
Toyotas top officers including Toyota Motors new
president,Fujio Cho have honed their skills within OMCD.
During their OMCD tenure,which can extend for a period
of years,Toyotas employees are relieved of all line
responsibilities and instead are charged with leading
improvement and training activities in the plants of Toyota
and its suppliers.By supporting all of Toyotas plant and
logistical operations in this way,OMCD serves as a training
center,building its consultantsexpertise by giving them
opportunities to solve many difficult problems and teach
others to do the same.
In 1992,Toyota founded the Toyota Supplier Support
Center (TSSC) in the United States to provide North
American companies with training in the Toyota
Production System.Modeled on OMCD,TSSC has given
workshops to more than 140 companies and direct
assistance to 80.Although most of these companies are
auto suppliers,few are exclusively Toyota suppliers;
participants come from other industries and from
universities,government organizations,and industry
associations.Indeed,much of the research for this paper
was derived from the experience of one of the authors,
who was a member of a TSSC team for ve months,
promoting the Toyota Production System at a plant that
supplies Toyota and two other auto assembly plants.
Toyot as Commi t ment t o Learni ng
ful l y expl ored al l thei r i mprovement opportuni ti es
because they had not questi oned thei r assumpti ons
deepl y enough.
There was a second reason for Ohbas persi s-
tence. He was tryi ng to show the group members
that thei r i mprovement acti vi ty had not been car-
ri ed out as a bona de experi ment. They had estab-
l i shed a goal of ve mi nutes based on the premi se
that faster changeovers and smal l er batches are bet-
ter than sl ower changeovers and l arger batches. But
here they were confusi ng goal s wi th predi cti ons
based on hypotheses. The goal was not a predi cti on
of what they bel i eved they woul d achi eve through
the speci c i mprovement steps they pl anned to
take. As a resul t, they had not desi gned the i m-
provement effort as an experi ment wi th an expl i ci t,
cl earl y arti cul ated, veri fi abl e hypothesi s of the
form, If we make the fol l owi ng speci c changes,
we expect to achi eve thi s speci c outcome. Al -
though they had reduced the changeover ti me con-
si derabl y, they had not tested the hypotheses i m-
pl i ci t i n thei r effort. For Ohba, i t was cri ti cal that
the workers and thei r supervi sor real i ze that how
they made changes was as i mpor tant as what
changes they made.
Who Does the Improvement. Frontl i ne workers
mak e the i mprovements to thei r own jobs, and
thei r supervi sors provi de di recti on and assi stance
as teachers. If somethi ng i s wrong wi th the way a
worker connects wi th a parti cul ar suppl i er wi thi n
the i mmedi ate assembl y area, the two of them
make i mprovements, wi th the assi stance of thei r
common supervi sor. The Ai si n team we descri bed
earl i er, for exampl e, consi sted of the assembl y l i ne
workers and the supervi sor, who was al so thei r i n-
structor. When changes are made on a l arger scal e,
Toyota ensures that i mprovement teams are created
consi sti ng of the peopl e who are di rectl y affected
and the person responsi bl e for supervi si ng the path-
ways i nvol ved.
Thus the process remai ns the same even at the
hi ghest l ev el s. At Ai si ns mattress factory, we
found that the pl ant manager took responsi bi l i ty
for l eadi ng the change from three producti on l i nes
back to two (the number had ri sen to three to cope
wi th an i ncrease i n product types). He was i nvol ved
not just because i t was a bi g change but al so be-
cause he had operati onal responsi bi l i ty for oversee-
i ng the way work owed from the feeder l i nes to
the nal assembl y l i nes. In thi s way, Toyota en-
104 h a r va r d bu si n ess r evi ew SeptemberOctober 1999
d e c o d i n g t h e d n a o f t h e t o yo t a pr o d u c t i o n s y s t e m
Toyota does not consider any of the tools or practices
such as kanbans or andon cords,which so many outsiders
have observed and copied as fundamental to the Toyota
Production System.Toyota uses them merely as temporary
responses to specic problems that will serve until a
better approach is found or conditions change.Theyre
referred to as countermeasures,rather than solutions,
because that would imply a permanent resolution to a
problem.Over the years,the company has developed
a robust set of tools and practices that it uses as
countermeasures,but many have changed or even been
eliminated as improvements are made.
So whether a company does or does not use any
particular tool or practice is no indication that it is truly
applying Toyotas rules of design and improvement.In
particular,contrary to the impression that the concept of
zero inventory is at the heart of the Toyota system,weve
observed many cases in which Toyota actually built up its
inventory of materials as a countermeasure.The ideal
system would in fact have no need for inventory.But,in
practice,certain circumstances may require it:
" Unpredictable downtime or yields. Sometimes a
person or a machine is unable to respond on demand
when a request is made because of an unexpected
mechanical breakdown.For this reason,safety stock
is held to protect the customer against random
occurrences.The person responsible for ensuring the
reliability of a machine or process owns that inventory
and strives to reduce the frequency and length of
downtimes so that the amount of the safety stock can
be reduced.
" Time-consuming setups. Difficulties in switching a
machine from processing one kind of product to another
can prevent a supplier from responding immediately.
Therefore,suppliers will produce the product in batch
sizes greater than one and hold the excess as inventory so
it can respond immediately to the customer.Of course,
suppliers will continually try to reduce the changeover
time to keep batch sizes and stores of inventory as small
as possible.Here,the owners of both the problem and the
countermeasure are the machine operator and the team
leader,who are responsible for reducing changeover times
and batch sizes.
" Volatility in the mix and volume of customer demand.
In some cases,variations in customersneeds are so large
and unpredictable that it is impossible for a plant to
adjust its production to them quickly enough.In those
instances,buffer stock is kept at or near the shipping point
Count ermeasures i n t he Toyot a Product i on Syst em
sures that probl em sol vi ng and l earni ng take pl ace
at al l l evel s of the company. Of course, as we have
al ready seen, Toyota wi l l bri ng i n external experts
as necessary to ensure the qual i ty of the l earni ng
process.
In the l ong term, the organi zati onal structures of
compani es that fol l ow the Toyota Producti on Sys-
tem wi l l shi ft to adapt to the nature and frequency
of the probl ems they encounter. Si nce the organi za-
ti onal changes are usual l y bei ng made at a very l ow
l evel , however, they can be hard for outsi ders to de-
tect. Thats because i t i s the nature of the probl ems
that determi nes who shoul d sol ve them and how the
organi zati on i s desi gned. One consequence i s that
di fferent organi zati onal structures coexi st qui te
happi l y even i n the same pl ant.
Consi der Toyotas engi ne-machi ni ng pl ant i n
Kami go, Japan. The pl ant has two machi ne di vi -
si ons, each of whi ch has three i ndependent produc-
ti on shops. When we vi si ted i n summer 1998, the
producti on peopl e i n the rst machi ne di vi si on an-
swered to shop heads, and the process engi neers
answered di rectl y to the head of the di vi si on. How-
ever, i n the second machi ne di vi si on, the engi neers
were di stri buted among the three shops and, l i ke
the producti on workers, answered to the vari ous
shop heads. Nei ther organi zati onal structure i s i n-
herentl y superi or. Rather, the peopl e we i nter-
vi ewed expl ai ned, probl ems i n the rst di vi si on
happened to create a si tuati on that requi red the
engi neers to l earn from one another and to pool en-
gi neeri ng resources. By contrast, the probl ems that
arose i n the second di vi si on requi red the produc-
ti on and engi neeri ng peopl e to cooperate at the l evel
of the i ndi vi dual shops. Thus the organi zati onal
di fferences reect the fact that the two di vi si ons
encountered di fferent probl ems.
Toyotas Notion of the Ideal
By i ncul cati ng the sci enti c method at al l l evel s of
the work force, Toyota ensures that peopl e wi l l
cl earl y state the expectati ons they wi l l be testi ng
when they i mpl ement the changes they hav e
pl anned. But beyond thi s, we found that peopl e i n
compani es fol l owi ng the Toyota Producti on Sys-
tem share a common goal . They have a common
sense of what the i deal producti on system woul d
be, and that shared vi si on moti vates them to make
i mprovements beyond what woul d be necessary
merel y to meet the current needs of thei r customers.
Thi s noti on of the i deal i s very pervasi ve, and we
bel i eve i t i s essenti al to understandi ng the Toyota
Producti on System.
When they speak of the i deal , workers at Toyota
do not mean somethi ng phi l osophi cal l y abstract.
They have a concrete deni ti on i n mi nd, one that i s
remarkabl y consi stent throughout the company.
Very speci cal l y, for Toyotas workers, the output
of an i deal person, group of peopl e, or machi ne:
" i s defect free (that i t, i t has the features and perfor-
mance the customer expects);
" can be del i vered one request at a ti me (a batch si ze
of one);
" can be suppl i ed on demand i n the versi on requested;
" can be del i vered i mmedi atel y;
" can be produced wi thout wasti ng any materi al s,
l abor, energy, or other resources (such as costs as-
soci ated wi th i nventory); and
" can be produced i n a work envi ronment that i s
safe physi cal l y, emoti onal l y, and professi onal l y
for every empl oyee.
We consi stentl y found peopl e at pl ants that used
the Toyota Producti on System maki ng changes that
pushed operati ons toward thi s i deal . At one com-
pany that produced el ectromechani cal products, for
exampl e, we found that workers had come up wi th
a number of i ngeni ous error-detecti ng gauges that
h a r va r d bu si n ess r evi ew SeptemberOctober 1999
105
d e c o d i n g t h e d n a o f t h e t o yo t a pr o d u c t i o n s y s t e m
as a countermeasure.The buffer stock also serves as a
signal to production and sales managers that the person
who works most directly with the customer must help
that customer eliminate the underlying causes of any
preventable swings in demand.
In many cases,the same type of product is held in
different types of inventory.Toyota does not pool its
various kinds of inventory,even though doing so would
reduce its inventory needs in the short term.That might
sound paradoxical for a management system so popularly
known to abhor waste.But the paradox can be resolved
when we recognize that Toyotas managers and workers
are trying to match each countermeasure to each
problem.
Theres no link between the reason for keeping safety
stock process unreliability and the reason for keeping
buffer stock uctuations in customer demand.To pool
the two would make it hard to distinguish between the
separate activities and customer-supplier connections
involved.The inventory would have many owners,and
the reasons for its use would become ambiguous.
Pooling the inventory thus muddles both the ownership
and cause of the problems,making it difficult to introduce
improvements.
generated a si mpl e, unambi guous yes-or-no si gnal
to i ndi cate whether thei r output was free of defects
as speci ed i n the i deal . At yet another pl ant, whi ch
manufactures i njecti on-mol ded parts, we found that
workers had reduced the ti me i t took to change a
l arge mol di ng di e from an al ready speedy ve mi n-
utes to three mi nutes. Thi s al l owed the company to
reduce the batch si zes of each part i t produced by
40%, bri ngi ng i t cl oser to the i deal batch si ze of one.
As Toyota moves toward the i deal , i t may temporar-
i l y hol d one of i ts di mensi ons to be more i mportant
than another. Someti mes thi s can resul t i n practi ces
that go agai nst the popul ar vi ew of Toyotas opera-
ti ons. We have seen cases where Toyota keeps hi gher
l evel s of i nventory or produces i n batch si zes l arger
than observers general l y expect of a just-i n-ti me op-
erati on, as we descri be i n the si debar Countermea-
sures i n the Toyota Producti on System.
Toyotas i deal state shares many features of the
popul ar noti on of mass customi zati on the abi l i ty
to create vi rtual l y i nni te vari ati ons of a product
as effi ci entl y as possi bl e and at the l owest possi bl e
cost. I n the fi nal anal ysi s, Toyotas i deal pl ant
woul d i ndeed be one where a Toyota customer
coul d dri ve up to a shi ppi ng dock, ask for a cus-
tomi zed product or serv i ce, and get i t at once
at the l owest possi bl e pri ce and wi th no defects. To
the extent that a Toyota pl ant or a Toyota workers
acti vi ty fal l s short of thi s i deal , that shortcomi ng
i s a source of creati ve tensi on for further i mprove-
ment efforts.
The Organizational Impact of the Rules
If the rul es make compani es usi ng the Toyota Pro-
ducti on System a communi ty of sci enti sts perform-
i ng conti nual experi ments, then why arent these
organi zati ons i n a state of chaos? Why can one per-
son make a change wi thout adversel y affecti ng the
work of other peopl e on the producti on l i ne? How
can Toyota constantl y i ntroduce changes to i ts op-
erati ons whi l e keepi ng them runni ng at ful l ti l t? In
other words, how does Toyota i mprove and remai n
stabl e at the same ti me?
Once agai n, the answer i s i n the rul es. By maki ng
peopl e capabl e of and responsi bl e for doi ng and i m-
provi ng thei r own work, by standardi zi ng connec-
ti ons between i ndi vi dual customers and suppl i ers,
and by pushi ng the resol uti on of connecti on and
ow probl ems to the l owest possi bl e l evel , the rul es
create an organi zati on wi th a nested modul ar struc-
ture, rather l i ke tradi ti onal Russi an dol l s that come
one i nsi de the other. The great benet of nested,
modul ar organi zati ons i s that peopl e can i mpl e-
ment desi gn changes i n one part wi thout undul y af-
fecti ng other parts. Thats why managers at Toyota
can del egate so much responsi bi l i ty wi thout creat-
i ng chaos. Other compani es that fol l ow the rul es
wi l l al so nd i t possi bl e to change wi thout experi -
enci ng undue di srupti on.
Of course, the structures of other compani es
have features i n common wi th those that fol l ow
the Toyota Producti on System, but i n our research
we found no company that had them al l that di d not
fol l ow the system. It may turn out i n the end that
you can bui l d the structure onl y by i nvesti ng the
ti me Toyota has. But we bel i eve that i f a company
dedi cates i tsel f to masteri ng the rul es, i t has a bet-
ter chance of repl i cati ng Toyotas DNA and wi th
that, i ts performance.
Repri nt 99509 To pl ace an order, cal l 1-800-988-0886.
106 h a r va r d bu si n ess r evi ew SeptemberOctober 1999
d e c o d i n g t h e d n a o f t h e t o yo t a pr o d u c t i o n s y s t e m