6

PHltlPs

PHlllPs

O 1994,1995by GOAL/QPC Methuen, MA, U.s-!L

Any reproduction of any part of
this publication without the written
permission of GOAL/QPC is prohibited-

PhilipsQuality

Memory
Joggef"
Additional copies
(minimum order size 100copies)
may be ordered ftom:
Philips Intemational B.V.
Corporate Quality Bureau
P.O.box 218,Building Vo-p
5600MD Eindhoven
The Netherlands
Tel. +31.40.2784935
Fax + 31..40.27867L0
E-mail: NLEVNCCM.C795628

12 ^c code4322275 205 0'l

E=
Philips Ouality

Tools for Continuous InDrovement

Philips Corporate Quality Bureau thanks

GOAL/QPC for permission to customize their
pocket guide Tre Meory IoggelrM ,based. o
The Memory loggerrMirst written in 1985.
Michael Brassardand Diane Ritter substantially revisedtheiworkin 1994topoduceTfte
Menory loggelrM It. Mnory logger' is a U.S.
hademark of GOAL/QPC.

Contents
This handbook is designed to help you and every
erson rn your organiztion to imptoDpdqily tha
Tocedur$, syslcms.quality.Lost,and yiel.lsrelaledto Wt
jb, This continuous improvement process js the foc; d
this bool-shring th philosophy and tools that ale
hndamental to this effort.

Building the Winning Company

Philips Quality
Continuous Improvement
Selector Charts
Using the Quality Tools
Activity Network Diagram (AND)
. Gantt Chart
Afinity Dia8ram
Brainstorming
Cause& Effect/FishboneDagram
Check Sheet
Control Chart
Data Points
Flowchart
Force Field Analysis
Histogram
Inte elationship DigraPh (ID)
Matrix Diagram
Nominal Group Technique (NGT)
. Multivoting
Pareto Chart
105
Prioritization Matrices
Problem-Solving/kocess-ImPiovement Model:
115
Improvement Storyboad
132
Process Capability
1.37
Radar Chart
1.41
Run Chart
145
ScatterDiagrm
150
Team Guidelines
. Team Stages
150
. Conducting Effective Meetings......................
154
156
.............
TreeDiagram.............
. Process Decision Program Chart (PDPC) . .. 16-0

a-

THE PHILIPS WAY...

We delight customers
Our customers'expectations
influmceall ow
decisions
andactions.

F.

.L
Ei-!
-1H.

We value people as our geatest resource

Nothing can replacepeopleas lhe man source
of prosperity.
We deliver quality and excellencein all
actions
Excellentaaluefor customersis uhat we unnt.
We achieve premium retum on equity
Profit s aital to attract the capital needeitfor
ouruture.
We encourage entrepreneurial behavior at
all levels
We takeactionsthat benefitthe company.

rL-l
f--a
-.-l
ft-.t
f1
IE

Building the Winning Company

To bLriJd.tlewinning compny,we need to hve shared
vaiues,ctearobjectives,distinctiveskills nd competitive
processes.Centorion integratesinitiatives in all four areas.
SHARED VALUES
Involvement begins when we believe in what we ate
doing; when our personal values are in line with ,,The
Philips War" our company values.

CLEAB OBJECTIVES
A cledrview ot what we are trying to chievetogetheris
essenfiar.
uur obiechre is to build a winning company
which improves the quality of life o its customers and
employees,
DISTINCTIVE SKILLS
We build and sustain our comptitive advantage by
developing and maintaining our teclurological andrga'_
nizationalcompetencies. World-class capbilities ba;d
on the skills o teams and individuals build the Winning
Company.
COMPETITIVE PEOCESSES
Our work consisls of processes.These processesmust
oelver products and services which delight the cus_
tomer and satisfy all other stakeholders in tie company.

H
.,. OUR SHAREDVALUS

aH

Philips Quality provides the structure, methods and

toob for working in better ways nd for building com_
Perltrveprocesses,
This pocket guide focuseson the euality tools.

PhilipsQualrty
/,NPLEMENTING
THECOMPANYVAUES

Ls ,nake things better.

- l -!t

H*"ff fJij ?;l'i:",#:

H*;tI
I

crEt rd etoueh to Build the Winnhgtom_

r-- a humc ausromers ilrsl Choice.
As
ag",her in a comp texo r8anjzatio n,
i*
Tai"S

l",i:iffij
ffi ,tr;ifl,:J,:""",.ff
i

.%rr.t

!t

y':_t:J,*t ''..
,/

V-x_r

tranework has three d imensions.

. Ekt

Deptoymenr srrucures the rnteractronbe_

"n*"-,i: ffi,[,1fi
ii:. i:

l!

..

l=

. .on*.oous lmprovement
is basedon the ,,plan,
uo, Lheck, Act,, cycle,which sbuctures theinterac_

sa.tre m the company.

acrs

Management structures the interclion

ru::
H'fr#:,:iJ',,'1.*,":.,'";.1*
-

'

l.r

--;'-'l-

-7*--ttii

F
__.
.rr--t

ffifri;tle

Pnilips
prrner<
Quatiiy
in rh;ir

'
P,olic),deplo),rnentand processmanagement
creale
\:
f4P.arency nd cohesion required to ocus
errorts
on bu9inesspriorities aJld satisfy customer
requrrements. The PDCA cycle prvides
the
shucture for impro\ ing *or proc.sse, and
ac_
celerating Ieamine.

Contin uous lmprovemqrt

The PDCA Cycle

STRIVE FOR EXCELLENCE

Striving for excellencemeanssystemadcally
oaly work anct continuallv improving prolr*,
5}
towmg a cteardiscipline or problem solving d h_
ins fom mistakes-

Setobjectives
ald agreeon
acnons

The basis or process improvement is the .n-,

Check, Act" rycle. The emphasisis on corscirr l
ning and prevention rather than on fue.fightiry
what the Quality Tools help us to do_
The Quality Tools make data and ideas visiblq.
ptterns in numbers, focus attention and build
sus among team members.

Irnplement
the action
plan

The Quality Tools help a team to:

. achieve zero defects in all products and sevi(e
. prevent all waste of material, time and eort
. reduce variability in processes
. reduce the cycle time of processes

Quafitytools htp us to make thingsfu.

.
:

=a
-

Ih-

E_

Pl-t

where you want to be and what you need to

do to get there.

fu

what you plamed on doing.

Ceck to see i the obiective was achieved.

Act

on the inormation. Insofar as you were

successful,standrdize;where'you seeroom
tor improvement, recycle.

Tool Selector Chart

This a-rt organizes the tools by lyPical imProv-'s
situations,suahasworking with numbers,with ideas'c
in teams.

+
+
+

o
o, o

Working
withldeas

-9
o o

o o

AND

Atfinity '12 o
Brainstorming1 9

oa
a a
oo
a
ID 76 a O
Matrix 85
a

Radar 137

Tree

+
=

C & E/Fishbone
Flowchart 56
ForcField 63
Gantt o

9t
NGT/Multivoting
Prioritization
PDPC t6(

#
Workinqwith
Numbeis

CheckSheet 3 1

controld6E 36

tr4

a
ao
aa
a

oa
a
a

_ 1

F-.!

Run 411
a
Scatter 451a

ft

(!

Data Points 52
Histogram 66
Pareto
al
ProcessCapability132

!r-

E
o,
o)

Workingin Tams

I.
E

PDCA Tool Selector Chart

lmprovement
\ Storyboard

8
I

P
3

E
g

3
3

Tools
3

o
o

o
o

(9
(9

oo

o
o
o

(9

o o o o
o

o
o

o
oo

(9

tJ

Q)

Moe. inJormation and examples of application can be

rormd in the Coacls Cuide, a guide ior facilittors who
suppo teams in leaning to apply the tools effectively.

ooo

The alphabetical table of contents plus the four selector

charts.presentedon the previous pageshelp you to put
r-our dnger on any tool within seconds

aoacH'sutDE

o
o

a) a
\:,

(9

(9

MEIORY JOGGEBI'I
This guide is a convenient and quick on-the-iob rcference.

(9

ID

USINGTHEQUALITYTOOLS

r-<

o o o

Ld

o
oo

tsr4
d

r!,

Tl Carh's Guide Esls additional matedal such as softF'aIe for c?eating flowcharts, videotapes and CD_ROM
to a(ommodate individual leaming atyles.
P'IIUPS QUALITY ACTTON LEARNING (PQAL)
The tools.can only_be ap?lied effectively in a learning
organization.PQAL workshops support teamson their
road to Total Quality. These *ortihop" dre manageFlelt-led,team-basedand directly releiant to the wrk
process. Acfual work issues are dealt with and practical
solutions are reached.

What .lo the diferent positions

o the runners mean?
,gp

ft '

Getting Readg-At important fust step is to sele.t

the dght tool for the situation. When you see tlE
"getting ready"
position o the rumer, expecta bEf
descdption of the tool's puipose arld alist obenfts
in using the tool.
C^i"i"g-Vnr*you
seethis runner, expectto fud
constuctionguidelinesandinterpretationtips.Tlis
is the action phase that provides you with stePb
step instructions and helpful formulas.
Einishitg the Cotlse-When you see this runrq,
expect to see the tool in its final orm. The a
examples from rnaflractuing, adminisffi
sert)ice,a d daily lie to illustrate tl widesprd
applications o each tool.

=
=

L
F{

&
-

F4
F4

-al

,3\. ffii'*iff8iScheduling sequential

& smultaneoustasks
Jtf
e
it?
To allol- tam to 6nd both the most eficient path and
=alirk schedule for the completion of any project by
geFit
showing total completion time, ahenecea5- squence of tasks, those tasks that can be done
.irli.!'<lt
and the critical tasks to monitor.
H@iado?
. AII teammembers have a chanceto qive a realistic
piature o what their piece of the plan requires,
basd on real expedence
. Everyone seswhy he or sheis critical to the overall
sucassof the project
. Unrealistic implementation timetables re discoveed and adiugtedin the planning stdge
. The entie team can think creatively about how to
shorten tasks that ae bottlenecks
. The entire team can ocus its attention and scalce
resoutces on the huly critical tasks
^a,
't-

How do t do it? )\l

L

Assembtethe right team o people with firsthand

knowledge of the subtasks

2. Brainstorm or document all the tasks needed to

complete a proiect. Record on Post-ib"

3. Find the fist task drat must be done, and Dlace tlte
card on the exkeme let of a large work uace
Job/fask card

PostjtD Notes

Detemine
larget _
auolenceo
newtopic

+
+
+
+
+

6. NEmber each tagk, draw the connecting arrowa,

md atrEe on a realistic tinre or the compledon oi
eacl task
. Recordthecompletion time on the
bottom hal o
each cad-

E
Bk "Are there any tasks that can be done sinol
laneouely with task #1?"
. IJ there ae simultaneous tasks, Dlacethe task cad
above or blow ta$k #1. ff not, go to the noc s+Aslq nvhat i6 the next task that mtrst bG dotra?
Can othes b done simultaneously?"
. Repeattiis questioningproces rmril all ereccded
tasks are placed in sequenceand in parand

EEIEE
I'iffiJI

| |-:l l-=.1l-:--l l=l

l=r
t:t t:l

tGe*l n tf tl fl

--_-j.-----\

tl

lh-

F--

ts--

:l
TiItAt each step always asl "Have we orgm
any other needed tasks that could be dor.? sinrul
taneously?"

lr{

7pBe sure to agreeon the standard time unit for

eachtask,e.g.,days,weeks.Elapsedtimeiseasier
rnan -deqrcted. hme, e.g.,g hours ofdedicted
nme versus 8 hours over a two_week period
(elapsed time).
7, DeterDine the proiect,s citical path
. Any delayto a taskon thec t icot
Wfh willbe added
ro the projecs completion hme, unless another
task is acceleratedor eliminated. Likewise, the
pro]ecs completion time can be reduced by accel_
erating any task on the critical Dath.
. There.are two options for caliulatrng
the total
critical parh nd the tasksincluded wiihin it.
In ngest,.c
umulatuepaI h. ldentify total proiect
comptetion time. Add up eachpath ofconneted
ac0vrues. lhe tongest cumulative path is the

H
Acdvlty NttrrorGint-

quickest possible implementation tine. This

rt.e proieds critical path.
"slack" in the startCalculatedslack.Calcttlate the
each
task. This id.!times
o
ine and completion
ti-fi'eswhich iasks must be completed 5cdy
schedded (on the critical pafh)and those that har
some latitude.

-I
_-+-_
t!-l

-|l

Findingthe ctical path by calculatingihe slad(

3. Assess
compelitois
ofieings

ES = The /argsr EF ol any prvtals co.hscEd td

.9

WhnES= LSANDEF = tF, ulat lask is on tne ctical @r. sd

rfieo h6 is no schdulflribililvin lhis Eslc

o)

r-{

F..-

'd-

F{

6
o

EF = ES + lhe lim lo complsie lhd tt

LS = LF - rhe tim !o oompli lhal tast

!-!f

tr.{

LF - Th smai/sst LS ol any conndd eri!

tip Determining the longest cumulative path is

simpler than calculating the slack, birt can
qurckly becomeconfusing in lrger ANDS.
The calculated slck option determines the total
poject and slack times; and therefore the total
projecttime and critical path are identified ,,duto_
maticallv"

F{
--

q)

H,
=
=

The constructede\ampte shown in this sectionis in

tie
Actr\rty on N ode (AON )format. For more inormation
on
ottle toruts such asActivity on Arrow AOA) and lye<e_
onceDugrm ODlvD,sR Th" Mpnory loggerplus+6.
wid.elyused,schedule.monitoing
,ftltg
(lann durt. It is smpletool that Lrges metlod is rhe
horizontalbarcto
!I|or\ wtuchtaskscdnbedonesiftultaneouslyoverthelife
oi tne propct. lts primary disadvantage is that it cannot
snon wruch hsks ale specificallydepdellt on ea other.

=
o)

r=

(!

o)

'-

=3!;

* i=9:9E
i

'P

R9R}

, - E

iltr*

s-

.g

g Ft

8 5
P

tF

E
3

sl
8

Activty Ne4wotldcann

61ss1co-Fc

!_

.E
6
E

m*- 3e

trt-

;=E:
p

E=

6
o
o

:
6

=
F

Y I E - !

sg="
>Q

hfonation

99

prcvided couttesy ot BGp

I
Activity Network /
-1
Phase|
ISO 9000 Certification Audit Schedub

E-=t

Activity Ntwork z

-)
Phase I
FO qm Ctificalion Audit Schedule (conflnued)

E iBfl
Eg

-! ..

-.E
F 9

R
rE
: E

F +*
gE

S . i ! a :

=r

t l o
P :

E +
6:=*

E=
:

fi

=:
o*

h-

g
! i c

u*;;

;i5
9;E

!.9
f i 3i

l"o
I
t-t
@

H
{

E:
e9T;

gTE
2.tA.P

z\'

Affinity Diagram
Gathering&
groupingideas

''f"#*,::'iin::l:::il#",".

Or trh rE
t-t t- c]t
Ir] t-t tEl
tEl l=l lEt

J"::'
i::.',:*i#'Jil;':i:'J;Ll

LIJI

Why use it?

To allow a team to creatively generate a large nutrd
ideas / issues and then organize and summarize natrd
groupings among them to understand the esserE d r
problem and breakthrough solutions.
What doet it do?
. ncourages creativity by everyone on the bD t
all phases of the process
. Bral(sdown longstanding comunicatio lFir.
. Encourages non-traditional connections amq
ideas/issues
. Allows breakthroughs to emerge naturally,6E
on long-standing issues
. Encourages "ownership" of results that 'Eg
because the team creates both the detailed irq
and qeneral results
. Overcomes"teamparalysis,"whichbrougbc
by an overwhelming array o options ard ladd
cnsensus
;

tlow do t do it?
1. Phrasethe i6sue under discffision in a full sr'
What are the issuesinvolvedin planning
un amilyvacations?

Br.instorm at least 20
ides or issues

;;:r:i;l"j"J:.d,,*,r""i
i"."fl;1iH.1.,i#1.*,,1,
*'iir,;+i;

mmr_ ffi
ffi@Jffi
L"q
rtl'",iilii
?,f"'
I r n ep s s t t - - - _ _ -

D4
I

ftffi

10to40mor
lli:1'L'gy,3$I'i",r

it isnot
?*"y, fi j;j,fJ;e160items;

E
=

-4ry

f=

$tf :ii:,tJiL;,'+idessimurtaneou6rvinro
",
fr'. i,l?,f_,,:-:r_"."whererheyfir besrroryou;
that voutirink
;"il;'ir:l,;1nv.noes

grouping; record it on a Post-it- note and place it

at the top of each grouping. Tese are ilralt
header cards.
b) For each grouping, agree on a concise sentence
that combines the grouping's central idea and
what all of the sDecific Post-it" notes add to that
idea; record it nd replace the draft version. This
is a final header card.
c) Divide large groupings into subgroups asneeded
and create appropriate subheade$.
d)Draw the final Affinity Diagram connecting all
finalized header cards with their groupings.

b) Sorting will slow down oi stop when each peEon

feels suficiently comfortable with the groupingr

Whatae the issuesinvolvedln

planningfun amilyvacations?

t"kkrdJ
o ida8

t"""rdtl
I hobbies I

lLoofirvl'
I

pictures

"
l=FL'd
pnce1I
rangor,"qd
AIIMINS
l-com6;-l
r.ipI
I business
toert.t
lloralbudstl

F"-'r r..ql

Whatare the issuesinvolvedin

planningun familyvacations?

lf'par 1

lllustrotion Note: Thea 5 to 10 moregot+ritgs

o ideasin a typicalAffinity
Diagram
ttp Sot in silence to focus on the meaning h'lrip.r
and connections among all ideas, instead of @tions and "history" tha t often arise in discltssiG.
7ip As an idea is moved back and fortll try tDthe logical connection that the oth6 perd b
making. If thjs movement continues treJdld a rEsonable point, a$ee to ceate a duplicate Fl'ip It is okay for some notes to stand alole- Th
"loners" can be as important
as othes that fr grcuPings naturally.
For-each grouping, .rete gunrDarJr fi *
caros usrng con6ensus
a) Gain a quick team consensuson a wotd fioh
that captu:res the cental idealthc
d d

Afinity

F*tr--l F".l"r:T',xlJ-l
EmFl

I'Tig:,{3ll''I
Jo"li,,{ii,,[
I ;:gm I
u"""1 |
I tA"kilJ I ltFnt;o*il. I t croativ
|

oridas

|
|
| |Consider | |
I lvervon,s
| |
| | hobbies | |

| | anoeoforic| |
l altmalives
ll
|,..----=....._
|
uomone ll
|
| lvacaonw nl |

I I Lookat tamitv
| | | -ll
p i c t u r e-sl'll I l l j e t e r m i nIe l l
ll
l'
I ltotatbudget
I ll

| |
| |
I llravlagenllI
|r:---'---:
l
I Flnobcaltons 1
| wrh acnvfles i
|
l R c a l l g o o dI
vacalios
ot I

t - l

lllustrationNote:Thereare5 to 10groupings
of
ideasin a typicalAfinity.
Thisis a paiialfii;ity.

Afrntty t5

Tip spend the e\ t rd time neededto d o solid hd

iards. Striue to capt ue the essenceof a]l o tlr ia
in eachqrouping.Sftorlcutshdeca geatly fu
the effetiuenes of the Jinal Affinity DiogrtIt is Dossible that a note within a grol4rtE
could bcome a header card- l{oweve' dc:
"closestone" becauseit's conerila
choosethe
The hard work of creating new header cads ''lla
leads to breakthough ideas'
vartafions
Another popula form of this tool, called the lQ
Method, i as developed by the JaPdneseanthroft'L'
nist tiro fa*aLito wile he was doing fieldwok m tb
qSO..L" <J Vethod, rdentiied with Kwakita s Er;
tiais.helped he anthropoloBistand his studenBgadE
nJ unuiyt" dta Th K]-Method differs from de
finity iug."* d"t.ribed above in that the cards are
iact ba"ed aid go th-rougha highly 'tructured re&ment proce.s be"forethelinal diagrm is 'rted'

Affinity

:=

lssues Surroundingv)
fnplenentation o the Business Plan

g
I

Hffiffi

E*j

g
6
E.

;:+:l

ffiH

ft l:e.llF
. l
:

l:llgl
l;:
l-EE
I
td5 |El

L eill-

F--

--

EC
R3

ht

rr{
-

16

Ainity

!-.1

o1ee4GoauoPc

Alfinltv

Afinitv
-

t/

lssuesSurroundingv)
lmplementatlon
of the BusinessPlan(cont)

+
+
#

Bainstorming
Creating bigger
& better ideas

|l'
e
if?
To establish a common rethod fot a team to ceatively
and ffciently geneEte a high volume of ideas on any
topic by ceating a process that is free of criticism and
Fdgment.

ffi@ ++
#
HHw +
+
HEEH
E4bq
+

E;-_-lrJ

l;gEllEl

63

Not: The Afinity helped the team bring focrls to rh -t

opinions on business plaming. The headers dEr s''A.I|
became the key issues in th ID example (shown in |h lD d

ry

E-

lfhat dos it do?

. Encou:ragesopen thinking when a team is sfuck in
"same old way''
thinking
. Gets all team membes involved nd enthusiastic
so that a few people don't dominate the whole
grouP
. Allows team members to build on each other/s
qeativity while staying focused on their joint
tllJsslon

tlow do t do tt?

There are two major methods for brainstorming.

. Structured. A process in which each team merrber gives ideas in tum.
. Utlstructured. Aprocess in which team members
give ideas as they come to mind.
Bither method can be done silently or aloud.

Unatructured
llhe process is the same as in the structued method
except that ideas are given by everyone at any time.
There is no need to "pass" since ideas are not solicited
in rotation,

Structurcd

b
1. The cental brainstorming questi@
aeeed on, and rl'ritten dor'vn tor ev6tn
B; sure that everyoneunderstands-the
issue,or Ptoblem Checkthisby askrng(
"L"t t" p"tuPt t"se it befoe
flipchart or board
irb
2. Each team member, in tu'n, giYs a
Ever!
is
.riticized.
idea
WiLheachrotation around the team'any El.n Dassat any time. lllLile this rotation F
ull PrticiPation' it may rl'
;;;;;;;;
E
i"i*rt*o""n"i",V r ine)'erimced or shy
meirbers.
l+
* As ideas ae geneated,wite eachore'[
3.
rriti
"i"r'i r"tt" on a flipchart or other
s!face
tb 5*
Make sure erery idea is recordedwith
*tt J ir't" tp"u, don't interPret or aErFi
tt.,'s,the Peson rwiting sdd
;;:.;;."t
""f. the speaker_if the idea h5 be
,i*""t
wored accuratelY'
ea P:
4. Ideas are Seneted in lul ytif, (or
E o1"""", itti"utiog tttat the ideas
ire exhausted
ffKeep the pocessmoving and retativd-v
f ,'zo-.inut"" *otts ielt, dependirg o b
comPlex the toPic is'
'l"trl5. Review the written li6t o ideaofo
disard any duPlicates
idtllbl 13
Discard only ideas that are virtualty

Jr't i-poituot.
l" fl"::,1",,:**r#T
that re revealedin slightly ctre

20

Brainstormhg

Variationa
There are many ways to stimulate cleative team thinking. The common theme among all o them is the
stimulation of creativity by taking advantage o the
combined brain power of a team. Herc ar thr examPres:
. Visual brainstorming. Individuals (or the team)
produce a pictue of how they see a situation or
ProDtem,
. Analogies/ree-word association. Unusual connections are made by comparing the problem to
seemingly unrelated objects, creatures, or words.
For example: "I the problem was an animal, what
kind would it be?"
. 6-3-5 method. This powertul, silent method is
proposed by Helmut Schlicksupp in his book
CreatioitVWorkshop.Itis done as ollows:
a) Based on a single brainstorming issue, each
person on the team (usually 6 people) has 5
minutes to write down 3ideas on a sheetofpaper.

b) Each person then passes his o he! sheet of

paper to the next person/ who has 5 more
minutes to add 3 more ideas that build on the
filst 3 ideas.

=e
Bralntonnlng

L ) T h i 5 r o i l i o ni < r e p e l e da - m a n \ .h r . c -: there are team members, e,9., 6 team r.-bers = 6 roiations, 6 sheetsof paper, 1Ei,r1.
per sheet.

This interesting processforcesteam memt'as:,

contiously build on each olher'r peripe._r.+
and input.

r=
r=
r=

f=

Ir=
Ii=
f;=
Ii=

F=
-i
l-.

lsr.

Cause& EfecU
FishboneDiagam
Find& curecauses,
NOTsymptoms

Why use it?

To allow a team to identift explore, and graphicaly
display, in increasingdetail, afof the possiblcausei
related io a problem or condition to discover 1$ root
cause(s).
What does it do?
. Enables a team to focus on the content of the problem, not on the history of the problem or difering
personalinterestsof team nembers
. Createsa snapshot of the collective knowledge
anclconsensusof a team around a problem. This
builds support or the resulting soiutions
. Focusesthe team on causes,not symptoms

*-

How do , do it? ))
1. Selectthe most appopriate cause& eect ormat.
Thee are trwomajor ormats:
. Dispersion Analysis Type is conshrctedby placing individual causeswithin each ,,major,,cause
categoryandthen asking of eachindividualcause
"Whydoesthiscause(dispersion)happen?,,This
question is repeated for the next level of detail
until ihe team runs out o causes.The graphic
examplesshown in Sfep 3 of this tool sectionare
basedon this format.

:
22

Brainstorming

-,1

. Proce$ Classiication TJ?e usesthe major stel=

f the process in place of the maior cause catthe
Ieoii"i ih" toot.u'e questioninPro'essis
sme as the Dispersion Analysls ryPe'
&
Geneate the causesneeded to build caus
Efect Diagram. Chooseone methocl:
. Bainstorming without previous PreParaflon
. Chek sheets based on data collected by tem
membersbefore the meeting
Constuct the Caus & fect'/Fishbone Diagraur
the
a) Placethe Poblem statemenlin-a bo\ on
rishthand side of the wrting surtace'
. Ailow plenty o[ space L'e a tipchart sheet'
j
t.,t.n"i popt, ot luge white board A PaPer
s*ra." il pi"f"tt"a ti"c; the inalCause& Effect
Diagram can be movecl'

Causes

_l

b) Dr\,,r-mdjorcuse(ategoriesor steps in the

pro_
ouctlon o serviceprocess.Conneci lhem lo the
_backbone'
of the fishbone chart_

ffi! I

T
r=

+
+

+
fr
r=

Methods

\raterials

. Be flerible in the maior

cause ,,bones..fhat are
u9ed. tn d production process the trditional
(dtegorjes.are:Machines (equjpment
), Vethods
Inor^ work ts done),Materials (compuncnrs
or
people
rw mteflals).
nd
{thehumaneiement).
process
rn a.sevic
the Lrdditiondl
methodsare:
ror rcle6(hrgher-leveldecisionrule5t,procedures
rsrepsIn a tdst), plant (equipmentnd
5p(e),
and People_In both t,?es t pioce.ses,Environ_
menr tbuttctlngs,
logistics,
dnd .pace),and Mea_
surement (cdlibrdtionand datj collection)
are
arsoJrequen y used. mere s 1o peiect
set or
nuhber ol rategortes,Make themfit the problem.

ts4
F4
tit Make sure eveyone agree5on the Problem
asmuchirormahonasPos$ole
siatemenl.lnclude
".tt"t" " "when," and "how much"
i tn";;*rto,,"
i-tne probtem. Use data to sPecify the Problem'

PoPIe

lllustationNote:Ina procssClassi,jcation
type
romal,eptacethemaior"bone-categones
w n:
-Cookin,.
uroer |akrng..pepaation."
and'Detjvery"

E:=

"Bones
(Majo cause categories)

Machinerv/
Equipmet

H
H
F4

t-

c & FJFrshbo't

--

01

o6rc

l-.1

ru

C& gFishbone

25

Machinery/
Equipment

more basiccausessuch as "Inaccurateordedng,"

"Poor use of space,"and so on.

ll.-{

-!
Mateials
c) Place the brainstormed or data-basedcausesin
the appropdate category.
. In brainstorming,possiblecausescanbeplacedin
a major cause category as each is generated,or
ody after the entire list has been created.Either
works well but brainstormine the whole list first
maintains the creativeflow oldeaswithoutbeine
(onslrained by the mdior cause ctegoriesor
where the ideas fit in each "bone-"
. Some caussseemto fit in more than one category.
Ideally eachcauseshould be in only one category but
someofthe "pmple" causesmaylegitimatelybelongin
two places.Placethem in both categoriesand seehow
they work out in the end.
7ip Iideasareslowin coming,usethemajorcause
categoiiesascatalysts,e.g.,"Whatin'mateials' is
causing...?"
d)Ask repeatedly o each cause listed on the
"bones," either:
. "Why doesit happen?"For example,under "Run
out of ingedients" this question would lead to
26

c & gFishbone

H
E

fL-q
H

E-

o_

>t
H
4

3
I

RP

P'

r.4

E
'-

;e 5

UJ

VX

3
3

E _.

.g

,s,

E s
!'

:r

I
C & gFishbone

27

. "what could happen?"For example,under "1.-=

out of ingredients" this question would lead 1. :
deeper understanding o{ the problem su =
"Boxes," "Prepared dough," '"Toppings,"and so.-rip For each deeper cause, continue to push f.-t:
deeperunderstanding,but know when to stoF ,:
rule of thumb is to stop questioningwhen .au.
is controlled by more than one level of marlagement removed from the group. Otheri{is. 'i
processcould becomean exercisein frustlad!-:,
LIsecommon sense.
b) oneormore. :
e) Interpretorte<to rootcause(s)
the ollowing:
. Look or causesthat appear repeatedly R'ithin !-:
acrossmajor causecategories.
. Select through either an unstuctured consenL.'.!.
processor one that is structured,suchasNomini
Group Techniqueo Multivoting.
. Galher data through Check Sheetsor other tomats to determine the relative frequenciesof fne
diferent causes.
Variation.
Traditionally, Cause& EffectDiagrams have beenc
"fishbone- :..
ated in a meeting setting.The completed
often reviewed by others and/or conimed wiih dE
collection.A very efective altemative is CEDAC! :
which a large, highly visible, blank fishbone chan :.
displayed prominently in a work area.veryonepr:':=
both potential causesand solutionson Post-it^ not 1.
each of the categodes.Causesand solutions ar :viewed, tested, and posted. This technique open-.the processto the knowledge and creativity of ereperson in the operation.

28

C & Fishbone

+
E
E
+

Cause & ElectlFishbone f

.")
Bed Assignment Detay

d=

I()

.9
7

E
E

.r= i
!

Ega

ZE

(J

!
o-E

c:

+
+',

--

:g

+
+

+,
+

5g

.4

gg

P 3

s E3
q
E

(!

i ;

E
p

q6

FE

tFtonatoa prcvded counes, ol

Bush-prctbyletian-S!.t ukes Medicatc;;rc1

C & gFishbone

29

Cause& EftecuFishboneJ
Causesor Bent Pins(PlugFlnSkb) '

-:
I_

f=

s
;

E=

l-t
=

E-

lh-al

e
a

F4

.9

E ^

(D

F-a
5

-c

r.9

AT&T

30

l-4

C & /Flshbon

.^\' :H,f:"" ffi

accumuiatinodata F+l r L!

Why us it?
To allow a team to systematicallyrecord and compile
data fuom histoical sources, or observations as they
happen,so thdt patternsand trends can be clearly
detectedand shown.
What does it do?
. Creates easy-to-understand data that come from
a simple, efficient process that can be applied to
any Key Pertorrnance areas
. Builds,with eachobservation,a dererpictureof "the
acts" as opposed to the opinions f each team
memoet
. Forces agreement on the definition of each condihon or event (every person has to be looking for
and recording the samething)
. Makes patterns in the data becomeobvious quickly

,tow do t do it?

*
-ta-

1. Agree on lhe definition o the events or conditions

being observed
o I you arebuilding a listofevents or conditionsas
the observations ae made, agiee on the overall
definition of the project.
Exmple: If you are looking for reasons for late
payments,agreeon the definition of "late."

II

. Ifyou are working rom a standard list of events

or conditions, make sure that there is agreement
on the meaning and application of each one.
Example: If you are tracking sales calls from
various regions, make sure everyone knows
which statesare in each region.
2. Decide who will collect the data; over rhat periodi
and ro[i what souces
. Who collectsthe data obviously depends on the
project and resources.The data collectionperiod
can ranse from hours to months. The data can
come frm either a sample or an entire popula. Make surethe data collector(s)haveboththetime
and knowledge thev need to collect accrn2ti:lrorrnauon.
. Collect the data over a suJficieni frii io i'e i-::.
the data represents "i]?icai' rulis ,lull:: e
"typical" cycle for your brrsiness.
. Sometimes there mav be imDortant diitsences
within a population ihat should be reilecied b
sampling each different subgoup indiiidualivThis is called stratification.
ExamDle:Collect comDlaint data r:om business
havelrs separately from other types of travelers.
Collectscrapdata from eachmachineseparately.
Iip tt must be safe to record and repo* "bad
news," otherwise the data will be iltered.

+
+
E=
t4
f=
4

t-4
-4
-{

i.

Design a Check Sheetorm lhat is clear,complele,

ancleasy to use
. -L complete Check Sheet,illustrated below, in.ludes the following:
Sorce Informalion
EName of project
ELocation of data collection
o personrecording data, i it applies
!Nane
@Date{s)
E Other important identifie$
Content Information
E Column with defect/event name
p Columns with collection days/dates
E Totals for eachcolumn
I Totals for eachrow
UGrand total for both the columns and rows
P.orec1.Adm s on De avs
@lion: ErercncvBoon

E """*n,

! -",

3r'11 3/12 3/13

ts--

a4

33

2A

36

30

25

r=

E -",

ShrftAll

3
5

12
I

47

'12

52

31

24

38

L ,",

b4
32

Check Shet

olee4GouoFc

-{

Chcksheet

33

Collect the data aonsistenily and acculately

. Make sure all entries are written clearly.
|iD Managersand/or teammemberscdndo their
".t to hefp th" dot" .ollecto(s) do their job well
by simply ihowing an interest in the Proiect Ask
ie.ollector(sthow the pojecl is working out
Show your support-tell the data collector(s)you
thinl it is imbrtant to collect the information
Abooeall-a onthe data as quickly aspossible!

va'iationa
Defect Location
Shows the concentration o defects by making a dra\r 'the
product each time a defect is fouJld.
ing of
irem: cltdEttr
[ttl: 11

Prolci: Monrc'c6rnq damas

LdaoAssemblv

slr' =

+
+

CheckSheet f
Kevboard ErrorJ
in ClssAssignment

+
+

(!

tr

+
+

r-*.t
H

Hl

---:

,t'1

' h .J )

f
GMic

@nP'tq

nonlol

Tsk Checklist
Tasksin producinga produclor deliverinEservicedre
lhis
checkedffas theyaie done.IncomplexProcesses
"mistake-Proofing."
is a form of

g
E-

Chd( Sheef

r-

E E E g F
.- Q , * E H -. t i- lo 0P g< '

. u =F ;

>

E-

a
l--

34

o ' o ) c . c ! h q a
c c o : :
o ft

Inbrnationprovidedcounesyot MillueekTownshipSchoolDistrict,
M [creek Town* ip, Pennsylvania

-r\

--ucL
Control Charts
f-t. ^
\a'^
Recognizine
tr
---LcL
sOUrCes
af variation l

Why use it?

To monitor,control,andimproveprocess
pormance
overtimeby studfng variationand its source.

+
#

+
a+t

What does it do?

. Focuses attention on detecting and monitcn::
process vdriation over time
. Di(tin8uishes special fom cornmon cauis ..i
vadation, as a guide to local or maiagemeEi
actl0n
. Serves as a tool for ongoing control of a proces-<
. Helps inprove a process to pefform constendt
and pedictably for higher quality, lower co.i.
and higher effective capacity
. Provides d common Ianguge for discussing process performance

*
,low do t do it? ))
There are many types of Control Charts. The Control
Chart(s) that your team decides to use will be determined by the type of data you have. Use the Tree
Diagram on the next page to determine which Contro.
Chart(s) will best fit your situation. Other t'?es of
Control Cha*s, which are beyond the scope of th
book, include the Pre-Control Chart, the Mor-ing
Average & Range Chat, the Cumulative Sum Chart,
and Box Plots.

Based on the typ o data and sample size you have,

choosethe appropdate Control Cha.

measurcd& ploltdon
e.9.,tim,tnperature,

*4

E=
=

* Defect = Failureto meet one of

the acceplanc crileria.
Adective unit might have multipledeecls.

*" Defeciive=An entire unitails

to meel acceotance
crileria,regardiessof the number of dects on the unit.

E@E

-4

ContolChas

37

t
-rt-:-

Constructing

Controt

Charts

1. Selectlhe pocessto be charted

2. Determine samPlint method nd PLn
. How large sample can be drah-r|l B.a'r time and-cost to cUecta sample trith tix a-
oi intbrmation you will gatir' Ser tlc ra f,li+
gnm on lhe priuious pagefor sugestc/ fi
. As much as possible, obtain the sarPls mds
the same technical conditions: the samernadic
operator, lot, and so on
. freouencv of sampling will depend on rr-helha
vou'are able to discem-patternsin the dta C.*
annuIlsider hourty, daily, sliifts, monthly,
"in control-'
lots. and sobn. Once the process is
you might consider reducing the frequenct rrr
whrch vou sample.
. Cenerally,colleit 20-25gouPs of samPlesHEre
clcultins the statisticsand control limits.
. Consider using historical data to set a base*i
7p Make sure samples are random- To e*blit
the inheent vaiahon of a process, allow tlprocess to run untouched, i.e., accolding to star
dard procedures
3. Initiate data collection
. Run the process untouched, and gather samPled
oata.
. Record dala on an appropiate Contol Chrt
sheetor other Eraph Pirr. Include any unusual
eventsthat occul.

ffi

Crlq at the appropriate statistics

: -11ou hare attribute data,use the Attribute Data
:a:1, Cental Line column.

:E

F
F
#
+
+

Attribute Data abte

Cqd

tt

-*r
-

Shple

\nbb,
@dly

F = npin
.*
_

csslanl,
usuay

"ucLo=-p+31/Eil:-F
*rcr"=p-o1fili-JJ
u c l - D= n + 3 V n p ( 1 + _ Lcr-e= nF 3y'np(l -p)

ucq=.+3V-

t *.s
a>5

Lc[=-3!G'ucL,,=!*3,F

ll\
:

=
- - 9:1.1"
..
:-"j::'*" sr4 wtrm
n: - empr
x- ' orslDgouos

'LcL,=a-31,F
rhis tomu.a ccares cr.ansirsconro ,m,rs.
ro voidrh.usdveragesampresizesn ror
r4osesmps lhal ac;{1n 4oo! ot rhe
av@Aesmpt siz. C,cutatindvduat
ttrrls tortho sampteserceFoingl2ooo

a
a
t-af

Conkol Charts

b)If you have vatiable data, use the Variable Data

Table,Central Line column.
Variable
TYpeContro
chat

Sample

& Range

< 1 0 ,b u t

Data abl
CentalLlne'

control Limlts

i = li,+i.+...i-t UCI'= X + &R

k

3lo5

LCli=X-ArR
UCLF = DlR

LCh = D3F

t = (&+&+...k) UCL'= X +&s

Slandat

k
>10

UCIT = B1S
LCL" = B3S

& Rang < 1 0 ,b u i

XandB

x = (xr+&+...xk) U C L ; = i + A f
LCLi=X-A,F

ucLu = D1R
LCLR= D.F

3 or5

& rvloving
Rang

t = (xi+x,+...xk)U C L I = X + E r R m
LCL' = X

Rn= (Xr,r Xr)l

i.=(Rr+R,+...RH

E,F,

LCl4, = D3B.

E+

+
.+
.

fr I
H

E=l

k = # o subsroups,I = medianvaluwilhinachsubgoup

F-l

^=1

P'

40

controlcharis

,o1rr4
coaLoPc

i.

Calcolate the control limits

: t: ! rru hal'e attribute data, use the Attribute Data
lable, Control Limits column.
b Ii vou have variable data, use the Variable Data
fable, Control Limits column for the correct for_
' Lula to use.
. L- the Table ofConstants to match the numeric
ralues to the constantsin the fomulas shown in
the Control Limits column of the Variable Data
Table. The values you wiD need to look up will
dpndon the typeot VariableControl
Chairyou
oo<e and on the size of the sampleyou h'ave
aharrn.
|ip I the Lower Control Limit (LCL) of an At_
fibute Data Control Chalt is a negativenumber,
st the LCL to zero.
7ip fhe p and u formi:las create changing contol
Iimits if the sample sizes vary subguf to sub_
group_To avoid this, usethe averagesamplesize,
n, for those samplesthat are within j20% of the
averagesample size. Calculateindividual limits
for the samplesexceedi]f.!20Ea.

6. Construct the Conhol Chait(s)

. For AtLibuteDataControlCharts,construct
one
chdrt, plotting ea(h subgroup,sproportion or
numOercletechve,
numbero defects,Or defects
Per unit.
. For VaidbleDlaConlrol Chart\,construct
two
charls:on the top chart plot each subgroup,s
mean,medin,or individuals,and on thebottm
chart plot each subgroup'\ range or standard

-E -4
ControlCharts 41

:-

able o Constants

D3

1.880

19*

1.023

2.574

a.729

2,282 1 , 6 2 4

0.577

0.483

84

3.267 1 9 ] 9
0

2.W

2266

2_114 1.427

2.OO4 1 . 2 8 7

oo
: :':

8A:

r:c

2 E3
;-:
--g

=:
=4

0 . 4 1 9 o 076

1.924

1142

0136

I 864

r a-ag

: '':

-:-:

-f-

':-

:=e

_- i

::-=

::

- -a

a=-

0.373

0.337 0 1 8 4
10

B3

D4

3.267 2.659

r-{

I and s chad

X and B charl

1.8i6

0.308 4.223 i :it

b-,r

fr
+,

t4

z
H

4
2
3

1.147

0.691

D.

3.67

2_659

2.471

2.242

1.772
'|.457

2.242 2_059

1.290

2 , 1 1 4 2.326

0.509 0.076 1.924

10

2.004 , . 1 8 4

D.

D.

D.

1.124

2.571

r.693

2.004 2,534

1 . 1 0 9 0.076 1.924 2.704

0 . 1 3 6 1_864 1 . 0 5 4

0 . 1 3 6 1 . 8 6 4 2.447

4.412 0.144

1.816 1.010

0 . 1 4 4 1 . 8 1 6 2.970

o.223

0.975

o223

3.074

H
=-

E=

* tlseulin esumarng the procsssstandaddevialion6'

42

control chals

olss4GoauoPc

bd

. Draw a solid hodzontal line on each chart. This

line conesponds to the processaverage.
. Draw dashedlines or the upper and lowet con_
trol limits.
tnteryrcting
Control Charts
. Attribute Data Conrol Charts are basedon one
chart- The charts or fraction or number defective.
numberof defect\,or numberot defectsDerunit.
medsurevriationbpf&,epr?
sar p/e<,Variable Dat
Control Chrts are basedon two (hrts: the one
on top, for averages,medians, and individuals,
measures variation ,etueefi subgroups ove ti\e;
the chart below, for ranges and standard deviations, measures vafialio ?tthrl subgfoups over
time.
. Detemine if the process mean (center line) is
where it should e relative to your customer
specifications
or your internlbu;inessneedsor
objecLives.
l not, then it is an indication that
somefhing has changed in the process, or the
customet requirements or objectives have
cnangect,
. Analyze the data relative to the control limits;
distinguishing between common catses and special causes.The fluctuation of the points witin
the limits results liom variation inierent in the
process.This vdriation results from common
cduse\within the system,e.g.,design,choiceof
machrne,preventjvemaintenance,
nd can onlv
be.aected.by
clangjng that system.Howeve,
pornrsour5rcteot the timits or patterns within the
limits, come lrom a special iause, e.g., hurnan
errors, unptanned events,freak occurrences,that
is not part of the wdy the processnormally
opertes, or is present becauseo an ullikelv
combinationof tro(ess steps.Specialcausesmut
Controlcharts

43

be eliminated before the Control Chart can be

used as a monitoring tool. Once this is done, the
processwill be "in control" and samplescan be
iaken at regular intervals to make srire that the
process desn't fundamentally change. See
7'Determininqif Your Processis ut
of Cntrol."
. Your p.ocessis in "statistical control" if the proces. ic not being aecLedby special cau.es, the
inlluenceo an individiual or machine.AII the
points must fali within the control limits and they
must be randomly dispersed about the average
line for an in-conol s';stem.
tip "Control" doesn't necessadly mean that the
product or seryice will meet your needs. It orily
means that the processis co11sfer7f,
Don't contse
conhol limits with specfication limits-specification Iimits are related to customer requfuements,
not processvariation.
|t:p Any points outside the control limits, once
identiied with a cause (or causes),should be
removed and the calculationsand chartsredone.
Pointswilhin the controllimit5,but showins indicationsof trend5,shrrts,o instability,dre;lso
specialcauses.
ip When a Control Chart has been initiated and
all specialcausesremoved, continue to plot new
data on a new cha, brrt DO NOTrecalculate the
control limits. As long as the process does not
change,the limits should not be changed.Control
limits should be recalculatedonly when a permanent, desired changehasoccurredin the process,
and only using data ael the changeoccurred.

44

control charts

orrr4 GoaLroPc

7D Nothing will changeiustbe(duseyou

chdrted
Ir. you need lo do gomething,l-t,rm a ream
to
,.Common
rnve.hgate.
See
eue\tionsjor lnvesti.
.
gdfing an ULrt_ol_Controlprocess
Dete''D,ining il your process is
"Out ol
Control"
A froces<is sdid to be , ouLof control,,il either
one o
lhese rs tru:
1. One or more points fall outside of the contrcl
limits
2. Wlen the Conlrol Chart is divided into
zones,s
shor4/n
betow,any of the ollowingpointsretrue:
ZoneA
ZoneB

UpperControl
Umit
(UcL)

9!99

n"er"se

ZoneA

Lower Control Limil

(LCL)
) lrvo poin ts,ou t o t l.Ueer onsecuh ve poin tr,
are on
tne samesrdeol the avedgein ZoneA or beyond.
,.
D' rour pornts,outof ftve consecutivepointr,
dre on
rnesamesrdeof L]teaveragein Zone B or beyond.
c) Nine consecutive
poinLsare on one side ol lhe
average.
o,
consecuhvepoints.incredsinsor
Ifj:-:,.:.,'.
oecreasjng.
e) I herere lourLeen
con<ecutive
porntsthatdlter.
nateup and down,
0 Thereareitteenronsecutive
pointswithin Zone
L (aboveand below the averget.

3
z

;
E

F
t

Yes d No

Yes

No

Yes f

No

Yes I

No

Yes I

No

Yes f

No

Yes C No

Yes f

Yes O No

a
o
,q)

o
e

46

Common Ouestions for tnvestigating

an Out-oi.Controt
Process

ContolohaB

No

+
+
+

f, Yes 0 No
O Yes D No

3 Yes al No

Are there differences in the measurement accuracy of inskuments/

methods used?
Are there differencesin the methods
used by diferent personnel?
Is the processafected by the environment, e.g./ temperature humidity?
Has there been a significant change
in the envionmentt
Is the processaffectedby predictable
conditions?Example:tool wear.
Were any untrained personnel involved in the process at the time?
Has therebeenachangein the source
for input to the process?Example:
raw matedals, information.
Is the process afected by employee
atigue?
Has there been a changein policies
or procedures?Example: maintenanceprocedures.
Is the process adrusted frequently?
Did the samples come from different
parts o the process?Shits? Individuals?
Are employees afraid to report "bad
news"?

A teamshould addresseach"Yes" answerasa potential

sourceo a specialcause.
ConaolCharrs 47

p Chart

fndividuafs& Movingf
RangeChart 1

GeneralDentistry:Perccntov)
Patients Who Failed to Keep Appointments

lV Lines Connection Time

PocesJop.atlo:VLesCon_eclor
lDepanmnl:
OnHanadm'ss'ns
I lnlenr,veC
FEquncy:l
StmPlSl:
y: tul:
Characterbllc
ls.nPl
'ne
in scondsI
On
|
chotint I Ew 6/10
Indlvlduals:k=26 2x=u7o
f;=325.n ucL=645 LCL=
n h d sn: = 2 t F = 2 9 c o R = 1 1 9 6 U C L - 3 s 2 l c l -

F=39

Historical Statistics:
U C L= 4 7
LCL= 31

_9_

630

\/\0,
270.

l/

140
90

424
360
300
140

A/

I-V

'-

*!/W-Y

tnfomation prcvided @udesy of U.S. Navy,

Navat Dental Centt, san Diogo
LCL

lnfornationprcridd counesyof
Pa*view EpiscopalMedicalCenter
Note Somethingin the procsschanged,and now it takesless
time to make IV connectionsfor patints being admitted for
open heart sugry.
e{q"oLoP
48
contolcharts

NotrProviding

nex dme for patients resulred in fewer appoinr-

u Chart

--1,
Shop ProcessCheck /
Solder Deects

II

ail
-).)-------LJCL

+
T

:r
T

aRcnart
Ovrall Course Evaluations

--L

r'
/

n = 10 evloalionsrandomlysampledeachweek
1-Notatall 2-Nolvery 3 Modeatey 4-very s-EnmeLy

E
#
+

: 9 : : : 9 =

! ! E E E E E
Checks(3 per day)
lnlomation prcvided coudesv o AT&r

+
+
+

a
ffi
50

Controlcharts

2.0-----Lg-

1.2

lnfamatianprcrided courtesyo Hamiltonstandard

Noter Weeks 1, 10 (from bottom chait), 16, and 22 should be
reviewdto undrstandwhy th ratingsar outsidethe control
limits.

ConaolChats

5'l

,,/.

DataPoints "I.l'",
-.*
turnlng oata

=:l

lnn lnfofmaLlon

E=

,,r,
3o7

What Type o Oata Oo You Have?

. Words?
. Numbers?
- Atttibute data?Attribute data can be counted
and plotted as discreteevents.It includes the
count of the numbersor percentagesofgoodor
bad, right or wrong, passor fail, yes or no.
Exmple:Number ofcorrect answeson a test,
number of mistakes per t'?ed page, percent
defectiveproduct per shift.
- Variabledata?Variable data can be measured
and plotted on a continuous scale.
Example: Length, time, volume, weight.

meningfulsubcdtegories
or classiications.
nd
from ttuspoint you canfocusyourproblemsoLving.
Example: Data often comesfrom lranv sources
but is treatedasjlcomingfrom one. Daton minor
rnjunes tor a plant my be recordedas a single
figurq but that number is actuallv the sum totaiof
injuriesby I ) t)?e (cuts,bums,sc;apes),21
location
eyec, hands, eet), and 3) department (mainten9:tce,
+pping, productionr.Bowis an example
ohow dataha( beenstratitiedby plantdepartrnef.

ra
r

f=
f=

5=

r=

Do You Ned to Cotlect Data?

. If you need to know the pe ormance of an entire
population, the more economicaland less time
consuming method is to dra\.a/a sample from a
population. With a sample, you can make inferences about, or predict, the performance of a

population. Basicsampling methods:

- Radotfi. Each and every observation o data
measurehas an equally likely chanceof being H
selected.Use a random number table or random number generatorto <eleclthe samples. --_
-a
spoucntfu!.lverv
nlh <amDlers<ele(ted.
Strutified.A sample is taken from stratified
data groups.
=

E=

EE@

Can You Catego vqr Data Inao s{l//ogrt,ups?

. Whenyou stratifydata,you breakit down into

:4

s 35
30
?5
20

Dpl.A andB
Combind

E 150
0
J

F 1 1 , 4i A
/ J

(appears airlv stabre overtime.,

.

35
30
25
20
10
5
0
J

r u r v . t

(DepL A is rLnning highr and

may be,ncreasing ove. tim.)

What Paltems are ,mportant in your Data?

Prediclable
ptternsor distributionscanbedeccribed
wlth statistics.

. Measures of location
Merr oravea8e).
Represenred
by I (or X-br),
the meanis the 5umo the vdluesof thesmple
(Xr, Xz, Xs . . . XJ divided by rhe torat nurne
(n) of sampled data.
ExamplefFor the samplet (3, 5, 4, 7, S)
(3+5+4+7+5)
= 4.8
DataPointE

53

- Medran.When sampled data arerank ordered

lowest to highest, the median is the middle
number.
Example: For the sample: (3, 5, 4, 7, 5)
Median o (3, 4, 5, 5, 7) = 5
When therc are an even number of values, the
median is the average o the middle two values.
Example: For the sample: (2, 5,7, 4, 5, 3)
Medir. o (2, 3, 4, 5, 5, 7) = 4.5
- Mod?. The most frequently occurring valu{s
in a sample.
Example: For the sample: i, 5, {. :. i
Mode = 5
. Measures o variation
- Ran8e.Represntedbr R. le rar-= s =r:ference betwen the highst dala r=-re X{E
and the lowest data value (Xdt,
Example: For the sample: (3. j. {, ;, i'

- StandardDaiation. Represented by 5, the standard deviation of a sample measuresthe variation of the data around the mean. The less
vadation there is o the data values about the
mear X, the closer swill be to zero (0).

A
.+

+
i+

Example:For the sample.(3,5,4,7,5, X=4.8

" = 1 ] t - i 3 - 4 8 ) " ( 5 - 4 s f. 4 - 4 3 f * ( ?- 4 . 8 ) , .{ s- 4 . 8 ) , I
\
5r
[ 3 . 2 4+ . 0 4 + . 6 4 + 4 . 8 4+ . 0 4 ]

=1E;
rl 4

="r'",
= 1.48
The squareo the standarddeviatiorL s, is referredto as
he osrance. Variar.ce is not discussed in this book.

#
+
+
+

54

Data Points

cllss4couoPc

Data Points

55

,,i'

Flowchart
l'tctultng

theProcess
Why use it?
To allow a team to identiy the actual flow or sequence
ol evenls in d pocess lhat any Product or service
follows. FlowchrtscanbeapPliedtoanythingrom the
havels of an invoice o the flow of materials, to the stePs
in making a saleor servicing a product.
What does it do?
. Shows unexpected comPlexity, Problem areas,
edundancy, unnecessarylooPs,and where simpliication and standardization may be Possible
. Compares and contrasts the achlal versus the
ideaiflow of a process to identify improvement
oPPOrtunities
. Allows a team to come to agreement on the stePs
of the process and to examine which activities
may imPact the ProcessPeltormance
. Identiies locations 'here additional data can be
'
collected and investigated
. Serves as a tnining aid to unde$tand the comDlete Dlocess
How do t do it?
1f Detennine lhe rame or boundaieof the Plocess
. Clearly define whete the Plocess under study
sta*s (input) and ends (final outPut)
. Team members should agree to the level o detail
they must show on the Flowchart to clearly understand the process and identify Problem areas.

l@il

. The Flowchart can be a simple macro_flowchaft

showing only sufficient information to undetstdnd the general process flow o it might be
detailedto showeverylinitectionnd de"(ision
point. The team might start out with a macro_
flowchart and thendd in detail later or only
t(/hereit is needed.

H
F

E
+
+

LGa

-t

2. Detemine the steps in the prccess

. Brainstorm a list o all major activities, inputs,
outpufs,and decijion<on a flipchrtsheetrom
the beginningothe processto the end.
3. Sequencethe steps
. Arrange the steps in the order they are
caried
out. Use Post-it' noles 50 you can move them
dround.Don't draw in the jrrows yet.
7ip Unless you are flowchafting a new process,
sequencewhat,s, notwhatsroald re orte idal.
This may be difficult at first but is necessaryto see
where fhe probable causesof the problem are in
me Process.
4. Diaw the Flowchart u6ing the appropriate s]rmbols
An oval is used to show the materials,informa_
tion or acLion(inputs)to stat the processor to
show the results at the end (oulput) of the
Process,

()
-

A box or rectangle is used to show a fask or

actvity perfomed in the prccess.Although mul_
tlple a ows may come into each bo>; usually
omy one output or aow leaveseachactivitv box.

A didmond shows those Points in the Process

where a ves/no question is beirg asked or a
decision i require.

A circle with either a letter or a number identiies

; break in the Flowchart and is continued elswhere on the same Pageor another Page
Arrows show the diection or flow of the Process'

. Keep the Flowchat simPle using the basic symbol;tsted above. As your exPefience gows, use
the
oth"t, -ot" graphic- symbois to rePrsent
jnclude:
rt"p.. ott"t "Iy-ibot" "mehmesused
- A half or tom sheeto paper for a report cornpleted and/or filed
- A can or computet taPe wheel for data entry
into a comPuter database
- A ld ge "D" o hal circleto identify Plcesin the
thereis dely orwait fo furthe
proce"sswhere
. Be consistent in the level o detail shown'
- A macrolevel flowchart will show key action
stepsbut no decisionboxes'
- An intermediate-level flowchat will show action and decisionPoints.
- A microlevel flowcha will show minute detail
. Label each Process step using words that ae
understandable to eveYone
. Add arrows to show the direction o the flow of
you
iirs itt ttt" pto."s. Atthough not a ule' i "no"
"yei" choicesbranchingdown and
shw att
choices brnching to the left, it is easier to ollow
ffi
58

Flowchart

+
:r
:r

#
+
+

+
F

the process.Preferencegand spacewill later dictate direction.

Don't forget to identify you work. Include the
title of your process,the date the diagram was
made. and the namesof the team members,
Test the Flowchart or completeness
. Are the symbolsused corectly?
. Are the processsteps (inputs, outputs, actions,
decisions,waits/delays) identified clearly?
. Make sure every feedback loop is closed, i.e.,
every path takes you either back to or ahead to
another step.
. Check that every continuation point has a corresponding point elsewherein the Flowchart or on
another page of the Flowcha*.
. There is usually only one output arow out ol an
activiwbox.lf thee is more than one arow, you
may need a decisiondiamond.
. Validaie the Flowchart with people who are not
on the team and who carry out the processactions. Highlitht additions or deletions they recommend. Bring theseback to the team to discuss
and incorporate into the final Flowchart.
Finalize the Flowcha
. Is this processbeing run the way it should be?
. Are people foltowing the processas charted?
. Are there ob\.ious complexitiesor redundancies
that can be reduced or eliminated?
. How different is the current processfromanideal
one?Draw an ideal Flowchart. Comparethe two
tcurrenl versusidel) to identify discrePn(ies
and opportunities for improvements.
Flowcha

59

Vafiations
reis sometimes
Thetype
_toof Flowchartiust described

ferre as a "detailed;' flowchart becauseit includes,

decisionPoints,and
in detdil, the inpuls, activjLie5,
output( o anv p.oce.s.Four other ormt, described
below, are also useful.
Macro Flowcha
Reerto the third bulleted item in Step 1 of this section
or d description. For a Erdphic e\amPle, seeSteP2 of
the Impror mentStoryboardin the troblem-Solving/
ProcessImprovement Model section.

Deployment Flowcha
This chart shows the people or deParlmentsresPonsible and the flow o the processstepsor tasksthey are
assiqned.It is useful to cladfy roles and track accountability as welt as to indicate dePendenciesin the sequenceof events.
Chris

l
1.2 Dcide on
guesl lisl

60

Flowchart

2 . 1D e c i d e
theme

2.2 Slcl
locallon

3.1 compLl

Lauren

ls there
timeto do
graphics?

Top-down Flowchalt
This chart is a picture o the major stePs in a work
process.It minimizes the detail to focus only on those
;i.epsessentialto the process.lt Llcuallydoes not in'
reworl dnd olher5lePcthatresultin
clue inspection,
quality problems.Tems sometimesstudv the toPdown'flwchartto look tor wd) 5 to simplify o reduce
the number of stepsto make the Processmore fficient
and effective.

1 . 1D e c i d eo n
budget

Karin

Worklow Flowchart
This type o chart is used to show the low of PeoPle,
materib,paperwork,etc.,within a worksetting When
redundaniiei, duplications, and unnecessary complexities areidentiiiedin a Path,PeoPlecantakeaction
io reduce or eliminate theseProblems.

3.2 Send

olee4coal/oPc

Flowchart

Ftowchan 1(

Prooosed Patient /
Appoiniment Procedure

at
lnftmation PrcviddcoutrsY
u s lhvv NavatDentalCenter'SanAgo
62

Flowchat

orse4coauoPc

E
#
#

'\' 1"":J'5'o
Positives &
negatives o change
Why use ia?
To identify the forces and factors in place that suPPort
or work against the solution of an issue or problem so
that the positives can be reinforced and/or the negatives eliminated or reduced.
What dos it do?
. Presentsthe "positives" and "negatives"ofasituation so they are easily compared
. Forcespeople to think together about all the asPects
of maling the desired change a Pemanent one
. Encourages people to agree about the relative
priodty o factorson eachside ofthe "balanceshee'
. Encourages honest ref'lection on the real underlying roots of a problem and its solution

llow do I do it? 4

1. Draw a lage letter "i" on a lipchat

a) At the top of the T, write the issue or problem that
you plan to analyze.
. To the far dght of the top of the T, wdte a description of the ideal situation you would like to
achieve.
b) Brainstorm the forces that are driving you towards the ideal sifuatlon. These forces may be
intemal or external. List them on the left side.

c) Bainstorm the forces that arc restraining movement toward the ideal state.List them on the right
side.
2. Prioritize the driving foces that cn be strentthened or identiy restraining focs lhat would
allowthemostmovementtoward the idealstate if
lhey were removed
. Achieve consensus through discussion or by using ranking methods such as Nominal GrouP
Techniqueand Multivoting.
7ip When choosing a target or change, remember
that simply pushing the Positive factors for a
change can have the oPPositeeffect. It is often
more helpful to remove ba(iers. This tends to
break the "change bottleneck" rather than just
pushing on all the good reasonsto change.

T=

FoceField

-t
:=

r:=
rt=
=

r=
r=

e
6>

E
, e :

E E

! t-

e E

: i ' 9 8 1 E!+i
f i : 9 , = ; ; E

:.:

!EE

-9

g i>

eiecee!
t r t t t t t t t l
9": A
h o;

iiiiititi

++t+++++
++

EHA

: 9pE
q c

iq "E E;c$ E
?u;
6

'

ts4

. : Y

E=

Forc Fleld

r-i

rr-

64

Fear of Public Speaking /

9
i
E

B
t

- 3 t

GoaUoPc
@1rr4

9 x

E E E

z Force Fild

65

lo

Histogram

|-l

rrocess cente ng. | )

l-1
spread, and shape lrl | | l-i

Why uso it?

To summarize data rom a process tht has been collectedoveraperiod ofhme,;nd gphically presentits
frequenry diitr;bution in bar for.'
What does it do?
. Displays large amounts of data that are dificult to
interpret in tbular form
. Shows the relative frequency o occurrence o the
various data values
. Reveals the centering, variation, and shape of the
data
. Illustntes quickly the underlying distribution of
the data
. frovides useful inormation for predicting future
pefiormance ot the Process
. Helps to indicate if there has been a change in the
Process
. Helps answer the question "ls the processcapable
oi meehng my customer requirements?"

*
How do , .to ia? ))
1. Decide on the poaessmeaeute
. The dat should be variable data, i.e., measured
on a continuousscdle,Fore\ample: temperafure,
time, dimensions,weight, speeA.

Gather data
. Collectat least50to 100data points i you plan on
looking for pattehs and calculating the
distribution's centeing (mean), spread (variation), and shape. You might also consider collect. ing data for a specified period o time: hour, shift,
day, week, etc.
. Use historical data to find patterns or to use as a
baselinemeasureof past performance.
Prepare a requency table ftom the ddta
a) Count the number of data points, n, in the sample.

10.2 9.4 't0.

9.9
1 0 . 1 9.8
9.8 9.8 1 0 . 1 9.9
9.8 9.9 1 0 . 0
9.4 9.6 10.0 9.8 9.9 1 0 . 1 1 0 . 4 10.0
10.2 1 0 . 1 9.8 1 0 . 1 1 0 . 3 10.0 10.2 9.8 10.7
9.9 10.7 9.3 10.3 9.9 9.8 1 0 . 3
9.9
9.3 1 0 . 2
9.9
9.9 9.8
9.4
9.0 9.5 9.7
9.8 9.8 9.3 9.6 9.7
10.0
9.4 9.8 9.4 9.6 10.0 '10.3 9.8
'10.11 0 . 0 9.8 '10.1
'10.6
1 0 . 1 10.2 9.8 9.5
9.6 9.4 1 0 . 1
'10.1 9.8 9.7 1 0 . 0
9.7
10.3
10.0
9.8 9.9 9.2 10.0 '10.0 9.7 9.7
9.9 10.4 9.3
9.7 9.7 10.7
10,2
10.2 9.8 9.3 9.6 9.5 9.6 't0.7
In this example,there are 125data points, n = 125.
b) Determine the range, R, for the entire sample.
Therange is the smallestvalue inthe setodata
subtracted from the largest value. For our
example:
R = X'",-Xa" = 10.7- 9.0 = 1.7
c) Determine the number of classintervals, k, needed.

Hierogram

67

. Metlbd 1:Takethe squareroot o the total numbs

of
clatapoints and round to the nearestwhole nunlbr.
k =fi25 = 1.18 = 11 intervals
. Method2: U\e therablebelowto providea
guide_
line for dividing your smpleinto a reasonable
number o classes.
Number o
Data Points

Number o
Classes(k)

50 - 100
100- 250
Over250

5-7
6-10
7-12
10-20

For our erampte.I2q dara points \aould be di_

vided into 7-12 classintervals.
Tirt These two methods are general rules of thumb
for determining classinterials. In both methods,
consider using k = 10 classintervals or easeoi
"mental" calculation.
7ip The number of intervals can inf,luence the pattern oFthe )dmple. Too few intervalswill pro_
du\e a ti8ht, high pdHern. Too many inte;ats
wllt producea rpreadout, ldt pttern.
d)Determine the classwidth, H.
. The formula or this is:

n = f= S = . r z
. Round yournumberto the nearestvalue with
the
samedecirnalnumbers as the original sample.In
our erample,$ e wouJdrourd up o.20.ltisuseful
to hve intervalsdetined to one more deciml
place than the data collected.
e) Determine the classboundaries,or end points.

ooffi

:t

. Use the smallestindividual measurementin the

sample,or round to the next appropdate lowest
round number- This will be the lower end point
or the fiisf class interval. In our example this
would be 9.0.
. Add the classwidth, H, to the lower end point.
This \.a/illbe the lower end point for the '1exclass
inteval. For oul example:
9 . 0 + H = 9 . 0+ . 2 0 = 9 . 2 0

HI
:-t
H

f=
=

fi

Thus, the first class inteNal would be 9.00 and

everythin8 up to, but not ncluding9.20,that is,
9.00 through 9.19. The second class inteNal
would begin at 9.20and be everything up to, but
not including 9.40.
Tip BachclassinteNal l]:.-llstbemutuallyexclusioe,
that is, every data point will rt i^to one,and ofily
oe classinterval,
. Consecutivelyadd the classwidth to the lowest
classboundary until the k classintervals and/or
the range of all the numbe$ are obtained.
) Constructtheftequencytablebasedon the values
you computed in item "e."
A requency table based on the data ftom our
example i9 shown below.

5
-

EF4

Mid-

1
2
3
4
5
6
7
8
I
10

1
9.00-9.19 9 . 1 |
9
9.20-9.39 9.3 t-\+t lt
16
9.40-9_59 9.s l-r+f 1-1+tl-l+ |
27
9.60-9_79 9.7 t+'fi ++11+++r t+11 +++1ll
9.80-9.99 9.9 l-t+f l+ft +++1ulI +t+1 ++l | 31
22
1 0 . 0 0 - 1 0 . 1 'to.1
9
).++1L+11)+r 1+fi |
-f-ltt 1++t |
12
1 0 . 2 0 - 1 0 . 3 10.3
9
2
1 0 . 4 0 - 1 0 . 51
9 0 . 5| |
'to.1
++t1
5
10.60-10.79
0
1 0 . 8 0 - 1 0 . 910.9
9
Histogram

69

4. Draw a Histogdm from the frequency table

. On the vertical line, (y axis), draw the frequencs
(count) scale to cover classinterval with the hiehest frequencycounl.
. On the horizontal line, (x axis), draw the scale
elated to the variable you are measuring.
. For eachclassinteNal, drawabarwith the heieht
equalto the requencytally of that class.

Interprel the Histogram

a) Cerfel1fl8.Where is the distdbution centered? Is
the processrunning too high? Too low?

b)Varialion. Whdl is the viation or spread of the

data? Is it too variable?
Customet

wilhin
requiremenis

too vaiabl

70

Histogram

Histogam 7.1

c) Stutpe.What is the shaDe?Does it look like a

normal, bell-shaped distdbution? Is it positivelt
ornegatively skewed,that is, moe data valuesto
the left or to the dght? Are there twin (bi-modal)
or multiDle Deaks?

Centringand SpreadComparedlo
CustomerTargeland Llmits
Speciication
Limit

ia)cenlered and we lwilhin

cuslomerlimits.
Mainla n pesent siate.

icrbrj

NormalDistribution
(bJNo margii lor or
Aclrori Feduce va alion.

Bi-Modal
Distribution l\4ulti-l\4odal
Distribution

fuJL

PositivelySkewed NegativelySkewed

ip Some processesare naturally skewed; don't

expect every distdbution to follow a bell-shaped
Tip Always look for tradnor multiple peaks indicating that the data is coming from two or moe
diferent sources,e.9., shits, machines,people,
suppliers. If this is evident, stratiy the data.
drProcessC.qpabIty.
Compare the results o your
Histogram to your customer requirements or
specifications.Is your processcapableof meeting
the requirements,i.e., is the Histogram centered
on the target and within the specificationlimits?
72

Histogram

rc) Processrunninglow.
Dellive productseruice.
.rbn: Bringaverage

1d) Pmess too variable.

Ol*tive poduc/serylce.
,qcrbnrReduceva alion.
e) Press ori cenle. and
toovaiable Deective
producvsevics.
crbnr Cenlerbetterand
reducevaralion.

Tip Getsuspiciousof the accuracyofthedata ithe

Histogram suddenlystopsat onepoint (suchasa
specification limit) without some Drevious decline in the data. It could indicate that defective
product is being sorted out and is notincluded in
the sample.

Hlstogram

73

T t The Histogram is lated to the Control ChaL

Like a Control Chart, a normally distributed IIrstosam will have almost all its values within a3
stahdard deviations of the mean. Seehocss Cqbility or an illushation of this.
vafialions
Stem & Leaf Plot
This plot is a cross between a ftequency distributib
and Histogram. It exhibits the shape of a HistograD"
but prese es the original data values---one of its lt
benefits! Data is easily rccorded by writing the trailint
digits in the appropriate row of leading digits.

z
a

.0517
.o6la

igs

. l i s s

7
7
3
2
2
0
4
4

482694
20435 9
1 9 3 7 3 8 6 6
4 8 0 8 9 0 5
52776
369
7

. 0 9 1 2 4 1 6879
.1010234579
. l t l t 3 3 3 6 7 8 9
. 1 2 1 0 0 2 4 5 8 8 9
. 1 3 1 2 2 5 677
. 1 4 13 06 9
.15147
.16t4

In this example, the smallest value is .057 and tlre

largest value is .164.Using such a plot, it is easy to find
the median and range of the data.
. Median = middle data value (or average of dre
two middle values) when the data is ranked fom
smallest to lagest.
For this example, there are 52 data points. Thereore,
the average of the 26th and 27th value will give the
median value.
Median= (.113+.1 16)/2 = .1145
. RangB=Highestvalue-lowestvalue=.164-.057=.107
74

Histogam

Histogam

Time Distributlon o Calls /

E
+

+
+
+

HOTrep data May 22 to August 4

tnfomatioDoroided couneevot
Snih4in B@cn
Note: The HistoSrm identied LtueepeaL calling periods at
the beginning oithe workday and berre and re rhe trdi
tional luch hour. This .an help the HOTrps synchronize
staing with thei customer neds.

EI
Hislogram

75

,rR.
'>\

Inteelationship,,K]..
Digaph (lD)
f-L:
Ioaktng tor dnvers
\IV
& oulcames
Ll U

Why use it?

Toallowa teamtogvstemati(lly
identity,
anllze,
and
classifythe (usednd eifectreltionshtps
thdte{st
amongallcriticalissuessothatkeydriversorourcomes
can becomethe heart of an efectivesolunon.

What does it do?

. Encouragesteam members to think in multiple
directions rather than linearly
. Explores the cause and effect relationships
among all the issues,including the most contro. Allows the Ley issuesto emergenaturllyrather
than allowin&the issue:to be forcedby'a dom;nant or po\,a/etful team member
. Systematically sudaces the basic assumptions and
reasonso disagreementsamong team members
. Allows a team to identiyrootcause(s)evenwhen
credible data doesn't exist
tlow

do t do it?
1. Agree on the issue/problem statement

E
#

#
&

. If using an original statement, (it didn,t come

fromaprevioustoolotdiscussion),createacomffi,P"

E-..

plete sentence that is clearly unde$tood and

agroed on by team members.
. If using input from other tools, such asan Affinity
Diagram, make sure that the goal under discussion
is still the same and clearly understood.
Assemble the right team
. The ID requires more intimate knowledge o the
<ubjectunder discussionthdn is neededor the
Afinity. This is important i the final cause and
effect patterns are to be credible.
. The idedl teamcile is generllv4- people.However. lhis numbe canie ircrJasedis lone as the
issuesdre still v isiblednd the meetine is we-llfcilitad to encourage participafion and naintdin o.us.
Lay outall o the ideas/issuecardsthat have either
been brought tom other tools ot brainslormed
. Arrange 5-25 cads or notes in a large circular
pattern,leaving d! much sp(e as possiblefor
drau ing arrowJ.Uselarge,bold priniing,including a large number or letteron eachidea for quick
referencelater in the Drocess,

l t" .-.' l

l%'lrylffi*l
tl
ry:,t

Wmr@
lceetadesl

-l- Look fo. cause/influencerelationships between

.Il o the ideas and draw relationship arrows
. Choose any o tle ideassa stating point. ! at_lo the
ideasare numberedorlettered,w;k throuehthem
in sequence,
. An outgoing aIlow from an idea indicatesthat
it
is the stronger causeor influence.
Askot 6achcombinaflon:
l) tsthera cause/intuence
rtationship?
2) ll ys,whichdirection
ocause/intunce
is stronqer?

v_ n>-rtil
El

-t'-lt r -t{
"B'
Dcision:
causesor
intluences,A"

(^
\:,/

\\l"i
tr m:\o4i

tr F t\Flrelationship

/:\

\s/_ ill\_

l'l i I l"l
loal

lElt i lcl

P]
Decision:
No relationship

"E'causes
Dcision:
or

continudnextpage

Interelationship

'orood
co '.e.

\
v-_+*|[R==

nf
-/

l,(

\
1y.i

z
l"l

tr-

Decision:
Nolationship.
"4" ls complted.

-a
L

'T 4
/
F,t

W
- t r -E . l
Dcision:'8"
causesor
inllencs
rc."
"B"
Nowbginwlih
ndrp.trh
queslions
or all rmalnlng
combinatiom.

Tip Draw only one-way relationship arrows in the

direction of the stroneercauseor influence.Make
a decision on the stro..neer
- d,ft ectio . Do not ilrau)
two-headed arrozos,
Review and revise the fir6t ound ID
. Get additional input ftom people who are not on
the team to confirm or modiy the team's work.
Either bring the paper version to othets ot reproduce it using available sotware. Use a different
size print or a color marker to make additions or
deletions.
Tally the number o outgoing and incoming rrows
and select key items or urther planning
. Record and clearly mark next to each issue the
number of arrows going in and out of it.
. Find the item(s)with the highestnumber oforfgoifig arrozosa d fhe item(s) r^rith the highest numDefoI rtcomlnSarfows.
. Oulgoingaftou'. A high number o outgoing rrows
indicatesan item that is a root causeor diver. This is
{enelll? the issue that teams tackle fust
Interelationship

79

c i''.:'.mitg.arrows.Ahigh number o incoming ar- - L

:o!{s indicates an item that is a key outcome.-ThL<
. dnbecomea to(usior planningeitherasa meaJ|- ,^-rngfulmed5ureofoverallsucces:orasaredefini tion of the original issueunder di5cussion.
7p Use common sen<ewhen you selectthe mosi
Lrili(dl icsuesto ocus
on. Issueswith vey close
tallies must be reviewed catefulty but in the end,
it is a judgment call, not science.'
1

+
+

7. Da ' the inal ID

.Identiy visually both the *cy driaph
\greles
numbeof outgoing?110.6rtand theAey;ulcomp.
(greatestnumber ot incomingarrowt. l ypical
methods are double boxes or old boxes.
What are ihe ssues related
lo educing littr?

+
+
#
+
+
+
+
C

Vafiations'
When it is necessaryto createa more orderly display of
allof therelationships,a matrix format isveryeffective.
The vertical (up) arrow is a driving cause and the
hoizontal (side)arrow is an effect.The examPlebelow
has added symbols indicating the strength of the relationships.
"relationshiP
The "total" column is the sum of all ofthe
strengths" in eachrow. This shows that you are working
on those items that have the strongest efect on the
ereatest number of issues.
lD - MatrixFormat

;E
J

b:

E!
.

6
3

=
I

o o A o
J J
o
o o o
o o
o o
J
A o o
o
J
o o o o
J

I;
;

J
1 16

24

18

22

0 24

O= I Siqnilicanl
O=SModlum
lntomaton prcvidd counesy or u.S. Ait Force, An Conbt Connan.t

Intelationship Il

fnterrelationship
Oigraph /

-)
lssuesSurrounding
lmDlementationo the Business Plan

+
E
E
E
+

Interrefationship
Digraph f
A vision of Andover
in the 21st Century

J)

See next page lor close up

+
+

lnomalion prcvided
counesyof Goodyear
Note: "The drivers" from this ID will be used as ihe goal in
theTreeexampleshown at the end o theTreeDiagram/PDPC

82

Interrelationship

+
)all

Intrrlationship

83

.z
DiSraPh{
Interrelationship
A Vision o Andover
in the 21st Century

,D(l

F
E
+
#,

Inltnalion provided cou ttsy

or oM of Andovet. MA

@ This is the driver. I the focus on the citizen as a

customerbecomesthe coreof the town'svisionthen
everything elsewill be advanced.
@
- This is the primary outcome.It puts the pregevation onaturein tetown a5 ky indicatorot the
vision working.

+
++,
g

a
a
L-.484

lnterielationshlp

b-4

Matrix Diagram
Ftndng relationships

Why us it?
To alloy a team or individual to systematicallyiden_
tilv, analyze,-and rate the presen;e and streigth o
relationshipsbetweentwo oi more setso inJorm"ation.
What dos it do?
. Makes patterns of responsibilities visible and
clear so that there is an even and appropriate
distribution of tasks
. Helps a team get conseruus on small decisions, enhancing the quality and support for the 6nal decision
. Improves a team's discipline in systematcally
taking a hard look at a largenumbei of importa
cteclsrontactors

Types o natrtces
Most Common
. L-sharydmatix.Twosets oitems dircctty compared
to each other or a single set compared fo itself.
OrintingNw Emptoyees

o
^

o
A

o
.A

o
o

Concluslon: SJprv,sorsand associatesfavota^en on rne

onellarclrote rate than fl^faditiondthumr resor,c uncton.

. The possible meanings of the s''mbols are almost

endless.The only requiement is that the team
comes to a clear undestanding and createsan
equally clear legend with the m;tri\.
5, Complete the matrix
. If dishibuting responsibilities, use only one "primary esponsibility" s)'rnbol to show ultimate
accountability.All other core team memberscan
be given secondaryresponsibilities.
Tip Focus the quality of the decision in eachmatrix
cell.Do not try to "stackthedeck" by consciously
building a pattem ofdecisions.Let thesepatterns
emergenaturally.
ip Interpret the matrix using total numedcal values only when it adds value. Often the visual
pattern is suficient to rnterpret the overall results.
Variations
The matdx is one o the most versatile tools available.
The important skilt to master is "matdx thinking." This
approach allows a team to focus its discussionon reltd factors that are explored thoroughly. The separate
conclusions are then brought together to create highquality decisions.Use your creativity in determining
which factors affect each other, and in choosing the
matrix format that will help ocus the discussion toward the ultimate decision.

rr
HI

t--

l--=

D-4

b-4

= =.
2r.+

CEE

=1

E
E}
E O

H
Matrix

t-

88

Matix

LogisticsAnnual Ptanv)

i4

*E.
*

E F 3 ;
6=z

oo< #
Maix 89


Logistics Annual Plan /

3<

'Ei

(9

o
ooo

sl>

-3

,rr

l!

s
ll,$
E
a

t 6
6 ll
l9-;;l
lcEEI
_ 1 . 5 t

SIEEEI
otoS

oEg

r,m'c

E
+
#

F
t! eF e
S b 6 BE

Matlx

z
d

f Its

ffi
90

e
E
E

E
8e

Ranking fot consensus

A. 2+2
B. 1+1
C. 3+4
D. 4+3

Why use it?

Allows a team to quickly come to a consensus on the
relative importanceofissues,problems,or solutionsby
completing individual importance rankings into a

(al

;^

F
3
:

ter",ru*ql

+
E

a' iSlli,X',?it'r

Matix

+
+
+

E=

fi h:l

nri^+ic

Whet does it do?

. Builds commitment to the team's choice through
equal participation in the process
. Allows every team member to rank issues without being pressuredby others
. Puts quiet team members on an equal footing
with more dominant members
. Iva*esa team's consensus(or lack o it) visible;
the major causes of disagreement can be discussed

How do t do it?
1. Generatethe list o issu6,problems, or solutions
to be prioritized
. ln a new team with memberswho are not accustomed to team participation, it may feel safer to
do wiitten, silent brainstorming, especially when
dealing with sensitivetopics.
2. Wite statementson a fliDchart or bord

+
I

Eliminate duplicates and/or clariy meanings o

any o the statements
. As a leader.a/?rayc
ask for the team'cpermiscion
and guidance when changing statements.
Record the final list o statemenls on a lipchart
or bord
xample:Why doesthe depaftment have mconsrstent output?
A
Lack of training
B
No documentedprocess
C Unclear quality standards
D
Lack _o cooperation with thr
oepafttnen$
E
High turnover
. Use letters rather than numbers to identiiv eac!
slatemenlso lhat teammembersdo not q;r a. :1fused by the ranking process that follo$s
Eachteammember records the corlespondirg letllF
on a piece o paper allrdrank orders the slatmG
Example: Larry's sheet o paper looks like _:
A 4
B
5

c 3
D 1
E
2
. This exampleuses"5" as the most important
ranking and "1" as the leastimportant.Since
individualrankingswill laterbe combined,this
"reverse

92

order"

NcTiMuhivoting

minimizes

the effect of team

olss4GoauoPc

:r
a-a

+
H
-

a
a
G1

ri

a
E-

membersleaving some statementsblank. Therefore, a blank (value = 0) would not, in effect,

in.rrs

i.

inn^7tin.c

6. Combine the .ankings of all team members

Total
Paige Si
Larry Nina Norm
A
4
5
2
2
1
8
5
4
5
3
5
c
3
1
3
4
4
D
I
2
1
.
5
2
3
8
E
2
3
4
'
I
"No documented process," B, would be the highest
priodty. The team would work on this first and then
move through the rest o the list as needed.
vafiations
One Hal Plus One
When dealing with a large number of choicesit may be
necessaryto limit the numberofitems ranked.The "one
half plus one" approachwould rank only a portion of
the total. For example,if20 ideaswere generated,then
team memberswould rank only the top 11 choices.If
needed,this processcould be repeatedwith the remaining 9 items, ranking the top 5 or 6 items, (half of 9 = 4.5
+ 1 = 5.5),until a manageablenumber are identified.
Weighted Multivoting
the relative impor
Eachteam member tates,r?ofrarTks,
tanceo choicesby dist buting a value, e.9.,100points,
acrossthe options. Each team member can distribute
this value among asmanyoras few choicesasdesired.

o@rre

Example:
Nina
I-ffry
A 2 0
8 4 0 8 0
c
3
0
5
D
5
1 0
8 1 0

Norm
10
50
10
10
20

Paige
100

Total
= 3 0
45 = 315
2 5 = 7 0
2 0 = 3 5
1 0 = 5 0

,i'

Si

With large numbers o choices,or when the voting for

ihe top ahoicesis very close, this processcan be repeatefor an agreeduPon number of items. StoP&'hen
the choiceis clear.

E
+
+
+
+
F+a

94

NcT,iMultivofng

Paretochart
Focus on key problems

Why use it?

To focus efots on the problems that offer the greatest
potential for improvement by showing their relative
frequencyor size in a descendingbar graph.
What does it do?
. Helps a team to ocus on those causesthat will
have the greatestimpact if solved
. Basedon the proven Paretoprinciple: 20%of the
sources(dule 80qoodny problerr
. Displays the relative importanceof problemsin a
simple, quickly interpreted, visual iormat
. Helps prevent "shifting the problem" where the "solution" emoves some causesbut wotsens otheis
. Progress is measured in a highly visible format
that provides incentive to push on for more improvement

,low do t do ,t?

*
))

1. Decide which problem you want to know moe

about
Example: Consider the caseof HOTrep, an internal computer network help line: Why do people
call the HOTrep help line; what problems are
people having?
Z

Choose th causesor problems thal will be monitore4 compared, and rank ordered by brinstomring or with existing data
a) BrainstorminS

Example: What are t)?ical problems that users

ask about on the HOTrep help line?
b) Based on existing data
Example:What problemsin the last month have
userscalled in to the HOTrep help line?
Choose the most meaningful unil of measuiement such as requency or co8l
. Sometimes you don't know before the study
which unit o measurementis best. Be DreDared
to do both frequen(y and cost.
Example: For the HOTrep data the most importdnt mesure is frequency becdusethe prject
team can use the information to simDlifu software, improve documenttion or triining, or
solve bigger systemproblems.
Choooethe time period o the study
. Choose a time period that is long enough to
representthe situation. Longer studies doi,t always translateto lreffelinformation. Look first at
volume and variety within the data.
. Make sure the scheduled time is tvpical in order
to take into account seasonality oieven dierent
pattems within a given day or week.
Example: Review HOTrep help line calls for 10
weeks (May 22-August 4).
Catherthe necessary
dtaon eachpoblem ctegory
either by "rerl time" o reyiewin; historicl dta. Whether data is gathered in "r-eal tirne" or historically, check sheets are the easiest method for
collecting data.
Example Gathered HOTrep help line calls data
basedon the review o incidnt repofts (historical).
96

Palo

'e1rr4nta

Tip Always ncl]J.de with the source dt and the

final chart the identifiers that indicate the source,
location, and time period covered.
Compare the relative requency or cosl o each
problem category
Example:
Frequency
ProblmCaiegory
3
Badconfiguration
68
Boolproblems
Filproblems
A
20
Lat.connection
Pinlproblms
16
24
Bolection
hang
Relection
sys.integdty
11
misc.
6
Relections
16
Syslemconiguration
19
Sysleminlegdly
15
Olhers
Total 205

Percnt(7")
1
33
4
10
I
12
5
3
I
9
7

7. Lisl the prcblem categories on lhe horizontal line

and ftequencies on the verlical line
. List the categodes in descending order from left
to right on the hozontal line with bars above
eachproblem category to indicate its ftequency or
cost.List the unit of measureon the vertical line.
(Optional) Draw the cumulative pecentage line
showing the portion of the total that eachpoblem
category reprcsentS
a) On the vertical line, (opposite the raw data, #, $,
etc.),record 100%opposite the total number and
50qoat the halfway point. Fill in the remainin8
percentagesdrawn to scale.

Paeto

97

b) Starringwilh thehighesL
problemcategory,
dra\
adotor mdrkan \ dl the u pper righthdndcomer
. Add the total o the next problem categoryto the
irst and draw a dot abovthatbar sho;ing both
thecumulativenumberand percentage.
Comecr
the dols and relord the remaiungiumuranr"
totals until 100%is rea.hd
220
200
180
P

160

o
3

120

& ioo

>

5 : E eg q
'
n 9

Fo 5 != E

9. Interpet the result

. Generully,thetallest bars indicate the biggest contributors to the overall problem. Dealing with
these problem categoies first thereore makes
common sense.Btf, the most frequent or exPen'
sive is not always the most important. Always
ask What hasthe most impact on the goalsof our
businessand customers?
Vartafions
The Pareto Chart is one o the most widely and creativelv usd imDrovement tool. The variations used most
frequntly are:
A. Maior Cause Brcakdowns in which the "tallest
ba/' is broken into subcausesin a second,linked
Pareto.
B. Before and Ate in which the "new Paleto" bars
are drawn side by side with the original Pareto,
showing the effect o a change. It can be drawn as
one chart ot two seParatecharts.
C. Change lhe Source of Data in which data is
collected on the same problem but from different
departments, locations, equipment, and so on/
and shown in side-by-sideParetoCharts.
D. Change Measuemenl scale in which the same
categoriesare used but measured differently.
Typically "cost" and "ftequency" arealtemated.

I E s : e . s E g "
3 F = * Es ! b
; i
t . * i 9 " cc ;

di

*
lnfamalian Navided coudesv of SnithKtine Beechan'

98

Parto

Pareto 99

II
Pareto

--.
A. MaiorCauseBreakdowns /
ReducedPaymentFightBills
ToralBills(329)
120
110
100
90

oo
3o
2a

HI

Pareto

HI
H

'ou

('

lqao

a
:

2
n

t-4
l-4

o E = : =e
r
3

(I)
IL

ReducedPaymentFreightBills
RalDispuls
Conlacl

/
i E L
\a 4ol

14

(E)

q)

uo[Eu[sp
tEurouo

I
sselc

: l ; c s g, i;

i r s - =
i
!

bj.=

:-:

stndsrp
lPlperluoc

qlqp#

4
hondhon ptavded catdesy ol Goodyea'

o ""i:. .o"

-l

om

E
Q

epeuEc

--

1 0l

1oo p*"t"

.a
lLt6!M

'csly'{

E 2a1

l
b

2
n

E-l

B. Beore and Ater /

Pareto
C. Changethe Source o Datal

Pareto

f
-:l

Reasonor FailedAppojntments
Sourceo Datais: ShoreCommands

:-t

,/

--L
C. Changethe Source o Data /

Reasonior FailedAppointments
Sourceof Datais: FleetCommands

:4r
f4
14
t4
4

F4

<6
h lotnat ion p rcr ided.a, desI ol
u s. Navy Nvat Dnlat Cntet San Otga

dd

>4
lv1

t,afeto

o o"j -oo-o1

Fd

lnlamalian prcvided coortesyo

U.S. Nary, NavalDentat Cente., San Diesa

Pareto 103

Paeto

{
D. Change MeasurementScale /
FieldServiceCustomeComplaints
lzz or.tt"orpt"int"

a :
6 g o
E

Costto RectiyFieldServiceComplaints
ts* otror"tcos

20,000

Weighingyaur oplions Ln--f]

Why use it?

To narrow down options through a systemahc aPproach of compadng choicesby selecting,weighting,
and applying criteria.
Whatdoes
do?
. Quickly sudaces basic disageements so they
may be resolved up front
. Forces a team to focus on the best thing(s) to do,
and not everything they could do, dramahcally
increasing the ances for implementation success
. Limits "hiddenagendas" by surfacingthe cdteda
as a necessarypat of the process
. Increasesthe chanceof follow-through because
consensus is sought at each step in the process
(ftom cdteda to conclusions)
. Reducesthe chancesof selectingsomeone's"pet
project"

*
How do , do it? ))

15,000
I
.E 10,000
6

There are three methodsor constructingPriodtization

Matrices. The outline that follows indicates typical
situations or using each method. Only the "Full Analytical Criteria Method" is discussedhere. The others
are covered ully in Ifte Memory logger PIus+@.

5,000
0
E

E
=

s
N

,5\. il'ji,1'jLf''"-_flffi

4 . <

"

Full Analltical

Full Analytical Crileria Method

Typically use when:
. Smaller teams are involved (3-8 people)
. Options are few (5-10 choices)
. There ale relatively few cdteda (3-6 items)
. Complete consensus is needed
. The stakesare high i the plan ails
ConsensusCriteria Method
This method follows the same steps as in the Full
Analytical Criteria Method except the Consensus
Criteria Method uses a combination of weighted
voting, and ranking is used instead of paired comparisons.
Typically use when:
. Larger teamsare involved (8 or more people)
. Options are many (10-20 choices)
. There are a significant number of critelia (6-15
items.)
. Quick consensusis neededto proceed
Combination ID/Matdx Method
This method is dierent ftom the othe! two methods
becauseit is based on cause and effect, rather than
criteria.
Typically use wheni
. lnterrelationships among options are high and
finding the option vrith the greatest impact is
cdtical

06 Prioitizadon

o1"4couorc

Criteria Method

1. Agree on the ultimate goal to be chieved in a

clear, concisesenlence
. Ifno other tools areused asinput, produce a clear
goal statement through consensus.This statement stronslv affectswhich criteria are used.
Choosethe most enjoyablevacation
or the wholefamily
Createthe list of criteria
. Brainstorm the list of criteria or review preyious
documents or guidelines that are available,e.9.,
corporategoals,budget-relatedguidelines.
.
.
.
.

Cost
Educationalvalue
Diverseactivity
Escapereality

|ip lhe team must rcach consensuson the final

oiteria aIld their meanings or the process is likely
to fail!
3. Using n L-shaped matrix, weight each criterion
against each other
. Reading acrossfrom the vertical axis, compare
eachcdtedon to thoseon the horizontal axis.
. Eachtime a weight (e.9.,1,5, 10) is recorded in a
mustbe
row cell, its reciprocalvalue (e.9.,75,1/lo)
recordedin the coresponding column cell.
. Total each horizontal row and convert to a relative
decimal value known as the "criteda weighting."

P.ioritization 107

Cdterionvs. Criterion
Educ.

cost
Cily

1
5

Educ.

Optionsvs. EachCriteon(CostCriterion)

5.3

10.2

.28

burg

20

.55

City

.60

.o2

l0

10

1
5

Total

1
5

f
0

10

5.3

.12

10

15,2

.33

_40

.01

25

.54

10

a
l0
10
Grand
Tll

r = Equarcosr

36.1

,15.9

10 = Much less expnsive

1 = Equallyimportanr

r/ro = Mucn lBe

I 0 = ll,luchmoeimpotant
elativDclmatvatu
Eachrowtorat+ by the grandtotal

r^o = [4uchtossimportant

Compare ALL options relative to each weighled

critedon
. For each criter1r1,create an L-shaped mattu
with all of the options on both the ;ertical and
horizontal axis and the criteda listed in the
lefthand corner of the rr.alffix.Therc uill be as
mafly optiofis matrices rs there arc criteria to
be applied.
. Use_thesameratint scale1,5, 10)a< in Step
3,
aUI customizethe wording for eachcriterion.
. The relative decimal value is the,,option
rating.,,
108 Prioritizaiion

--

oPr

Cdrir-e S

epnsive

a thoughhe moe@ionvcddon matices,lik6this:

DiveBe activity
_l
opt'o"s El optffi-l

fT-fT_n fT-frn

Educlionalvlue

|d

_l
optons

--|-I-r]

r-FrF
r-rrln-rn

l'f++fll"1-+++]
I'Fi-fl
p The whole number (1,5,10) must always represent a desirablerating. In some casesthis may
mean "less," e.9., cost, in others this may mean
"more," e.9.,tasty.

Priorilization 109

5. Using an L-shaped summary malrix, compae

each option based on all criteria combined
. List all criteria on the hodzontal axis and all
optons on the vertical axis.
. In each mahix cell multiply the ',cdteria weighting" of eachcdterion (decimalvalue from Step3)
by lhe "oplion raHng"{decimalvalue from Step
4). This createsan "option scote."
. Add eachoption scoreactossallcriteria for a row
total. Divide eachrow total by the grand total and
convert to the final decimal value. Compare these
decimal values to help you decide which option
to pursue.

Summary/latrix
Optionsvs. All Criteria
(.15)

(.23)

{.55t

\.o7)

1.22)

(.01)

. 3 3x . 1 5
(.0s)

(.10)

(.06)

(0)

.01x.15
(0)
(.10)

(.27)

(0)

lr-I
f-{

(0)

.3

.09

.09

a
I

\-.--

F=r

Oplionse

Prioidzation

_ -

frl

variations
*e The Memory log&et Plts+o for full explanations of
both the Consensus Cdteria Method and the Combination lDlMatrix Method. The Full Analytical Criteria
\tethod, illustrated in this book, is recommended becar.r,<e
it encouraees full disorssion and consensus on
criti.al i!"-rrs. Th; Full Anal'tical Criteria Method is a
sinpled dptation of an even more rigorous model
L1--lr-:: us iie -{nalr-tical Flierchy Process.It is based
.--:e'- ork.-f Ihomas Saatv,which he describesin his
r.1Je:1--:-": -\,ziin3 hr ltaders.h a y case/use comr
ad to l31o\^_rr'hen a sifuation is important
dvl: to rr'arrant su thorough processes.

32

.54
x
.15
(rrom
slp4narax) {rom
srepsmarx)
(.08)

.l1o

6. Choose the best option(s) acossall criteria

ip yy'hile this is more systematic than baditional
decision malong, it is not a science.Use common
senseand judgment when options ate rated very
closely, but be open to non-traditional conclusions.

i un.'usl, rulo,' ,ol or" uu

{.oB) ,0, ,.r,
lHnry!
|
I

."""

+
fL4

(.02)

(.02)

+
+

,1rr4
coar,rorc

Prioritizationlll

Prioritization -.-\f

Prioritization -.,\,/

Choosing a Standard I
Corporate SpreadshetProgram

Choosing a Standard /
Corporate SpreadsheetProgram (cont.)

O Weightingciteria(describedin Step 3)
intotal.
Thisis a portion
oa ullmahixwith14criteria
o

@ Comparingoptions(describedin Step4)
Theseareiust2 of 14 maices.
ad

Program Program Program Total

(6

.q

Criteria

B
1 0 .20 3.7

Program

1.00

ro

Program

.0'1

Program
1.00
c

.20 .20 35.4 .08

1.00

1.00

2.OO

.33

1.00

2.OO

.33

2.00

.33

1.00

Grand 6.00
toltl

lVaximum
functionality1 0 . 05.0
Best 5.0 5.0
Derormance

Lowesl Program Program Program

ongorng
cgt

Program

.10

tolel

Relaliv
Decima

.20

.30

.02

lnfomation prcvidedcouesy of t'lovaco.Chenicals

Note: This constructed example, illustrated on the pa8s,
represents only a ponion o the pnoitization procss and only
a portion o Novacor's spreadsheet valuation process.
Novacor Chemicals assembled a 16-person tam, comprised
mainly of system users and some information systems staft
The team developed and weighted 14 standard critria and
then applied them to choices in word processing, spreadsheet,
and presentationgraphicsproSrams.
ThisxanPtecontinuedDenPg

Program
10.00
B
Program
c

5.00

5.00
.20

'15.00
5.20

Grand 20.50
tolal
lnfomation provided courtsy o Novacot Chemi.ls
his exanple cntiiued nn page

112 Prioitizadon

Prioritization 113

Prioritization

tl

E:

{
Choosing a Standard /

Corporate SpreadsheetProgram (cont.)

4 'g:it:!8,'
a-

OptionRatingsAcrossAll criteria -{
@ Summarize
(described
in Step5)
This is a portionof a ull matixwith 14 criteriain total
Bost
Tcal
Easylo
Decima
ongorng
use(.08) ntogratiol icosr
.oa)
Program
.16
.18
.03(.01) .33(.03) .02(0)

x:4,

Program

.48(.04) .33 (.03) .73(.06)

Program

.48(.04) .33(.03)

.25(.02\

.30

.33

.44

.49

E=

lnlornationprcvidedcoudesyof NovacotChmicals
Result Piosram C was.hosen. Even though l4 out of the 16
team memblers were not cu[ently ushg this pogram, the
prioritiztion process changed their minds. nd Prevented
them from biasinthe final decision.

There are many standardmodels or making improvements. They all attempt to provide a repeatableset of
steps that a team or individual can leam and follow. The
Improvement Storyboard is only one of many models
that include tt?ical steps using typical tools. Follow
this model or any other model that createsa common
lnguage for continuous improvement within your

ot8"'''t"no"
Plan

E=
E-

Ea
Ea
E=r
E-a

1, Select the problem/process that will be addressed first (o next) and des.ribe the improvement opportunity.
2. Describe the current prcce6s suounding the
impiovement opportuaity,
3. Describe all of the possible causesof the problem and agtee on the root cause(s).
4, Develop an efective and wokable solution and
action plan, including targets or improvement.
Do
5. Implmenl the solution ot prccesschang,
Chck
6. Review and evaluate the result oI the change.
Act

114 Priorilization

\"-/

li

F4

Grand .90
Total

[3:"J3:i"J3]sJ,ent
Moder
1.- \

b{

7. Relect and act on learnings.

r'!
Depending on your formal process structure, Step 1
may be done by a steering committee, management
team, or improvement team, I you are an impovement
team leader or member, be prepared to start with Step 1o/
Step2.
1. Select the problenr./process thal will be addressed
irst (or next) and describe the improvemenl oppohlnity.
. Look for changesin important businessindicators
. Assembleand support the right team
. Review customerdata
. Narrow down projectfocus.Deveiopprojectpurpose statement
Typical tools
Brainstorming,Affinity Diagram, CheckSheet,Control
Chart, Histogram, Interrelationship Digraplr Pareto
Chart, Prioritization Matrices, ProcessCapability, Radar Chaft, Run Chart
Situation
Stop 'N Go Pizzanis a small but rccently growing pizza
delivery business with six shops. Ater a period of rapid
growth, Stop'N Go Pizzaexpedencedasix-month decline
in volume. Customerswere leaving. Top management
formed a mixed team o store managers,kitchen staft
and delivery personnel to find out why, and to generate
an implementation plan to coect the situation. The
team used both the Run Chart and ParetoChart.
' l h e n a m e \ t o D l \ C o l v z . a n d t h d t a s < o c i d t e d$ i t i t l i s
case study arficiional. Any similiariiy to an actual company
bv this nme is Durelv coincidenil.

ffi

Bun Cha
Averagemonthlyvolumeo deliveis
(pe shop)

-i

E=
5=
E=

My rr

uq s;,

oa

Nv Dc

Pareto Chart
T\9eso customecomplaints
Total=2520
Oclober-December

r-

J;

(across6 shops)

2ah-

EI

FG4
E'
E=

lllu$rarion note: DeliveNlime was delinedbv lh

lotall me lron whnlh odrwas Dlacedlo vihn
lhe cusloher r&eived il.

PgPl I'Iodel I17

Parto Chart
400
350
300
Q
' 250
200
o 150
100

Lat delivey complaints

Tolal=1890 octobr-Dcembe
(across 6 shops)

r-

= rrioav

= saturoay

E=r

f] = ote'
Shops

Pocess to poducing and delivering

Stop ' Go Pizza

EA

Decision
Late deliveries, (late from the time of o!der), were ,yfar
the most frequent customer complaint across all loctions, especially on Fridys and Satuldays.

5=

Tem purpose statement

Reduce late delivedes on Fddays and Saturdays.

ra

2. Descdbe the current proce8s surrounding the improvement oppounity,

. Select the relevant process or process segment to
define the scope of the project
. Describe the process under study

that contributed to the proiect focus. The team used a

Macro Flowchart.

F=r

5a
E=r

Decision
It became obvious that the '1ate delivedes" went far
beyond the physical delivery process. Everything in
the Macro Flowchart afected the "order-to-eating"
time. This total process improvement became the
team's focus.
3. Descibe all of the possible causeso the problem
and agreeon the root cause(s),
. Identify and gather helpful facts and opinions on
the cause(s) of the problem
. Confirm opinions on oot cause(s) with data
whenever possible

Typical tools
Brainstorming, Macro, Top-dowlL and Deployment
Flowcharts, Tree Diagram

Situation
The team began to understand the overall process for
producing and delivedng their product and problems

Typical tools
Affinity Diagram, Bainstorming, C & E/Fishbone Diagram, Check Sheet Force Field Analysis,Interelationship Digraph, Multivotin& Nominal Group Technique, Pareto Cha, Run Chart, Scatter Dagram

E=r

@.rr.co,Jo'c

18

PgPl Model

PgPlModl 11g

Situation
The team brainstormed all of the possible causes for
"late deliveries" and then continued to ask "Why?" so
that possible cause pattems could emerge. The team
used a C & E/Fishbone Diagram and Run Charts.

+
+
efr

EA
E t!!

iF

9
o;

O N

E*

:
3t

.g

E
;
urb

{s
o.c

I t l
P l r

Eg E

E
'5

-:

+'

ut

.E

I
=
120 PgPlModel

E=
Eg

va

Ad rt

Jf

Ju' rrq Sep Od Nov Dec

e
9

-:

EE=l
E=l
E=
E=t

Jan Fob Ma Apr May Jun Jul ug S6p Oct Nov Oec

PS/PIModel

'l2l

Decision
The C & E/Fishbone Diagram repeatedly pointed to
"turnove/' and "lack of training" as root
causes.This
applied to Ll areas of the operatiorL NOT just in the
actual delivery portion of the process. Subsequent data
showed that as the business rapidly grew, less time was
put into training all employees.With this lack of adequate
haining, many employeesfelt a great deal of pressure at
the busiesttimes. They alsofelt that they were unable to do
their jobs well. This combination of work pressure and
lack of self-confidenceoften causedemployees to quit.
4. Develop an efective and workable solution and
action plan, including tagets or improvement.
. Define and rank solutions
. Plan the changeprocess:\ /hat? Who? When?
. Docontingencyplanningwhendealingwithnew
and sky plans
. Set targets for improvement and establish
m^nif^riho

Situation
The team used the combination of the Tree Diagram,
Prioritization Matrices, Responsibility Matrix, and
GanttChart to createa workable plan that attackedthe
nf rhp

nr^hlm

Decision
The team focused on the most efective, efficient way to
deliver the new training. They developeda four-monfh
22 pgptModet

rt

Tee Diagram
Reducetumovrrateof employees
(company-weJ

{
{

f4

mah^Ae

Typical tools
Activity Network Diagram, Brainstorming,Flowchart,
Gantt Chart, Multivoting, Nominal Group Technique,
PDPC, Priodtization Matrices, Matrix Diagram, Tree
Diagram

harr

-:l

F=l
F.
FA

r
E=a
E-e

imDlementationDlan that feafured the creativeuse of

rideotapes, role plays, peer ratings, and so on. The team
set the following targets based on past performance:
. Reduce turnove! rate from 62 to 30 percent
. Reduce average time of order-to-delivery ftom 40
to 25 minutes
. Reducecustomercomplaintso late deliveriesby
50 percent, without increasing other complaint
categones
. Increaseaveragemonthly volume to 2400units
per shop from the current 1891units
PSi/Pl Modl

123

Prioatization llatix
program
Selecling
thebesltraining
components

ilatix & Gantt Chart Combined

Newlrainingprogramtimeline

,-
':
uJ
(.60)

Tin beore shit

Tainduingshit
Tain afteshift
Tainat th shop

Dovlopvidos

use terals
Use ole play

St cla psro-

(;)

( . 1 e ) (.01)

eoC o
C A o o
A o o o
oooo
o o o rn
eooo
ooCo
eooo
o oo o
eooo
oooo

7.70
3.74
1.77
7.70

5.29
8.83
4.t6

e=

8.89
7.70
8.83
8.89

[=l eoor
O= eExettenr Q=srair
=
Th total th sum o lrating valus x crihna wightingl
For exampl, to ind tfte total o the "Train beore shit" row,
do th fotlowing:
I o ( 9 ) x . 6 0 1 +t O ( 9 )x . 1 9 1 +t O ( 3 ) x . 1 9 1+ I O ( 3 )x . 0 1 1 = 7 . 7 0
Nore: Weighting values of eachcriterion comefrom a cdteda
matnx not shown,
Task oDtions comefrom th most detailed level o the
Tre Diagram on the previous page.

124 PgPl Model

a=

Fa
.

>

=
Pc
=
:
9 t > E

, E H i
l r z

oo<
PS/PIModel 125

5. Implement the solution or processchange.

.Itis.often recommendedto try the solution on a
small scaleirst
. Follow the plan and monitor the milestonesand
measures
Typi(l tools
Aciivity Network Diagram, Flowcha , canrt Chart,
Matrixbiagram, and oirt". p"o;".t u";;d;:
ods,aswell asgtheringongoingddtawiur Nur'\ ndrrs,
Check Sheets,Histograms,frocess Capability,and
controlchats

Eh";l

h+_

6. Review and evaluate the result o the change.

. Confirm or establishthe meanso monitoring the
solutron-Are ihe measuresvalid?
. Is the solution having the intended effect?Any
unintended conseque-nces?

. +--=l
- ^'l:

,
Ill:"Ij"-:t'
c:-iiheet.

conirol chaft, Flowchart, Pareto chat,

I=
, ---.
t4

Siluation
The team used the Responsibility Matrix and Gantt
charfto guide rhe rraining rollor.t.Th" ;.ig;t
plu;

.-.--ts4

i n g t i m e a s w e l l s s o m e m a n a g e r s n o t g e tt i n g c o v e r d q e
during training.

_ -_ r >

-,- -?li.li-,

ry and r-ebruo'" " * r. developedin Jdnud

--' a:ci rolled out in March and APdl The team went
:-=:\-:t' ihe ori8inal Paretochart-data to comPareit to
"omPlaint data ln addition, they
*ttot"t
:l:::::T

E= ;:::i:i"f,::Hffi1il*i''ii.1:.'"#
nini:ll^:X,lnjl'.W"mxn:ti'*i
pareto Chart
Beformptoyelrainrng

il;=
i,""'i1,o,,*u",ooaytheemproyeesroro^"-harrrhe
Ea
fraining time and to setup overlapping shifts for behe

rorar=2510
_.i.i,p!,
across6s'"ops)

October-December

1SOO]

_a
F=r
1;;;]
rn:i;jrx,n*ur*r,lii;:r*.;;
uj::;ji,i:"?i*:i".*:"#'"**
E
; ;;;;1
beneritsthat it was expandedto att 'll
lil;:""0"0

-i
--+l'.-r
126 pgre

-----,
H

soo
]

n*

,r"

* + F :o i. : r

ur"

3./.

The Pareto Chart on the next page shows the t'?es of

customer complaints after the training plan was rolled out
in March and April.
5.;; no^,o".

psipr rrodet i2z

Pareto Chart

-t

Ater employee taning

-l
.t ao0
-g
E

Total=1310
600

(across6 shops)

H
{

400

E 200
-6

i g 5

5e

Decision
The teamreviewed the original targets,which were set
in Step4.
Tumover: While not at the 30 percent average, it
had decreasedfor six consecutivemonths ftom a
high of 62 percent to 44 percent last month.
Delivery Time: Steadilydeclinedftom a high of40
minutes to its most recentlevel o 28 minutes.
Cuslomei Complaints: Overall, complaints were
reduced by 52 percent and within "late delivedes"
by 63 percent
SalesVolume: The averagevolume last month was
at 2250units, up or the third straight month.
Unexpectedesult
For thefirst time, customerscomplainedaboutthelack
ofa good selectionon themenu. WhileStop'N GoPizza
was working toward increasingits speedopizz deliveriesand standardizingits processes,the variety of the
menu was perhaps too smplified.
128

PgPl Model

f4
4
t4
Averagedeliverylime
(company-wide)

L4

q
E=l

-=r
-a

.E
Ezs
.E

r---f

o'qe4GouoPc

PgPlModel

129

*=l
7. Relect and act on lernings.
. Assessthe results and problem-solving process
and recommend changes
. Continue the impovement process where
needed;standardizationwhere possible
. Celebratesuccess
Typical toots
Affinity Diagram, Brainstorming, Imptovement
Storyboard,Radar Chart
Situation
Six months ater the new haining was started,the teammet
to evaluateits resultg and process,Team membersused a
Radar Cha to illustrate their assessmentof the team.
Decision
The Radar Chart showed strong agreement among
team memberson the perormanceof "Results," "Use
oftools," and "Impact on cusiomerc-"Theperformance
and consensusamong team membetswere both lower
in "Standardization"and "Teamwork-" When the team
presenteditsstoryboard to top management,themajor
result was a completeoverhaul ofbasictraining content
and delivery, aswell asthe new performancemeasures
that would continue to be monitored-The final celebra,
tionwas.. . whatelse. . . anall-expense-paidStop'N Go
Pizza party!
Future Possibilities
More efficient mapping, routing, and dispatching of
pizza deliveries,as well as more staff closs-training.

l3O

PgPlModel

olrr4coalroPc

Radar Cha
Teamvaluation
o itselaflernewtaining

:l
L-

-al
-'-!l

b-al

ta

Note: The "x" markindicats the team's avera8e Prformance

rating whil the shaded area indicates the range of ratings

a
a

G1

r
PS/PIModel 131

z3\'

ProcessCapability

E=-

Measuring conformance
ta customet requirements

r-

Why use it?

To determinewhether a process,given its natural variation, is capable of meeting establishedcustomer !equirements or specifications.
What dos it do?
. Helps a team answer the question, "Is the process
capable?"
. Helps to determine if there has been a changein
the process
. Helps to determinepercent of product or service
not meeting customerrequirements

ts=
r:-r
4
4

*
,lowdo, do it? -a

1. Determine the process grand avelage,X, and the

average iang, R
. Use a stable Contlol Chart, which means the
process is stable and normally distdbuted.

2. Determine the Uppe Specification Limit (USL)

and the Lower Specifi.ation Limit (LSL)
. The USLand LSL are basedoncustomet requirements. Recognize that these specification limits
are based solelyon customer requirements and do
not reflect the capacity of the process.
3. Calculate the processslandard deviation
. Processcababilityis bsedon individualpoints
ftom a processunder study. Information rom a
Control Chart canbeusedto estimatetheprocess'

IfrrI@a4M

averageand vadation (standarddeviation, s).

. o is a measureof the process(population) standard deviation and can be estimatedrom information on the Control Chart by
6=lord=i
02

where R and s are the averages o the subgroup

rangesand standard deviation, and d2and ca are
the associatedconstantvalues basedon the subgroup samplesizes.Seethe Table of Constantsin
ContnlCharl<.
. The processaverageis estimatedsimplyby X, X,
X.
1- Clculet the processcapabitity
. To measuJethe degeelo which a processis or is
not capableof meeting customer requirements,
c:pbjli' indices have been developed to comFare t}le distribution of your process in relation to
i;

Ea

a
ET

Fa

<i.i^n

lin+<

. A stableprocesscanberepresenledbya measure
of its rariation-six standard deviations. Comparint six standard deviations of the process
rariation to the customer specificahons provides
a measureo capability. Somemeasuresof capabililv indude Cpand its inverse Cy, Cp , Cpu,and Cpl.
ce (simpleprocsscapability)

=a

ca

. _ USL- LSL
6a
ip While Cprelatesthespread o the processrelative to the specificationwidth, it DOESNOT look
at how well the process average is centered to the
target value.

re

USL

LSL

it / l 1
i /
-

l,/

^'P-'-r

t=
b=l

r
\

\l

Theprocess
variation
xceeds

]-

specificalion. Defectivesar
binq made.

The processis just mting

speciication.
A minimumof
.3%de{ectives
will b made,
moreif the processis nol
centered,
The processvrialionis less
however,
thanspeciicalion,
defectivesmightbe madei
process
is not cnteredon
the
the targetvalu.

Cpr, Cpu,and Cpl(procsscapabilityindices)

The indices Cpl and Cpu (or single-sided

specificiationlimits) and Cpk(ortwo-sided specification limits) measure not oLly the process
variation with respectto the allowable specification. thev also take into account the location o the
ptocessirr".ug". Cpk is considereda measureof
the processcapability and is taken as the smaller
of either Cpt or Cpu
^

"p

X - L S L UeJ
^
e "

U S L _ Xwek

--l;_

||n|| lwp|' UpJJ

7p If the process is near normal and in statistical

control, Cpkcan be used to estimatethe exPected
percent of defective rnaterial. Estimating the percentagesofdefectivemate al isbeyond the scope
of this book and canbe found in statisticalbooks.
34

Proce-ssCapability

F-r

tsa
ts
t4
t4
b4

EE=
-7
E

Ea
Ea
Ea

. I the process is not capable, form a team to

identify and correct the common causes o the
variation in the process.
Processcapability,basedon individual data of
the processpopulation, is used to determine if
a processis capable o meeting customer requirements or specifications,It represents a
"snapshot" o the process for some specific
period of time.
- Control Charts use small sample sizes over
time and look at the averages.The control
limits are natural limits of variation of the
aoerages
wfhifi thesatlple.Theselimits arenot to
be contused with specification limits, which
are or ndiaid al datapointsin thepopulation.
VaJiations
The construction stepsdesc bed in this sectionare based
on process capability of a Vadable Data Control Cha.
The processcapability of an Attribute Data Control Chart
is representedby the processaveragesp, nB,c, and O.

ProcessCapabitity 35

Process Capability /
DieCuttlngProcess
1
A ControlChartwasmalntained,
producing
thollowing
statistics:
X = 2 1 2 . 5R = 1 . 2n = 5
S p e c . = 2 1 0 r 3U S L = 2 1 3L S L =2 0 7
6 = = 1 . 2= . s r o
d2 2.326

us!; LSL=2ll-. ?97=....3 = r.see

co
' = 60
6(.s16)
3.096

-207 _
._, _ x - LSL- 2'12.s
_55 _
=ffi
"0,=-3-==16)-:
= ".
s.5s3

=u!-x =215q;ft1-2.5
=
ceu
ffi =o.szs
Cpr = min {Cp1,Cp,} = 0.323

Radarchart
HAnng organEalon
performance

Why use it?

To |isually show in one graphic the size of the gaps
arrong a number of loth cuffefit orga izatio perormance area<and idealperormanceaeas.
?/hat does it do?
. Makes concentrations of strengths and weaknessesvisible
. Clearly displays the important categodesof performance
. If done well, clearly defines full performance in
eacn caregory
. Capfures the different perceptions of all the team
members about organization perormance

Since Cpk<1,defectivematerialis being made.

LSL

,fir

USL

,low do, do it? 't)

1. As6emble the dght teanftaters
iP It is crtical to get varied perspectives to avoid
organization "blind spots."
2. Selectand deine the rating categories
. Thechartcanhandlea-ridenumberocategodes,
with 5-10 categoiesas an average
. Brainstorm or bring headers from an Affinity
Diagram to createthe categodes.
. Define both non-performance and full performance within each category so tings are done
consistently.

136 PocsCapability

3, Construct the chart

. Draw a large wheel on a flipchart with as many
spokesas there are rating categories.
. Wite down each nting category at the end of
each spoke around the perimeter of the wheel.
. Mark eachspoke on a zero to "n" scalewith "0"
at the center equal to "no performance" and the
highest number on the scale at the outer dng
equal to "full perormance."Perormancecan b
measured either objectively or subjectively.

t:
E=
E=
:<

ip Make the team rating highly visible on the

chart. Be sue to dierentiate the team atings
from individual ratings on the chart by color or
type ofmark.
Connect the team iatings or each calegory and
highlight as needed

rl-!f
l--

Rate all perormanc calegories

a) Individual Each person rates in silence,using
multicolored markers or adhesivelabelsdirectly
on lhe flipchart.
b) Tea:Throughconsensusoranaverageof individual
score6,get a tem mting. Take into ac$unt both the
clustering and the spread of the individual ratings.

f-
=

a
tri

138 Radar

io1994GOUOPC

EC
I
E'
I
E
I

ip gap scorecan be added to eachcategoryby

subtractingtheteamrating scoreromthehighest
number on the rating scale,e,g.,ona scaleof "10,"
a team ratingof "4" producesa gap scoreof "6" in
categodesB and E.
6. Interpret and use the results
. The oveiall iatings identify gaps within each category but not the relative importance of the categories themselves. Work on the biggest gap in
the most critical caegory,
. Post the resulting Radar Chart in a prominent
place,review pogressregularly, and update the
chart accordingly. It is a great visual "report
card."

Radar 39

Radar

F=f

TOC Review Scores /

r+f

F+r

Planning
Procss

E4

chart 1'.,...-.
,l5\. lun
tracKtno Lrenos v
Why use it?
To allow a team to studyobserved data (a perormance
measure o a process)for trends or pattems over a
specifiedpe od of time.
Whet does it do?
. Monitors the performance o one or more processeso\.er time to detecttrends, shifts. or cycles
. -{llolrs a team to compare a pefformance measure before and after implementation of a solu;n-

4
f4
&

E-

Banseo ratingswiihinth lam

Ez

Compny'sqoal: lo have80o/.o allnlilies

(34)chievean overallscoreo >3.5.

To computeovrall score:
Sumo aveage
ea9hcarsoy = 13 =
z s (.",,inun i" s)
ii or careqones
lnfomattan Dtovided @utlesv ol Hewlett Packard

EI

E3a
j{

40

Radar

olee4coal-/oPc

Ea

rn mp:<rrrp

c imn:.f

. _!-jis aitstion on truly vital changesin the process

. l.a.ls userul intbrmation for predicting frends

,b,

*
&, do it? ,.6'
I

1. Decide on the processperormance measure

Z

Cather data
. Generally, collect 20-25 data points to detect
ma,hin-

'l

n:l_lcrhc

3. Create a graph with a vertical line (y axi6) and a

horizontal line (x axis)
. On the vertical line (y axis),daw the scalerelated
to the variable you are measudng.
- Armnge the y axis to cover the fuIl range o the
measurcmentsand then some, e.9., 1rl, times
the range of data.

. On the horizontal line (x axis). draw the time or

sequencescale,
4. Plot the data
. Lookat the data collected.If there areno obvious
tends, calculatethe averageot a thmetic mean.
The averageis the sum oithe measuredvalues
dividedbythenumberof datapoints.Themedian
value can also be used but the mean is the most
fuequently used measure of the 'tentering" of the
sample. (SeeData Points fot mote information on
averages.)Draw horizontal line at the average
value,
Tip Do not redraw this average line every time
new data is added. Only when there has been a
significant change in the plocess or prevailing
conditions should the averase be recalculated
and redrawn, and then only using the data points
after the vedfied chanee.

E
L=e
<
a
b4

5. Intepret the Chart

. Note the position of the averageline. Is it where it
shouldbe relativeto a customerneedot specification? Is it where you wnt it relative to your
businessobiective?
Tip A danger in using a Run Chart is the tendency
to seeevery variation in data asbeing impo{ant.
The Run Chart should be used to focus on truly
vital changesin the process.Simple testscan be
used to look or meaningful trends and pattems.
These tests are found in Control Charts in the
"Determinine if Your Processis Out of Control"
section. Remmber that for more sophisticated
uses, a Contol Chart is invaluable since it is
simplv a Run Chart with statistically-basedlimits.

),-,

F4

ra

+
F

o l

Average

< D l

I
---___________
,, (x axis)
me or sequence

142 Run

F.
Fi
E'
ET
E=

GoauoPc
@1ee4

Run

143

Run

+
+

Average Number o Day)

or Determining Eligibility or Servlcs

+
r+

EE$
o

0)

E 8

; , ! !
i
=
3
5

.r ss
.

3p'

<-

^' ;i::ff;3'i!,li,J
"n,
o"
between variables

Why use it?

relationshiPbetween
Ic stud)-ndidentif-vthepossibte
:le::;::olsen edin two differentsetsofvariables.
ll't &s

C&
iL

Co'se
E{P
<--9

I
Fi

P
A

I
S

P
R

I
p

9
I

+
=

\ o l e : E l i g i b i l i r yr e q u i r e m n r sc h a n p e dr n M a y . m a k i n 8 i {
muh s'mpler io lhe departmenr st; (o make derermi;a
tions.The trnd isstatistically significantbecauserhereare six
or more consecutivepoints declining.

144

Run

@1994
GOAUOPC

+
d

*
-^l

--t-fr?

F
=

it do?

. ;.:::-:=_-::i3iatoconiimahypothesisthattwo
;:-,r--=
r-. ::laii
. :-:-:-=
:'-=--. : i_-:i:l and statistical means to test
relationshlP
--=:=
:: : r::-.:l
. =:rrr=
io a Caus & Effect
: .-:., t:i.ai-::
l. !..ore than iust a
= =c :-- : :-a
g:
:r-^ -- -'-L-and iheeffect
G:==E:E=-T

rr

o data thal you

iO-1S p.ird -nples
nd coDstruct a data sheet
b !lrd

A\erage
Average
Exoerience
of
SssbnRatina
\on a 1-5 scal) TrainingTeam(days)

L2

22o

2
3

3.7
4.3

270
270

40

3.9

625

Theorv: Thereis a oossibleelalionshigbetweenthe

numb of days ol dxperincethe trainingteam has
receivedand the ralinosof coursesessions.

-Etqt[

2, Draw the hoizontal (x axis) and vertical (y axis)

lines o the diagam
. The measurement scaiesgenerally increase as
vou move up the vertical axis and to the right on
he hodzont axis.

+
+
+
F_

tsa

Plot the data on the diagam

. If values are rePeated,circle that point as many
times as aPProPriate
.9

4.5

4.3
4.2

@ 1 .

fr

Interpret the data

. There are many levels of analysis that can be
applied to ScatterDiagram data.Any basicstatistical processcontlol text, like Kaoru Ishikawa's
Gudeto Qualit! Cofittol,describesadditional correlation tests.It is imDortant to note that all o the
examples in this chapter are based on straiShtIine correlations. There ae a number of nonlinear pattems that can be routinely encountered,
e.g- y = er, y = x').These types f analysesare
bvond the scopeof this book.
. The iolloriine five illustrtions show the various
Riterns and meanings that Scatter Diagrams can
i\'e- The exnnple used is the training session
pre riouslv shown. The pattems have
rais:re:
:e's. a-=red ;or illustrative Dul:Doses.Pattem #3
= +:= :.::-:al sanPle.
fp :r 5:ati Diagram doesnot predict ca]used
*e,:T ritionships. It only shor's the strength of
'-:e -i-eironihiD between two--variables.-The
sn:g
the relationship, the geater the likelii11d that chanse in one variable will affect
chanse in anothe; variable.

3.9
3.4
3_7

=r

o
300
Aveage Expeience of Taining Team (in clays)
tnomation prcvicte.J counasv of Hanilton Standatd

ffi
'146 Scarief

A
g
Scter 147

. PositiveCoilation.An increasein y maydepend

on an
increasein x, Sessionratings
arlikelylo incrase
astrainer
expenceIncreass.

150

400

2. Posslbl PosftlveCor.letlon. lf x is increased,

y may
increas
somewhat.
olhervariablesmayb involvdin the
lvl of raiing in addition to
trainorexperienc.
650

#
+
+

scatter

2.O

400

650

.'.t.';:*tt"

.ir"'

!:

4s

55

5i

60

70

65

-rE WldtlvSpacing(;n microns)

4- PossiblNegativCorrelation.Asx is inceased,
y may
clcreasesomewhat.Other
vaiabls,bsidesAainrexpedence,mayalsobe aflect
ing ratings.

.t.tttl!'t9-

--z:e150

2.1

=
3, NoCofielation.Thereisno
clemonstaled
connection
betweentrainrxperienc
and
sessionratings.

Capacitance vs. Line Width

davided tut14

ol A&T

iFb--- th?t th@ is a skong Posihve

]:-- !:aiE
3*...!
dE th o 1?rbl5 in Producing micro-t@*+
re
capacitarce measures a critical Perfor.h!co.
-.:isthat affects ii Positively or negaanftiDnt
md rrrun
Th diasrm shows that Line width/
trel
also sitical
sgocins is sonething to warcrr dosely, Perhps using d Co ntol
C}urt r anodrer n'oe o statistical Drocess (ontrol ISPC) tool

5. NegativeConlation.A d6creasein y maydepndon

an
inceasein x. Sessionratings
are likelyto lall as trainerexpenceincreases,
TBinerxpnnc
49

scsttr

o rss4GoarroPc

Scatter 49

' [111,,"""
From "me" to "we"

Starting feams
Themostcritical taskfor any new teamis to establishits
purpose,process,andmeasures
of teamprogress.Once
the team hds developed the ollowing guidelines and
chrlersspecilicto its purpose,they should berecorded
on a flipchart dnd posted at each tedm meeting for
. Develop a Team Behvior Charter
- Groundrules.Deyelop consensusgroundn_desof
acceptableand unacceptableindividual and
team behavior.
- Decisionmaking.Determine whether decisions
will be made by consensus,majority rule, or
anarchy! Discusswhether there are, or should
be, exceptionsto when the team should not
follow its usual process.
- Cornm fiicatotl.Recognizethe value of listening and constructive feedback, and make the
effo every day, to communicate constuctively!
- Rolesand pdrticipatiori.Discuss how the team
will choose a leader, and generally how the
team processwill be led. The individuals and
team must take responsibility to encourage
equal paticipation.
Valaes.Acknowledge and accept the unique
insight of eachmember of the team.
. Develop a Purpose Chaer
- Establishthe answer to why the team exists.

rIIlI@gEEq

- Bring together the individuals who would

work well together as a team. Determine
whether eachpersonhasthe knowledge,skills,
and influence required to participate effectively on the team.
- The teamshould discusswho its customercare.
I the team has multiple customers, decide
which customershave the highest priority, or
at least how their needswill be balanced.

+
L+r
E
E=
FA

. Deyelop Measuies o Team Progress

- Dl.c1rssarld agree on the desired sitnals, which
ie team can assessboth objectively aJld subjecri. +,ai rtill indiGte theteamismakingFogress.
- ]-*:.-r.s :r,.d alfee on the tvPes of measuresand
:_::.ome: ihai $iIl indicate the team has
:=i.!i :uicess or tailtne- L-=r:: _:1edate rshen the Project should be

fr
a

,rylo'''errtum
\Gnr ieams elF)- terrific strts and then soon fizzle.
TIle rel cin]lgrge G to keep a team ocused on its
DrDo6eand not the lustories o its membe$ and their
iehonships to one another.
. Agre on the Improvemenl Model to Use
- Sta ilard sfeps.Use your organization's standard step-by-step improvement process or
choose from the many published oPtions. (See
the ImDrovement Storvboard in the Problem
Solvin!/Process Imprvement Model section
for one such standard process.)

-+a

z
F=

ffi

Dafa. Gather relevant data to analyze the current situation. Deine what you know, and
what you need to know, but know when to
stop. Learn, as a team,to say vahenyou work
is good enough to proceed to the next step in
the process.
- Deoelopa plan. Use your organization's standard improvement model toprovide the overall structure of a project plan. Estimate times
for eachstep and for the overall project.Monitor and revise the plans as needed.
. Use Povn Methods Based on Both Data and
Knowledge
Data basedmethods.Use tools in this booklet,
e.9., Run Chart, ParetoChart, that reveal pattems within data. These tools often take the
emotion out of discussionsand keep the processmovrng.
- Knouleilge-basedmelrods.Many o the methods
in this booklet, e.9.,Affinity Diagram, Interrelationship Digraph, help to generateand analyze ideas to reveal the important in-formation
within. They help createconsensus,which is
the ideal energy sourcefor a team.
. Manage Team Dynamics
- Useacilittors.A facilitator is someonewho
monitors and helps team members to keep
their interactionspositiveand productive.This
is the stagewhen a acilitator canhelp the team
stay focusedon its purposewhile improving its
working relationship.

152teamffi

+
+
+
+
E
.

Mafiageconlct.Asteamsgrow, so do conflicts.
This is a natural Pocess as communi.ation
becomesmore open.The entire team can learn
techniques for conflict resolution and use the
facilitator as a resource.
- RecrytlizeagTeefient.Ma agir.8 agreement is
oiten as much of an effort as managing diseSeement.Test or agreementoften and wdte
i:1.: :he points of agreementas they occur.
- :-:,- ; :. it partclpafior1.
Eachteam member
- -- :.::.Frallt tale responsibility
for partici
in all discussions.Like::-_::::.:'tentlv
__: ::::. ieam should be constantly
'-'the
1:-.r_::.
dominantmembers
:i ::-k

tG+ot

/E
rdir
t{
JIr

=a5
'l r..
I

,t
rcfe --1

-:.-r.llt end. Both

6{ L :- - .. - -.:
"oficiallv
....
rF<-: :.:.:
- _ ---- -::+r'i.rr':hi
kr':
?_:-i
=a':
:r= :+--:
= --s:-::r+-.-----i

:::-:-.:

.-.lI !.riginal

i-, --t.n.r.-sl<tobedone.

rfe stablished

responsibility

:h.hino.r

rim

for monitodng

I Wedocumentedand trainedpeople,when necessary,in the new process.

TeamGuidelines 153

O We communicated the changes to everyone

affected by them.
O We reviewed our own team's accomplishmentc for re9o improvement,
d We celebrated the eorts o the team with a
lunclL newsletter article, special presentafion to
the company, or other expressiono celebration.
O We feel proud of our contribution and accomplishments, our new capabfities, and our
newly defined relationshipswith coworkers.
Co nduc,i n q Ee ctive Meet ings
Preparation:
. Decide on the purpose of the meeting
. Develop a meeting plan (who, what, where,
when, how many)
. Identify the meeting leader
. Prepareand distribute the agenda
. Setup the meeting area
BeginninS:
. Start on time
. Introduce the meeting leader
. Allow team membersto introduce themselves
. Ask or a volunteer timekeeper
. Ask for a volunfeer recorder
. Review, change,order the agenda
. Establishtime limits
. Review pdor meeting action items
'154 TamGuidelines

+
+
+
+
r+
.+
+
+
+
+
F+

+
+

Meeting Etiquettel
. Raise your hand and be rccognized beore
speaking
. Be brief and to the point
. Make your point calmly
. Keep an open mind
. Listen rvithout bias
. Understand what is said
. {i oid side conversations
. ?r*r
.

!1.,i.1

^rhar ^nini^nc
#lsn:l

,dn.|,<

. C..:. .rpredto do what's good for

--

L:r:?t\'

j,,:-:
:.i:

F".i-g
. l"FL-: ion iisls (Ir'ho, what, when, how)
r 3r-l*ize
tire meeting *ith the goup
. L<{atlh t} dafe and time for a ollow-up meeting
. Ei_ahte the meeting
. End on time
. Clen the meeting atea
Next Steps:
. Prepare and distdbute the meeting activify report
. Follow up on action items
. Go to "Preparation"

TemGuidlineg t55

^.

Ilee Diagram
Mappin.g the tasks
ror tmptemenaflon

,_E
1_E
*E

Why use it?

To break any broad goal, graphically, into increasing
levels of detailed actionsthatmust or could be done to
achievethe statedgoals.
Wha does it do?
. Encourages team members to expand their thinking when creating solutions. Simultaneously, this
tool keeps everyone linked to the overall goals
and subgoalsof a task
. Allows all participants, (and reviewers outside
the team), to check all of the logical links and
completeness at every level of plan detail
. Moves the planning team from theory to the real
world
. Reveals the reai level o complexity involved in
the achievement of any goal" making potentially
overwhelming projects manageable,as well as
uncovering unknown complexity

*
tlow do t .to ia? )\l

+
5

+
.

1. Choosethe Tree Diagam goal stalement

Goal:Increaseworkplacesuggestions

IIU

+
+

' TyPical sources:

- The root cause/driver identrfied in an Interrelationship Di$aph (ID)
- An Affinity Diagram with the headersdsmjo
subgoals
- Any assignmentgiven to an individual or team
. \\-hen used in conjunction with other manager-ri .ni DLanningtools, the most typical source
L<--r1
roct.aL. driver identified in the ID.
r?| Rardlr of the source, work had to crete---tfuough consensus-a clear, action-oriented staterent.
Assmble the right team
. The team should consist of action plannerc with
detailed knowledge of the goal topic. The team
should taketheTree only to the level of detail that
the team's knowledge will allow. Be prepared to
hand further details to others.
. 4-6 people is the ideal goup size but the Tree
Diagram is appropiate for larger groups as long
as the ideas are visble and the sessionis well
facilitated.
Geneate the maiorTree headings, which are the
maio subgoals to purgue
. The simplest method for creating the highest, or
first level o detail, is to brainstorm the major task
areas. These arc the major "means" by which the
goal statement will be achieved.

@1994GoAUoPc

Tree iS7

. To encourage creativity, itis often helpful to do an

"Action Affinity" on the goal statement. Brainstorm action statements and sort into grouPings,
but spend less time than usual refining the header
cards. Use the header cards as the Tree's firstIevel subgoals.
Goal

Increaseworkplace
suggestaons

Means

:\-i

Createa workable
process

Createcapability

Measure
results

Break each maior heading into geater detail

. Working from the goal statement and fftst-level
detail, placed either to the extreme let, dght or
top o the work surface, ask of each first{evel
item:
-_",_-_-:r
=e.:s : t addred to achieve the soal
L';iquestion for each successivelevel o

. ]=E i breildown of each level when there are

ringnable tasks or the team reachesthe limit to its
..-n erpertise. Most Trees are broken out to the
third level of detail (not counting the overall goal
statement asa level). However, some subgoals are
Iust simple thn others and don'l require s
much breakdown.

Providerecognition
tip Use Post-it" notes to create the levels of detail.
Draw lines only when the Tree is finished. This
allows it to stav flexible until the Drocessis finished. The Tecan be oriented ftom left to right,
right to left, o! top down.
|ip Keep the first level of detail broad, and avoid
"If
iumping to the lowest level of task. Remember:
you start with what you already know, you'll
end up where you've already been."

158 rce

#
+
E

ra oq4uoaLoP-

Tr".

15g

5. Review the compleled Tee Diagram ot logical

flor and completeness
. At each level of detail, ask "ls therc something
obvious that we have forgotten?"
. As the Treebreaksdowninto greaterdetail (from
generalto specific)ask, "If I want to accomplish
theseresults,do I really need to do thesetasks?"
. As the Tree builds into broader goals (rom the
specific to the general) ask, "Will these actions
acfually lead to theseresults?"
. Draw the lines connectingthe tasks.
r'p The Tree Diagram is a great communication
tool. It canbe used to get input from thoseoutside
the team. The team's final fask is to consider
proposedchanges,additions or deletions,and to
modiy the Tree as appropdate.
Variationt
TheProcessDecisionProgram Chat (PDPC)is a valuable tool for improving implementation through contingency planning. The PDPC,basedon the Tree Diagram, involves a few simple steps.

+
+

Choos the mo6l eective counterme,a-sr

build them into a revised plar

+
E+

+
+

--!-G-

X = DficlMmpos.sible
O = Selecled

br

1. Assemble a team closestto the implementation

2, Determine proposed implementation step6
. Li5t4.l0broadsiepsandplcethemin sequence
in the irst Tree level.
3. Ban.h likely problems o each step
. Ask "What could go wrong?"
4. Branch possible and reasonable responses off
each likely problem

60 Tree

Tree 161

PDPC(TreeVaiation)

Awardino unest c-tedFinancialAidl

+
ts=t

,::i.-r

BusinessPlanningInteraction

4!
41
Irrrf-!
|-1,
research- exislinodala\
. Additionlquantitvdata
Addllionalqualitiivedala

. Sel. quickordecislonon

. Plan for vaiety on rtes ol

rrrrt'", into-.tlon "*i"t" Uutl" norg-l*,

lnlomation pravided courtesy of St. John Fisher Cotteg
Note: The PDPC surfaced a lack of accurate information as a
major problem. By anti.ipating
this and filling the mosi .dti.al
information Baps, ihe budget can be more a.curate.

fr

ffi

re 163

PDPC(TreeVariation)

Alrarding Unsticted FinancialAidr)

,:::"_(
BusinessPlanningInteraction

. Incrcasenrollmenl \
esearch- xistinadata\
. Addnionalquantitaivdala
qualitalivedata .
Additional
:

rr:rttrer lntormationexistsbut is norsi"-wn

lnfomationprovidedcounesyof 5t. John FsherCottege
Note The PDPC surfaced a lack of accuat information as a
major prcblem. By anti.ipating this and filting the lost .riticai
information gaps,the budget can be more acuate.

re

Tree 63

Tree f

tmprovel
BusinessPlanninglnteraction
(continued)

H
+r

o
a

64 Tee