You are on page 1of 39

1. Introduction...........................................................................................................................................................

2
1.1. Quality............................................................................................................................................................2
1.1.1. Quality levels...........................................................................................................................................2
1.1.2. Cost of poor quality.................................................................................................................................2
1.2. Six Sigma.......................................................................................................................................................3
1.2.1. Trained teams..........................................................................................................................................3
1.2.2. Sigma levels............................................................................................................................................3
1.2.3. Focusing of te !s...................................................................................................................................3
1.2.". Te five pases........................................................................................................................................3
1.2.#. $FSS.......................................................................................................................................................3
1.2.%. Customer&centric metrics........................................................................................................................3
1.2.'. $FCI........................................................................................................................................................"
1.2.(. Six Sigma at te customer......................................................................................................................."
1.2.). *ean Six Sigma......................................................................................................................................."
2. $efine...................................................................................................................................................................."
2.1. Step +, ldentify pro-ect CTQs........................................................................................................................#
2.1.1. .oice of te Customer /.0C1.................................................................................................................#
2.1.2. $etermine 2riority CTQs........................................................................................................................#
2.1.3. 2roduct32rocess $rill&$o4n Tree...........................................................................................................%
2.2. Step 5, Carter...............................................................................................................................................%
2.2.1. 06-ectives................................................................................................................................................%
2.2.2. Five elements of a carter.......................................................................................................................%
2.2.2.1. 5usiness Case...................................................................................................................................%
2.2.2.2. 2ro6lem and 7oal Statement............................................................................................................%
2.2.2.3. 2ro-ect Scope....................................................................................................................................'
2.2.2.". 8ilestones........................................................................................................................................'
2.2.2.#. 9oles.................................................................................................................................................'
2.3. Step C, 2rocess 8ap.......................................................................................................................................'
3. 8easure.................................................................................................................................................................(
3.1. Step 1, Select CTQ caracteristics.................................................................................................................(
3.1.1. Tools........................................................................................................................................................(
3.1.2. QF$ /Quality Function $eployment1.....................................................................................................(
3.1.2.1. :ouse of Quality..............................................................................................................................(
3.1.2.2. QF$ flo4do4n.................................................................................................................................)
3.1.2.3. ;se of QF$......................................................................................................................................)
3.1.3. F8<+......................................................................................................................................................)
3.1.". 2rocess 8ap..........................................................................................................................................1=
3.1.#. Failure 8odes and <ffects +nalysis, severity> occurrence> detection...................................................12
3.1.%. Types of $ata........................................................................................................................................12
3.2. Step 2, $efine performances standards........................................................................................................12
3.3. Step 3, 8easurement systems analysis........................................................................................................12
3.3.1. Ta?t time................................................................................................................................................12
3.3.2. 06servation Seet..................................................................................................................................13
3.3.3. 7age.......................................................................................................................................................13
3.3.". 8easurement Systems +nalysis Cec?list............................................................................................1"
3.3.#. Test&9etest Study..................................................................................................................................1"
3.3.%. 7age 9@9.............................................................................................................................................1"
". +nalyse................................................................................................................................................................1#
".1. Step " & <sta6lis process capa6ility............................................................................................................1#
".1.1. Continuous data.....................................................................................................................................1%
".1.1.1. Some definitions.............................................................................................................................1%
".1.1.2. Tools...............................................................................................................................................1%
".1.1.3. normality testing.............................................................................................................................1'
".1.1.". Aormal distri6ution........................................................................................................................1(
".1.2. $iscrete data..........................................................................................................................................1)
".1.2.1. $efinitions......................................................................................................................................1)
".1.2.2. Bield...............................................................................................................................................1)
".1.3. 2rocess capa6ility..................................................................................................................................1)
".1.3.1. Special .ersus Common Cause .ariation.....................................................................................2=
".1.3.2. 9ational Su6grouping.....................................................................................................................2=
".1.3.3. C sift.............................................................................................................................................2=
".1.3.". Csort term vs Clong term..............................................................................................................2=
".2. Step # & $efine performance o6-ectives.......................................................................................................2=
".3. Step % & Identify variation standard..............................................................................................................21
".3.1. Tools......................................................................................................................................................21
".3.1.1. Fis6one /cause and effect diagram1..............................................................................................21
".3.1.2. 2areto cart.....................................................................................................................................22
".3.1.3. 2rocess 8ap +nalysis....................................................................................................................22
".3.1.". F8<+.............................................................................................................................................23
".3.2. :ypotesis testing.................................................................................................................................23
".3.2.1. :ypotesis Testing, Continuous BD $iscrete !.............................................................................2%
#. Improve...............................................................................................................................................................2(
#.1. Steps ' & Screen potential causes..................................................................................................................2)
#.2. Steps ( & $etermine relationsip..................................................................................................................2)
#.2.1. Improvement stategy.............................................................................................................................2)
#.2.2. Tools......................................................................................................................................................2)
#.2.3. 2rocess of experimentation...................................................................................................................2)
#.2.". $0<, $esign 0f <xperiment................................................................................................................3=
#.2.".1. +nalyEing a full factorial & replicated design.................................................................................3=
#.2.".1.1. plot te ra4 data......................................................................................................................3=
#.2.".1.2. plot te residuals......................................................................................................................3=
#.2.".1.3. examine factor effects.............................................................................................................31
#.2.".1.". confirm impressions 4it statistical procedures.....................................................................31
#.2.".1.#. summariEe conclusions............................................................................................................31
#.2.".2. Compute 2rediction 8odel............................................................................................................31
#.2.".3. 9educing siEe of experiments.........................................................................................................31
#.3. Step ) & <sta6lis operating tolerance..........................................................................................................33
#.3.1. Tolerancing............................................................................................................................................33
#.3.2. simulation..............................................................................................................................................33
%. Control.................................................................................................................................................................3"
%.1. Step 1= & define and validate measurement system on !s in +ctual +pplication........................................3#
%.2. Step 11 & determine process capa6ility.........................................................................................................3#
%.3. Step 12 & Implement process control............................................................................................................3#
%.3.1. ris? management....................................................................................................................................3#
%.3.2. mista?e proofing /po?a yo?e1................................................................................................................3#
%.3.3. control plan............................................................................................................................................3'
%.3.". control carts.........................................................................................................................................3'
1. Introduction
1.1. Quality
1.1.1. Quality levels
3F capa6ility & istorical standard
"F capa6ility & current standard
%F capa6ility & ne4 standard
1.1.2. Cost of poor quality
Tangi6le Costs
Inspection
Scrap
9e4or?
Garranty
Intangi6le Costs
<xpediting
*ost Customers
*onger Cycles
1.2. Six Sigma
1.2.1. Trained teams
Several roles ave 6een defined in te Six Sigma strategy,
Campions, 6usiness leaders> provide resources and support implementation
8aster 5lac? 5elts, experts and culture&cangers> train and mentor 5lac? 5elts37reen 5elts
5lac? 5elts, lead Six Sigma pro-ect teams
7reen 5elts, carry out Six Sigma pro-ects related to teir -o6s
1.2.2. Sigma levels
2F, logic and intuition are enoug> no special tool is required
3&"F, quality programs used 6y ;S companies for years /seven tools1
#F, $8+IC> process caracteriEation and optimiEation
%F, $FSS /$8+$0.1> design for Six Sigma
1.2.3. Focusing of the Xs
! are te inputs of te process> B te output
Traditionally> companies focused on te B> Six Sigma forces us to understand te relationsip 6et4een te !s
and B to find te vital !s and control tem.
1.2.. The five phases
$efine & Customer expectations of te processH
8easure & Gat is te frequency of defectsH
+nalyEe & Gy> 4en and 4ere do defects occurH
Improve & :o4 can 4e fix te processH
Control & :o4 can 4e ma?e te process stay fixedH
1.2.!. "FSS
$esign
1 Identify product3process performance and relia6ility CTQs and set quality goals
8easure
2 2erform CTQ flo4do4n to su6systems and components
3 8easurement system analysis3capa6ility
+nalyEe
" $evelop conceptual designs /6encmar?ing> tradeoff analysis1
# Statistical analysis of any relevant data to assess capa6ility of conceptual designs
% 5uild scorecard 4it initial product3process performance and relia6ility estimates
' $evelop ris? assessment
$esign
( 7enerate and verify system and su6system models> allocations and transfer functions, lo4 Cst> lac? of transfer
function> un?no4n process capa6ility
) Capa6ility flo4&up for all su6systems and gap identification
0ptimiEe
1= 0ptimiEe design, analysis of variance drivers> ro6ustness> error proofing
11 7enerate purcasing and manufacturing specification and verify measurement system on !s
.erify
12 Statistically confirm tat product process matces predictions
13 $evelop manufacturing and supplier control plans
1" $ocument and transition
1.2.#. Customer$centric metrics
In general> te customer sees te performance of its o4n process> not te performance of 7<Is one 4ic is -ust a
part of it. +n example,
customer removes
defective component
from its macine and
sends it to 7<
7< repairs te
component and
sends it to
customer
customer
replaces repaired
component
7< vie4
customer vie4
Ge must understand te customerJs perspective and expectations regardless of o4 muc of it 4e control.
1.2.%. "FC&
$esign For Customer Impact
$elivery span, te customers feel te variance> not te mean.
1.2.'. Si( Sigma at the customer
2ro-ects done 6y 7< 55375s or customer 55375s> trained and3or mentored 6y 7< 55375s
+ddress customer Bs
+ddress customer or 7< !s
customer GE
Ys K te customer 7<Is $8+IC3$FSS pro-ects
Xs K te customer K te customer
1.2.). *ean Si( Sigma
:unt for 4aste
2. Define
06-ectives,
define te product or process to 6e improved
identify customers and translate te customer needs into CTQs
o6tain formal pro-ect approval
tools
.0C tools and data tecniques
team carter
ig&level 2rocess 8ap
2ro-ect selection
+ good pro-ect sould
6e clearly 6ound 4it defined goals
6e aligned 4it te 6usiness goals and initiatives
6e felt 6y te customer
4or? 4it oter pro-ects for com6ined effects /may 6e driven 6y te funnel pro-ect of a 551
so4 improvement tat is locally actiona6le
relay to daily -o6
Sources of pro-ect ideas
Quality Function $eployment /QF$1
Customer $as6oards
Surveys and Scorecards
+ctive 5eta Temes
0ter 2ro-ects +vaila6le for *everage
5rainstorming
+nalysis of Critical 2rocesses
Six Sigma Quality 2ro-ect Trac?ing $ata6ase
customer removes
defective component
from its macine and
sends it to 7<
7< repairs te
component and
sends it to customer
customer replaces
repaired component
$iscussions 4it Customer
Financial +nalysis
Internal 2ro6lems
2.1. Step A: ldentify project CTQs
C02IS, customer centric vie4 of te process
Customer, Gomever receives te output of your process /may6e internal or external1
0utput, Te material or data tat results from te operation of a process.
2rocess, Te activities you must perform to satisfy your customerJs requirements.
Input, Te material or data tat a process does someting to or 4it.
Supplier, Gomever provides te input to your process.
2.1.1. +oice of the Customer ,+-C.
.0C, 4at is critical to te quality of te process according to te customer.
pros cons
Surveys *o4er Cost approac
2one response rate '= & )=L
8ail surveys require least amount of
trained resources for execution
Can produce faster results
8ail surveys & can get incomplete results>
s?ipped questions> unclear understanding
8ail surveys & 2=&3=L response rate
2one surveys & intervie4er as influential
role> can lead intervie4ee producing
undesira6le results
Focus
7roups
7roup interaction generates
information
8ore in&dept responses
<xcellent for getting CTQ definitions
Can cover more complex questions or
qualitative data
*earningIs only apply to tose as?ed & difficult
to generaliEe
$ata collected typically qualitative vs.
quantitative
Can generate too muc anecdotal information
Intervie4s Can tac?le complex questions and a
4ide range of information
+llo4s use of visual aids
7ood coice 4en people 4onJt
respond 4illingly and3or accurately 6y
pone3mail
*ong Cycle Time to complete
9equires trained> experienced intervie4ers
Customer Information Issues
9eal Aeeds vs. Stated Aeeds, !erox, Focus on copiers or documentsH
2erceived Aeeds, + :ersey 6ar or 7odiva cocolatesH
Intended vs. +ctual ;sage, Is a scre4driver also a ammerH
Internal Customers, Turf 4ars and Mnot invented ere.N
<ffectiveness vs. <fficiency Aeeds, MBou 4ant it rigt or you 4ant it fastHN
Te +ffinity $iagram and Structure Tree can elp to organiEe and translate .0C into customer needs.
2.1.2. "etermine /riority CTQs
Te specific needs statement provides te foundation for te " elements of te CTQ /0utput Caracteristic>
8easure> Target3Aominal .alue> Specification *imits1.
0nce CTQs are identified> te team sould re&evaluate teir carter. Gill addressing te CTQs impact te
issue/s1 identified in te 2ro6lem StatementH $oes te Scope of te 2ro-ect allo4 for focus on te top 1 or 2
CTQsH
+ successful pro-ect is related to one or more of te four vital customer CTQs,
Customer 9esponsiveness3Communication
8ar?et 2lace Competitiveness O 2roduct32rice3.alue
0n&Time> +ccurate> and Complete Customer $elivera6les
2roduct3Service Tecnical 2erformance
2.1.3. /roduct0/rocess "rill$"o1n Tree
Te 2rocess32roduct $rill&$o4n Tree is a 4ay to integrate CTQs 4it 6usiness strategy.
2.2. Step B: Charter
2.2.1. -23ectives
Clarifies 4at is expected of te team
Peeps te team focused
Peeps te team aligned 4it organiEational priorities
Transfers te pro-ect from te campion to te improvement team
2.2.2. Five elements of a charter
2.2.2.1. Business Case
<xplanation of 4y to do te pro-ect
Gy is te pro-ect 4ort doingH
Gy is it important to do it no4H
Gat are te consequences of A0T doing te pro-ectH
Gat activities ave iger or equal priorityH
:o4 does it fit 4it 6usiness initiatives and targetH
2.2.2.2. Problem and Goal Statement
$escription of te pro6lem 3 opportunity or o6-ective in clear> concise> measura6le terms
2ro6lem Statement 2urpose, $escri6es 4at is 4rong
Gat is 4rong or not meeting our customerJs needsH
Gen and 4ere do te pro6lems occurH
:o4 6ig is te pro6lemH
Gat is te impact of te pro6lemH
Pey Considerations32otential 2itfalls
Is te pro6lem 6ased on o6servation /fact1 or assumption /guess1H
$oes te pro6lem statement pre-udge a root causeH
Can data 6e collected 6y te team to verify and analyEe te pro6lemH
Is te pro6lem statement too narro4ly or 6roadly definedH
Is a solution included or implied in te statementH
Gould customers 6e appy if tey ?ne4 4e 4ere 4or?ing on tisH
7oal Statement 2urpose, $efines te teamJs improvement o6-ective
$efines te improvement te team is see?ing to accomplis.
Starts 4it a ver6 /reduce> eliminate> control> increase1.
Tends to start 6roadly O eventually sould include measura6le target and completion date.
$oes not assign 6lame> presume cause> or prescri6e solutionQ
S8+9T
Specific
8easura6le
+ttaina6le
9elevant
Time 5ound
2.2.2.3. Project Scope
2rocess dimensions> availa6le resources
Gat process 4ill te team focus onH
Gat are te 6oundaries of te process 4e are to improveH
Gat resources are availa6le to te teamH
Gat /if anyting1 is out&of&6ounds for te teamH
Gat /if any1 constraints must te team 4or? underH
Gat is te time commitment expected of team mem6ersH
2.2.2.4. Milestones
Pey steps and dates to acieve goal
+ preliminary> ig&level pro-ect plan 4it dates sould 6e,
Tied to pases of $8+IC process
+ggressive
9ealistic
2.2.2.. !oles
2eople> expectations> responsi6ilities
:o4 do you 4ant te campion to 4or? 4it te teamH
Is te teamJs role to implement or recommendH
Gen must te team go to te campion for approvalH Gat autority does te team ave to act
independentlyH
Gat and o4 do you 4ant to inform te campion a6out te teamJs progressH
Gat is te role of te team leader /5lac?37reen 5elt1 and te team coac /8aster 5lac? 5elt1H
+re te rigt mem6ers on te teamH FunctionallyH :ierarcicallyH
2.. Step C: !rocess "ap
;se C02IS, start 4it te customer and 4or? 6ac?4ards.
Tis is a ig level process map,
Aame te process.
Identify te outputs> customers> suppliers> @ inputs.
Identify customer requirements for primary outputs.
Identify process steps.
Some C+2 tools can also 6e used.
3. Measure
06-ectives,
select te measura6le CTQs to improve
map te process
determine te specification limits for B
define te measure system and ensure te tat it is adequate to measure B troug te use of a 7age 9@9
collect te data
.1. Step 1: Select CTQ characteristics
3.1.1. Tools
Te follo4ing tools are usa6le to select te CTQs on 4ic to focus,
QF$
Fis6one
2rocess 8ap
2areto Cart
F8<+
+ deplacer
2recision is te consistency of a process as measured 6y te standard deviation.
+ccuracy is te a6ility to 6e on target as measured 6y te mean.
2rocess caracteriEation descri6es te distri6ution of te data.
3.1.2. QF" ,Quality Function "eployment.
QF$F is a structured metodology to identify and translate customer needs and 4ants into tecnical
requirements and measura6le features and caracteristics.
It is used to identify CTQs.
3.1.2.1. "ouse of #ualit$
step 1, find 4at te user 4ants /use te .0C tools descri6ed 6efore1 and 4rite tis list in te left and
column /te R4atIsR1
step 2, rate customer importance of eac needs34ants /use factors in S1&#T or S1&1=T range1
step 3, list parameters tat ave an effect on te process3product /te Ro4Is to satisfyR te R4atIsR1
step ", evaluate te relationsip 6et4een a factor and a customer need,
dou6le circle U strong U )
circle U medium U 3
triangle U 4ea? U 1
step #, define te target direction,
up4ard arro4 U more is 6etter
do4n4ard arro4 U less is 6etter
circle U a specified amount
step %, indicate te target values> quantities and units> of te factors /tey sould come from te customer
requirements1
step ', for eac column> sum up te products / customer importance V relationsip 1> tis gives te tecnical
importance of eac parameter
step (, fill te correlation matrix 6et4een te parameters
dou6le circle U strongly positive
circle U positive
cross U negative
dou6le cross U strongly negative
11 customer needs
21 customer importance
4

a
t

o
4
31 process3product
parameters
"1 relationsip
strong )
medium #
4ea? 1
#1 target direction
more is 6etter
less is 6etter
a specified amount
%1 target values
'1 compute tecnical
importance
(1 correlation matrix
strongly positive
positive
negative
strongly negative
+nalyEe te ouse of quality,
6lan? ro4, 4e donIt ?no4 o4 to satisfy a customer need
6lan? column, te parameter as no impact on te customer needs
no design constraint in o4Is
resolve negative correlations
finaliEe target values
4at parameters sould 6e deployed to te next level of flo4do4n
3.1.2.2. #%D flo&do&n
2roduct,
customer requirements W ?ey functional requirements
functional requirements W ?ey part caracteristics
part caracteristics W ?ey manufacturing processes
manufacturing processes W ?ey process varia6les
Service,
customer CTQs W ?ey pro-ect delivera6les
?ey pro-ect delivera6les W ?ey process steps
?ey process steps W ?ey tas?s
3.1.2.3. 'se of #%D
2itfalls,
$o not use QF$ for every pro-ect.
Set te rigt granularity.
Inadequate priorities.
*ac? of team4or?.
Too muc cart focus, te cart is te mean> not te o6-ective
R:urry up and get doneR.
QF$ must 6e regularly revie4ed and updated.
3.1.3. F456
2reparation
1. Select 2rocess Team
2. $evelop 2rocess 8ap and Identify 2rocess Steps
3. *ist Pey 2rocess 0utputs to Satisfy Internal and <xternal Customer 9equirements
". *ist Pey 2rocess Inputs for <ac 2rocess Step
#. $efine 8atrix 9elating 2roduct 0utputs to 2rocess .aria6les
%. 9an? Inputs +ccording to Importance
F8<+ 2rocess
'. *ist Gays 2rocess Inputs Can .ary /Causes1 and Identify +ssociated Failure 8odes and <ffects
(. *ist 0ter Causes /Sources of .aria6ility1 and +ssociated F8@<s
). +ssign Severity> 0ccurrence and $etection 9ating to eac Cause
1=. Calculate 9is? 2riority Aum6er /92A1 for eac 2otential Failure 8ode Scenario
Improvements
11. $etermine 9ecommended +ctions to 9educe 92As
12. <sta6lis Timeframes for Corrective +ctions
13. Create MGaterfallN 7rap to Forecast 9is? 9eductions
1". Ta?e +ppropriate +ctions
1#. 9e&calculate +ll 92As
1%. 2ut Controls into 2lace
9is? 9atings, Scale, 1 /5est1 to 1= /Gorst1
Severity /S<.1
:o4 significant is te impact of te <ffect to te customer /internal or external1H
0ccurrence /0CC1
:o4 li?ely is te Cause of te Failure 8ode to occurH
$etection /$<T1
:o4 li?ely 4ill te current system detect te Cause or Failure 8ode if it occursH
9is? 2riority Aum6er,
+ numerical calculation of te relative ris? of a particular Failure 8ode.
92A U S<. x 0CC x $<T.
Tis num6er is used to place priority on 4ic items need additional quality planning.
Te F8<+ ta6le contains te follo4ing columns,
2rocess Step 3 2art Aum6er
2otential Failure 8ode, *ists Failure 8odes for eac 2rocess Step
2otential Failure <ffect, *ists te <ffects of eac Failure 8ode
S<., 9ates te Severity of te <ffect to te Customer on a 1 to 1= Scale
2otential Cause, *ists te Causes for eac Failure 8ode, <ac Cause is +ssociated 4it a 2rocess Input 0ut
of Spec
0CC, 9ates :o4 0ften a 2articular Cause or Failure 8ode 0ccurs, 1 U Aot 0ften> 1= U .ery 0ften
Current Controls, $ocuments :o4 te Cause is Currently 5eing Controlled in te 2rocess
$<T, 9ates :o4 Gell te Cause or te Failure 8ode can 6e $etected, 1 U $etect <very Time> 1= U Cannot
$etect
92A, 9is? 2riority Aum6er /92A1 is, SevV0ccV$et
+ctions 9ecommended, $ocuments +ctions 9ecommended 5ased on 92A 2areto
9esp., $esignates Go is 9esponsi6le for +ction and 2ro-ected Completion $ata
3.1.. /rocess 4ap
2rocess map, a grapical representation of steps> events> operations and relationsips of resources 4itin a
process.
;se C02IS, start 4it te customer and 4or? 6ac?4ards.
<lements of a process
Control, Te material or data tat is used to tell a process 4at it can or sould do next.
8ecanism, Te resources /people> macines> etc.1 tat come to 6ear on a process to cange te input to an
output.
2rocess 5oundary, Te limits of te process> usually identified 6y te inputs> outputs and external controls
tat separate 4at is 4itin te process from its environment.
5enefits of 2rocess 8apping
Can reveal unnecessary> complex> and redundant steps in a process. Tis ma?es it possi6le to simplify and
trou6lesoot.
Can compare actual processes against te ideal. Bou can see 4at 4ent 4rong and 4ere.
Can identify steps 4ere additional data can 6e collected.
8ost of te time> tere are tree maps,
4at 4e tin? may 6e appening
4at is actually appening
4at 4e 4ant to 6e appening
IS0 )=== sym6ols
prepare,
<sta6lis te process 6oundaries
06serve te process in operation
*ist te outputs> customers> and teir ?ey requirements
*ist te inputs> suppliers> and your ?ey requirements
6uilding te map
determine te scope, 4at level of detail 4e 4antH
determine te steps in te process, no order> no priority
arrange te steps in order
assign a sym6ol
test te flo4
are te process steps identified correctlyH
is every feed6ac? loop closedH
does every arro4 ave a 6eginning and ending pointH
is tere more tan one arro4 from an activity 6oxH 2eraps it sould 6e a diamond
are all te steps coveredH
validate te map
4al? te process again
as? te questions, Gat appens ifXH Gat could go 4rongH GoXH :o4XH GenXH
update map
evaluating a 2rocess 8ap
does eac step add valueH
are controls and measurement criteria in placeH
are M9eJsN occurringH 9e4or? 9evise 9epeat 9evie4
is te step necessaryH
3.1.!. Failure 4odes and 5ffects 6nalysis7 severity8 occurrence8 detection
See ere.
3.1.#. Types of "ata
continuous, measured on a continuum or scale
discrete
count, counted discreetly
ordered categories, ran?ings or ratings
6inary, classified in one of t4o categories
2refer continuous to discrete, continuous data often requires a smaller sample siEe tan discrete data. For
example> for delivery time> it is 6etter to consider te actual times deviated from target rater tan te num6er of
late deliveries. Count and> may6e less easily> ordered categories> can 6e transformed in continuous data.
.2. Step 2: #efine performances standards
+ 2erformance Standard is te requirement/s1 or specification/s1 imposed 6y te customer on a specific CTQ.
Te goal of a performance standard is to translate te customer need into a measura6le caracteristic.
2roduct32rocess Caracteristic
0perational $efinition
Target
Specification *imits
$efect $efinition
0perational $efinition /S02> Standard 0perating 2rocedure1, clear definition of 4at to measure /i.e. CTQ1 and
o4 to measure it.
Its purpose is,
9emove am6iguity so tat everyone as te same understanding
2rovides a clear 4ay to measure te caracteristic
Identifies 4at to measure
Identifies o4 to measure it
8a?es sure tat no matter 4o does te measuring> te results are essentially te same
8ust 6e useful to 6ot te company and te customer
$efect, te definition of a defect must 6e provided.
Try to get a continuous measurement instead of a discrete one.
.. Step : "easurement systems analysis
3.3.1. Ta9t time
Ta?t Time, Te frequency at 4ic eac unit sould 6e completed in order to meet customer demand. Ta?t Time
esta6lises te necessary rytm of te process.
$emand Customer
Time 2rocessing +vaila6le
Time Ta?t =
Cycle time /touc time1 U manual 4or?ing time for one cycle of te process
8easure te cycle time for eac step of te production flo4.
If te cycle time is greater tat te ta?t time> 4e cannot meet customer demand. If te cycle time is smaller tan
te ta?t time> te difference is te 4aiting time> tis is an opportunity for improvement 6y redistri6uting te
4aiting time on oter production steps.
3.3.2. -2servation Sheet
Te o6servation seet is used to compute cycle time.
Te o6servation seet records,
Set&up Time, 8easure te set&up time and divide 6y te num6er of parts made 6et4een setups to get te set&
up time per part.
8anual Tas? Time, 8easure and enter te ands&on time /uman 4or?1 it ta?es for te operator to perform
te operation 4it te macine /te process1. Gal?ing time is not included ere.
+uto&9un Time, 8easure and enter te time from 4en te macine is s4itced on and it processes te part
until te time 4en te macine returns to its original ome position.
Travel Time, 8easure and enter te time it ta?es te operator to move to te next station and pic?&up or put&
do4n parts or tools. *eave tis space 6lan? if tere is no travel time.
Gaste3Aon&.alue +dded +ctivities, 9ecord any non&value added activity tat you o6serve and any
recommendations for improvement.
3.3.3. :age
In fact> 4e su6mit te output of a first process /te manufacturing process1 to a second process /te measurement
process1. To address actual process varia6ility> te variation due to te measurement system must first 6e
identified and separated from tat of te manufacturing process.
2 2 2
observed measuremnt actual
= +
accuracy, te differences 6et4een o6served average measurement and a standard
repeata6ility, variation 4en one person repeatedly measures te same unit 4it te same measuring
equipment
reproduci6ility, variation 4en t4o or more people measure te same unit 4it te same measuring
equipment
sta6ility, variation o6tained 4en te same person measures te same unit 4it te same equipment ever an
extended period of time
linearity, te consistency of te measurement across te entire range of te measurement system
Te gage sould ave,
precision, te gage sould 6e a6le to resolve te tolerance into approximately ten levels /Rte num6er of
significant digitsR1
accuracy, te gage noise must 6e less tan te process noise /Rte measured value is te real oneR1
<lements of 8easurement System .aria6ility
8en and Gomen
8etod
8aterial
8easurement
8acine
<nvironment
+ Fis6one $iagram may 6e used to identify sources of variation in te measurement system.
3.3.. 4easurement Systems 6nalysis Chec9list
1. Gat is te measurement procedure usedH
2. 5riefly descri6e te measurement procedure. Gat standards applyH +re usedH
3. Gat is te RprecisionR /measurement error1 of te systemH
". :o4 as te precision 6een determinedH
#. Gat does te gage /measurement system1 Supplier state is te deviceJs,
U $iscrimination /9esolution1H
U +ccuracy /5ias1H
U 2recision /8easurement <rror1H
%. $o you ave results of a,
U Test&9etest StudyH /$etermines 8easurement <rror or Mlac?&of&precisionN1
U 7age 9@9 StudyH /+llocates te error 6et4een device and operator/s11
If so> 4at are teyH '. +re different measurement systems /gages> scales> etc.1 used to gater te
same dataH Identify 4ic data comes from 4ic device.
3.3.!. Test$;etest Study
5est practice, do a test&retest study 6efore te gage 9@9 study to get a quic? loo? at te situation.
9epeatedly measure te same item 4it te same conditions> operator> device> and MlocationN on item and
completely mount and dismount item for eac measurement /MexerciseN gage troug full range of normal use.1
2erform 2= or more measures /1=&1# may 6e 0P if te measurements are expensive to perform1.
dra4 a plot cart
device resolution,
compute measurement it sould 6e less tan 131= of te tolerance
device accuracy, if 4e ?no4 te true value of te test unit
True.alue inacurracy =
t measuremen
X
3.3.#. :age ;<;
Collecting te $ata
In order to get and estimate variation in te M9ealN 8easurement System,
Follo4 actual process.
;se te people tat usually measure.
Follo4 te planning for te -o6.
2erform te study in te usual environment.
;se te gages used for te -o6.
7age 9eproduci6ility /+ppraiser .ariation1 @ 9epeata6ility /<quipment .ariation1

2
equipment Y
2
appraiser U
2
total/9@91
<quipment .ariation, /Sources of variation from 4itin te process1 te variation introduced into te
measurement process from 4itin one or more elements of te measurement process /4itin operator variation>
4itin gage variation> 4itin part variation> 4itin metod variation1
+ppraiser .ariation, /Source of variation from across te process1 te variation introduced into te measurement
process 6y effects going across te measurement process /different appraisers> different part configurations>
different cec?ing metods1
9ules of tum6
Lcontri6ution /or 7age 9@9 Std$ev1 is o? 6elo4 2L and unaccepta6le a6ove (L
Ltolerance is o? 6elo4 1=L and unaccepta6le a6ove 3=L
num6er of distinct categories
signal&to&noise ratio U round/1."1VStd$evparts3Std$ev79@91
1 no value for process control> part all loo? te same
2 can see t4o groups> ig3lo4> good36ad
3 can see tree groups> ig3mid3lo4
" accepta6le measurement system
effective resolution
control limits U Y3& 3 Zs[
more #=L of measurements are out of control limits
3 metods for 7age 9 @ 9 computation
sort form
a fe4 operators measure te same parts
for eac part> te range of te measure is computed. Te sum of ranges is computed and divided 6y te
num6er of parts to get te average range. 8ultiply 6y #.1# and divide 6y te num6er of te follo4ing ta6le
to get te 7age 9 @ 9.
Aum6er of operators
2 3 "
A
u
m
6
e
r

o
f

p
a
r
t
s
1 1."1 1.)1 2.2"
2 1.2( 1.(1 2.1#
3 1.23 1.'' 2.12
" 1.21 1.'# 2.11
# 1.1) 1.'" 2.1=
% 1.1( 1.'3 2.=)
' 1.1' 1.'3 2.=)
( 1.1' 1.'2 2.=(
) 1.1% 1.'2 2.=(
1= 1.1% 1.'2 2.=(
te sort form 7age 9@9 does not provide a 4ay of separating gage repeata6ility and reproduci6ility.
long form
+A0.+
to 6e 4ritten
4. (nal$se
06-ectives,
Current process 6aseline as 6een calculated.
Te improvement goal of te pro-ect as 6een statistically defined.
+ list of statistically significant !s as 6een generated as a result of analyEing istorical data.
$.1. Step $ % &sta'lish process capa'ility
Some distri6utions
;niform distri6ution
Triangular distri6ution
Aormal distri6ution
<xponential distri6ution
Te distri6ution sape is not so important as to understand 4y te o6served measure as tis distri6tuion and
not anoter one.
.1.1. Continuous data
4.1.1.1. Some definitions
2ro6a6ility,
outcomes of num6er total
value 4it tis outcomes of num6er
Te sum of all pro6a6ilities is 1.
Te sym6ols used for te population and for te sample are not te same,
population sample
mean /average1

/mu1 U
N
x
N
i
i
=1
x /x 6ar1 or
\
/mu at1 U
n
X
n
i
i
=1

standard deviation /sigma1 s or \ /sigma at1
variance
2

U
( )
N
X
N
i
i
=

1
2

2
s
or
2
\
U
( )
1
1
2

=
n
X X
n
i
i
8ode, most frequent value
9ange U 8ax & 8in
8ean ,
N
X
i
$eviation U /!i & !6ar1
Sum of square /SST1 U sum of te square deviations
.ariance U average SST U
2
U / !i & !

1
2
3 A
Standard deviation U square root of te variance U
( )
N
X Xi


=
2

Coefficient of variation U ratio of stdev to mean expressed as percentage U 1== V


X

Sum of square SST U SS4 Y SS6


SST, sum of squares total
SSG, sum of square 4itin
SS5, sum of square 6et4een
4.1.1.2. Tools
mettre cacun des tools dans un paragrape ind]pendant
o :istogram to display variation in a process
2urpose, To display variation in a process. Converts an unorganiEed set of data or group of measurements
into a coerent picture.
Gen, To determine if process is on target meeting customer requirements. To determine if variation in
process is normal or if someting as caused it to vary in an unusual 4ay.
:o4,
Count te num6er of data points
$etermine te range /91 for entire set
$ivide range value into classes /P1
$etermine te class 4idt /:1 4ere : U 93P
$etermine te end points
Construct a frequency ta6le 6ased on values computed in previous step
Construct a :istogram 6ased on frequency ta6le
o $ot 2lot to display variation in a process
2urpose, To display variation in a process. Quic? grapical comparison of t4o or more processes.
Gen, First stages of data analysis.
:o4,
Create an ! axis
Scale te axis per te range in te data
2lace a dot for eac value along te ! axis
Stac? repeat dots
o 5oxplot to get an overvie4 of te data
2urpose, To 6egin an understanding of te distri6ution of te data. To get a quic?> grapical comparison of
t4o or more processes
Gen, First stages of data analysis
:o4,
minimum o6servation tat fall 4itin te lo4er Q1 & /Q3&Q11V1.#
Q1 & first quartile, 2#L of o6servations is 6elo4
median, #=L of o6servations is 6elo4
if te num6er of o6servations is even> te median is te average of te t4o center values
Q2 & tird quartile, '#L of o6servations is 6elo4
maximum o6servation tat fall 4itin te upper Q3 Y /Q3&Q11V1.#
outliers, o6servations outside te lo4er and upper limits
o 9un Cart to display data trends over time
2urpose, To trac? process over time in order to display trends and focus attention on canges in te process
Gen,
To esta6lis a 6aseline of performance for improvement
To uncover canges in your process
To 6rainstorm possi6le causes for trends
To compare te istorical performance of a process 4it te improved process
:o4,
$etermine 4at you 4ant to measure
$etermine period of time to measure and in 4at time increments
Create a grap /vertical axis U occurrences> oriEontal axis U time1
Collect data and plot
Connect data points 4it solid line
Calculate average of measurements> dra4 solid oriEontal line on run cart
+nalyEe results
Indicate 4it a dased vertical line 4en a cange 4as introduced to te process
o 8ulti&.ari Cart
2urpose,
To identify te most important types or families of variation
To ma?e an initial screen of process output for potential !s
Continuous B> $iscrete !
use it to see impacts of several families of variations, variation 4itin /positional or location related1>
variation 6et4een /cyclical or 6atc3piece related1 or variation time&to&time /temporal1
4.1.1.3. normalit$ testin)
o mean U median U mode
o point of inflexion at
o normal pro6a6ility plot
o Te result of te +nderson&$arling Aormality Test tells if te data is normal or not. + p&.alue of greater
tan =.=# means te data is normal. + p&.alue of less tan =.=# means tat te data is not normal.
8ost of te time 4en te data is not normal> tis is 6ecause 4e are measuring te process as a 4ole and te
results are te aggregation of several normal su6&results /i.e. te process is an aggregate of several su6&
processes1.
Gen te data is not normal> it is may 6e 6etter to use te median and te span instead of te mean of te
standard deviation to descri6e te data.
To descri6e te central tendency,
mean
median /if te distri6ution as a long tail or extreme outliers 4ic inflate3deflate te average.1
Q1 /if te data is s?e4ed to4ard te lo4 values1 or Q3 /if te data is s?e4ed to4ard te ig values1
To descri6e te variation
standard deviation
span /2)# ^ 2#1 /used 4it long&tailed distri6utions1
sta6ility factor SF U Q13Q3 /used 4it s?e4ed distri6utions1, te closer tat te resulting num6er
/Sta6ility Factor1 is to 1> te less variation is in te processD te closer te num6er is to = te greater te
variation.
4.1.1.4. *ormal distribution
computing in Excel
A098SIA./1&$201 compute C from defect rate
1&A098S$IST/C1 compute defect rate from C
+ s,ort term + lon) term DPM-
2 =8! 3='!3%8!3
3 18! ##'=%823
28! #2=)8#'
! 38! 2328#%
# 8! 38=
52U+2&1>#
C2U/1&A098S$IST/5211V1=_%
S?e4ness
S?e4ness is a measure of asymmetry. + value more tan or less tan Eero indicates s?e4ness in te data.
5ut a Eero value does not necessarily indicate symmetry.
Purtosis
Purtosis is one measure of o4 different a distri6ution is from te normal distri6ution. + negative value
typically indicates a distri6ution more pea?ed tan te normal. + positive value typically indicates a
distri6ution flatter tan te normal.
C6enc
Specs
*S*
>S*
4easurements
mean
stdev
?usl @,C2$C3.0C
?lsl @,C3$C1.0C
/usl @1$*-&.A-;46*5.ST6A"6;",C#.
/lsl @1$*-&.A-;46*5.ST6A"6;",C%.
/total @S-445,C'7C).
?2ench @*-&.A-;46*5.ST6A"6;".&A+5;S5,1$C1=.
if data is normal> 8inita6 Six Sigma [ process report
if data is not normal> minita6 Six Sigma [ product report
.1.2. "iscrete data
4.1.2.1. Definitions
unit, num6er of elements inspected or tested /;1
opportunity, an inspected or tested caracteristic /021
defect, a non&conformance /$1
defect per unit, $2;U$3;
total opportunities, T02U;V02
defect per opportunity, $20U$3T02
defect per million opportunities, $280U$20V1=
%
Clong term U 1 & A098$IST/$201
Csort term U Clong term Y 1.#
4.1.2.2. .ield
Classical Bield /Bc1, num6er of defect&free parts for te 4ole process divided 6y te total num6er of parts
inspected. If 4e say te yield is 33" or '#L> 4e lose valua6le data on te true performance of te process. Tis
loss of insigt 6ecomes a 6arrier to process improvement.
First Time Bield /BFT1, te num6er of defect&free parts divided 6y te total num6er of parts inspected for te
first time. If 4e say te yield is 13" or 2#L> 4e are really tal?ing a6out te First Time Bield /FTB1. Tis is a
6etter yield estimate to drive improvement.
BT2 is te percentage of units tat pass troug an operation 4itout any defects. Tis is te 6est yield estimate
to drive improvement.
Trougput Bield,
BT2 U 2/=1 U e
&$2;
U e
&2.2#
U .1=#" U 1=.#"L
percentage of units tat pass troug an operation 4itout any defects. Tis is te 6est yield estimate to drive
improvement.
B9T is te 9olled Trougput Bield
tis is te yield of a process consisting of several steps
B9T is te product of te yield of eac step
Quantification of $efects
inspection effectiveness

$2;
S

$2;
o

$2;
<

2ercentage of li?eliood of detecting a default, <
Su6mitted $2; *evel, $2;S
<scaping $2; *evel, $2;0
06served $2; *evel, $2;<
$2;S U $2;0 Y $2;<
$2;0 U $2;S V <
$2;< U $2;S V /1&<1
.1.3. /rocess capa2ility
Gat is te 6est te process can 6eH
4.1.3.1. Special /ersus Common Cause /ariation
special /assigna6le1 cause variation,
It is non&random variation 4ic can 6e assigned to specific causes
It is controlla6le variation
common /random1 cause variation,
It is an inerent> natural source of variation of te process
It is not controlla6le variation
4.1.3.2. !ational Sub)roupin)
coose su6groups so tat,
Tere is maximum cance for te measurements in eac su6group to 6e ali?e. + su6group sould only
contain common cause variation.
Tere is maximum cance for su6groups to differ from one to te next. Te difference 6et4een te
su6groups is te special cause variation.
4.1.3.3. + s,ift
Te difference 6et4een a long&term data collection and sort&term data collection can 6e demonstrated 6y C
sift. 0ver te sort&term> common cause variation is present> 6ut not special cause variation or mean sifts. Te
typical amount of variation added 6y including te long&term variation and mean sifts is 1.#F.
4.1.3.4. +
s,ort term
0s +
lon) term
;se +nova 1&4ay in 8inita6, SSG is te sum of squares for error> SS5 is te sum of squares for group
Clong term ta?es into account SST
Csort term ta?es into account SSG
*ong&term is 4at is actually going on in te process. It is defined 6y tecnology and process
control.
Sort&term is te 6est te process can do. It is limited 6y tecnology only. Tis is te entitlement of te process.
Csift U Clt & Cst
Te larger te Csift > te greater te control pro6lem. + typical value is 1.#.
$.2. Step ( % #efine performance o'jecti)es
5encmar?ing is te process of continually searcing for te 6est metods> practices and processes> and eiter
adopting or adapting teir good features and implementing tem to 6ecome te R6est of te 6estR possi6le
6encmar?ing,
<valuating te Competition
$etermine te 5est 2rocess or 2roduct
$etermining te 5est 5usiness Strategy
Pey 2arameter 4itin a 2rocess
5est 2ractices
5encmar?ing sould not 6e used as a single event> 6ut sould 6e utiliEed on a continuous 6asis.
6encmar?ing
internal, similar activities 4itin 7< 7<> 6ut in different locations> departments> operating units> country...
competitive, direct competitors selling to same customer 6ase
functional, organiEations recogniEed as aving 4orld class processes regardless of industry
cec?list
identify process to 6encmar?
Select process and define defect and opportunities.
8easure current process capa6ility and esta6lis goal.
;nderstand detailed process tat needs improvement.
select organiEation to 6encmar?
0utline industries3functions 4ic perform your process.
Formulate list of 4orld class performers.
Contact te organiEation and net4or? troug to ?ey contact.
prepare for te visit
9esearc te organiEation and ground yourself in teir processes.
$evelop a detailed questionnaire to o6tain desired information.
Set&up logistics and send preliminary documents to organiEation.
visit te organiEation
Feel comforta6le 4it and confident a6out your ome4or?.
Foster te rigt atmospere to maximiEe results.
Conclude in tan?ing organiEation and ensure follo4&up if necessary.
de6rief @ develop an action plan
9evie4 team o6servations and compile report of visit.
Compile list of 6est practices and matc to improvement needs.
Structure action items> identify o4ners and move into Improve pase.
retain and communicate
9eport out to 6usiness management and %F leaders.
2ost findings and3or visit report on local server3%F 6ulletin 6oard.
<nter information on 7< Intranet 6encmar?ing pro-ect data6ase.
$.. Step * % +dentify )ariation standard
.3.1. Tools
cause and effect /fis6one diagram1
2areto
process mapping
F8<+
4.3.1.1. %is,bone 1cause and effect dia)ram2
2urpose,
To provide a visual display of all possi6le causes of a specific pro6lem
Gen,
To expand your tin?ing to consider all possi6le causes
To gain groupJs input
To determine if you ave correctly identified te true pro6lem
:o4,
$ra4 a 6lan? diagram on a flip cart.
$efine your pro6lem statement.
*a6el 6rances 4it categories appropriate to your pro6lem.
examples of categories,
measurements> materials> men @ 4omen> moter nature /environment1> metods> macines
" 2s /...1,2olicies> 2rocedures> 2eople> 2lant
5rainstorm possi6le causes and attac tem to appropriate categories.
For eac cause as?> MGy does tis appenHN
+nalyEe results> any causes repeatH
+s a team> determine te tree to five most li?ely causes.
$etermine 4ic li?ely causes you 4ill need to verify 4it data.
4.3.1.2. Pareto c,art
2urpose,
To separate te vital fe4 from te trivial many in a process. To compare o4 frequently different causes occur
or o4 muc eac cause costs your organiEation.
Gen,
To sort data for determining 4ere to focus improvement efforts.
To coose 4ic causes to eliminate first
To display information o6-ectively to oters
:o4,
Collect data /cec?seets> surveys1.
Total results and arrange data in descending order.
$ra4 and la6el a 2areto Cart.
Te !&axis so4s te categories of pro6lems.
Te left B&axis so4s te frequency3cost3impact in te units of measure /count> `> time> etc.1.
Te rigt B&axis so4s te percentage /frequency of occurrence3total count of all occurrences1.
Te 6ars so4 te value & frequency of occurrence.
Te line so4s te cumulative percentage.
$ra4 te largest frequency on te left and 4or? your 4ay do4n to te smallest on te rigt.
+nalyEe results.
<valuate improvement effectiveness after cange initiated 6y comparing 6efore and after 2areto carts.
4.3.1.3. Process Map (nal$sis
value&added 4or?, steps tat are essential 6ecause tey pysically cange te product3service and te
customer is 4illing to pay for tem
value&ena6ling 4or?, steps tat are not essential to te customer> 6ut tat allo4 te value&adding tas?s to 6e
done 6etter3faster
nonvalue&added 4or?, steps tat are considered non&essential to produce and deliver te product or service
to meet te customerJs needs and requirements. Customer is not 4illing to pay for step.
Internal Failure, Steps tat are related to correcting in&process errors due to failures in current or prior
step in te process. <xample, 9e4or?
<xternal Failure, Steps 4ic relate to fixing errors in te product tat te customer as found and as
returned to you. <xamples, Customer follo4&up> 9ecall
Control3Inspection, Steps for internal process revie4 often referred to as Mte cec?er&cec?ing&te
cec?er.N <xamples, Inspection> +pproval3revie4> 5ureaucracy
$elay, Steps 4ere 4or? is 4aiting to 6e processed at tat step. <xamples, 5ac?logs> Queues> Store>
5ottlenec?s
2reparation3Set&;p, Steps tat prepare 4or? for a su6sequent activity in te process. <xamples,
<ntering into a computer> 9etrieve policies3pricing
8ove, Steps tat entail te pysical transport or transmit of outputs 6et4een activities in a process.
<xample, Fax3mail
process time Y delay time U cycle time
Te delay may 6e due to,
7aps, responsi6ility for a given step in te process is unclear> or te process seems to go off trac?.
9edundancies, duplication of efforts suc as 4en t4o people or groups approve a document. 9edundancies
occur 4en different groups ta?e action tat tey are una4are is 6eing done some4ere else in te process.
Implicit or unclear requirements, operational definitions do not exist or clear differences exist in perspective
or interpretation.
Tric?y and&offs, no cec? is in place to assure te process is continuing 4itout delays. For example>
$epartment + sends someting to $epartment 5 6ut neiter as a 4ay to ?no4 if it 4as properly received.
Conflicting o6-ectives, te goals of one group cause pro6lems or errors for anoter. For example> one group
is focused on process speed 4ile anoter is oriented to error reduction ^ te result may 6e tat neiter
group accomplises its o6-ectives.
Common pro6lem areas, occurs 4en steps are repeated in a variety of places in process. Aoting tese areas
may provide insigt into potential solutions.
$ra4 a matrix,
te columns are
te 4or?flo4 steps
te total
te total in percentage
te ro4s are
average time
.alue&+dded
Aon&.alue&+dded su6classified into
Internal Failure
<xternal Failure
Control3Inspection
$elay
2rep3Set&;p
8ove
.alue&<na6ling
4.3.1.4. %M3(
.3.2. Bypothesis testing
+n assertion or con-ecture a6out one or more parameters of a population/s1.
To determine 4eter it is true or false> 4e must examine te entire population. Tis is impossi6leQ
Instead use a random sample to provide evidence tat eiter supports or does not support te ypotesis
Te conclusion is ten 6ased upon statistical significance
It is important to remem6er tat tis conclusion is an inference a6out te population determined from te sample
data
:ypotesis Testing 2rotocol
state te null ypotesis := /using U> Z or [1
state te alternative ypotesis :+ /using a> b or c1
test alternative ypotesis 4it Statistical Test
6ased on te test result> 4e re-ect or fail to re-ect te null ypotesis :o.
transfer follo4ing diagram
Te 2&.alue,
d is te maximum accepta6le pro6a6ility of 6eing 4rong if te alternative ypotesis is selected.
Te p&value is te pro6a6ility tat you 4ill 6e 4rong if you select te alternative ypotesis. Tis is a Type I
error.
;nless tere is an exception 6ased on engineering -udgment> 4e 4ill set an acceptance level of a Type I
error at d U =.=#.
Tus> any p&value greater tan =.=# means 4e accept te null ypotesis
any p&value less tan =.=# means 4e re-ect te null ypotesis.
Seven examples of ypoteses
te mean is on target
:o, e U constant U T
:a, e c constant U T
te variance is on target
:o, F
2
U constant
:a, F
2
c constant
t4o populations ave te same mean
:o,f1 U f2
:a, f1c f2
a population as a mean smaller tan te mean of anoter population
:o, f1 a f2
:a, f1 [ f2
t4o populations ave te same variance
:o, F1
2
U F2
2
:a, F1
2
c F2
2
te populations ave te same mean
:o, f1 U f2 U . . . U fn
:a, at least one not equal
te populations ave te same variance
:o, F1
2
U F2
2
U . . . U Fn
2
:a, at least one not equal
:ypotesis tests
single ! several !s
continuous discrete continuous discrete
s
i
n
g
l
e

B
c
o
n
t
i
n
u
o
u
sn
o
r
m
a
l
scatterplot
single
regression
curve fitting
/a?a
correlation1
T&test
omogeneity
of variance
1&4ay
+A0.+
multiple
regression
$0<
2> 3> "> #..&4ay
+A0.+
n
o
n
&
n
o
r
m
a
l omogeneity
of variance
*evineIs
8oods median
correlation
d
i
s
c
r
e
t
e
logistic
regression
goodness of fit
test of
independence
multiple
logistic
regression
multiple
logistic
regression
s
e
v
e
r
a
l

B
s
multivariate statistics
in te analysis pase> 4e only loo? at te xingle ! 3 single B> oter tests 4ill 6e seen in te improve pase
g mettre sous forme textuelle
4.3.2.1. "$pot,esis 4estin)5 Continuous .6 Discrete 7
to compare t4o or more distri6utions,
Study sta6ility
in 8inita6 > use 9un Cart
loo? at grap sape
cec? p&values for clustering> mixtures> trends> and oscillations /:o, $ata is random> special causes not
presentD :a, $ata is not random> special causes presentD so all p&values sall 6e greater tan =.=#1
Study sape
in 8inita6> use Stat[5asic Statistics [ $isplay $escriptive Statistics [ 7raps... [ 7rapical summary
in 8inita6 and loo? at sape and p&value for eac distri6utions
p&value 4ill indicate if data is normal or not, :o, Te data is normally distri6utedD :a, Te data is not
normally distri6uted /p[=.=# indicate tat te distri6ution is normal1
Study spread
in 8inita6> stac? te distri6utions in a column /and te distri6ution indices in anoter column1> use
Stat[+A0.+[:omogeneity of .ariance
to get te p&value,
o loo? at F&test /for 2 populations1 or 5artlettIs test /if more populations1 is data is normal
o loo? at *eveneIs test if data is not normal
p&value Z =.=# means tat te distri6utions are different
Study Centering
4e compare te means of te populations
o to compare t4o populations
te t4o populations must ave te same variance
in 8inita6> Stat[ 5asic Statistics[2&Sample t
/do not forget to cec? +ssume <qual .ariance1
verify tat te populations are different 6u cec?ing tat
& tat = is not in te )#L confidential interval of mu 0per 1 & mu 0per 2 or
& te p&value of te T&test /ypotesis is mu 0per 1 U mu 0per 21, p Z =.=#
o to compare te mean of more tan t4o populations
te populations must ave te same variance
in 8inita6> Stat[ +A0.+[0ne Gay
cec?
& tat te )#L confidential intervals for te mean overlap
& te p&value
loo? at excel document +nalyEe tool guide.xls
4at a6out discrete data
non normal data,
loo? at plotcart
cec? tat data is normal
sometimes> te data appears as not normal 6ecause te measurement resolution is not ig enoug
remove outlayers
try to find some possi6le su6groups
cec? tat tese groups are really different 6y performing a median test /8oods median test1
ten performing :0.
data transformation sould 6e used very carefully, it is easy to manipulate data 4ic is meaningless.
pair testing ,
4e 4ant to test if a cange on a population of very different specimens as an impact
4e must renormaliEe te data 6y computing te difference of te ne4 and old values and perform
a 1&sample T&test to see if = is in te )#L confidence interval of te mean or
perform a T&test of te mean
Stat [ 5asic Statistics [ + Sample T and select Test 8ean
o4 to dra4 te normality plotH
discrete B & discrete !
Ci&Square test
Te population are different if p Z =.=#. /If p[=.=# 4e re-ect te alternative ypotesis.1
in 8inita6> to compute a sample siEe given te po4er of a test
Stat [ 2o4er and Sample SiEe [ 1&sample t
comparison of means
1&sample T&test is used 4en comparing a population to a target
2&sample T&test is used 4en comparing t4o populations
+A0.+ is used 4en comparing several populations
continuous B & continuous !
use Scatter 2lot to loo? at te data
in 8inita6> 7rap[2lot
in 8inita6> Stat [ 9egression [ Fitted *ine 2lot [ 0ptions [ h $isplay Confidence 5ands> $isplay prediction
6ands i
red lines indicate tat 4e are )#L sure tat te regression line is 6et4een tese t4o lines
6lue lines indicate tat 4e are )#L sure tat te measurement are 6et4een tese t4o lines
n + confidence 6and /or interval1 is a
measure of te certainty of te sape of
te fitted regression line. In general> a
)#L confidence 6and implies a )#L
cance tat te true line lies 4itin te
6and. S9ed linesT
n + prediction 6and /or interval1 is a
measure of te certainty of te scatter of
individual points a6out te
regression line. In general )#L of
te individual points /of te population on
4ic te regression line is 6ased1 4ill 6e
contained in te 6and. S5lue linesT
general linear model
see if te data is un6alanced
Stat [ Ta6les [ Cross Ta6ulation
if te data is un6alanced> 4e 4ill us 7*8
Stat [ +A0.+ [ 7eneral *ineal 8odal
te sum of squares depend on te order of te factors> te ad-usted sum of square sould 6e used instead.
pZ=.=# indicates tat te factor as an impact on te output
if te sum of squares is too small compared to te total sum of squares> tis means tat some factors ave 6een
forgotten
analysis of residuals
Stat [ 9egression [ 9esidual plots...
Gat is te Aormal 2lot of 9esiduals dra4ingH
. Impro0e
To develop a proposed solution,
o Identify an improvement strategy
o <xperiment to determine a solution
o Quantify financial opportunities
To confirm tat te proposed solution 4ill meet or exceed te quality improvement goals,
o + pilot, includes one or more small&scale tests of te solution in a real&4orld 6usiness environment
o To statistically confirm tat an improvement exists /ypotesis tests1
To identify resources required for a successful full&scale implementation of te solution.
To plan and execute full scale implementation including training> support> tecnology rollout> process and
documentation canges.
(.1. Steps , % Screen potential causes
(.2. Steps - % #etermine relationship
!.2.1. &mprovement stategy
operating parameters,
!s tat can 6e set at multiple levels to study o4 tey affect te process B
Canges in teir settings impact te B directly and influence variation
8ay 6e continuous and3or discrete
use Statistical 5rea?troug Strategy
you need to ?no4 o4 tey are related to eac oter and to te B to develop an
appropriate solution1
develop a matematical model> or
determine te 6est configuration or com6ination of !s
critical elements,
!s tat are independent alternatives
!s tat are not necessarily measura6le on a specific scale> 6ut ave an affect on te process
use *ean 2rocess Improvement
you need to develop and test several practical alternatives to determine 4ic is te 6est solution
optimiEe process flo4 issues> or
standardiEe te process> or
develop a practical solution
!.2.2. Tools
depend on te pro6lem sopistication /complexity> 6usiness impact> ris?> data availa6ility1
6asic
o fis6one
o 6oxplot
o linear regression
o ypotesis testing /E&test> t&test> +A0.+> ci&square> :0.1
o process map
o time order plots
o mista?e proofing
o multi&vari plot
o force fields
o action 4or?&out
intermediate
o $0< /full> fractional1
o multi&variate regression
advanced
o response surface
o Taguci /inner3outer array1
!.2.3. /rocess of e(perimentation
Tere are seven steps,
1. $efine 2ro-ect
Identify responses
2. <sta6lis Current Situation
3. 2erform +nalysis
Identify factors
Coose factor levels
Select design
9andomiEe runs
Collect data
+nalyEe data
$ra4 conclusions
.erify results
". $etermine Solutions
#. 9ecord 9esults
%. StandardiEation
'. $etermine Future 2lans
!.2.. "-57 "esign -f 5(periment
2ossi6le ad&oc improvement strategies,
Stic?&Git&a&Ginner
ma?e a cange, if tis cange results into an improvement> ?eep itD oter4ise> cange 6ac? and try anoter
factor
W ma-or dra46ac? , no guarantee to find te optimal solution> solution depends to some degree on 4ic
factor 4e decide to cange first
0ne&Factor&at&a&Time
Cange only one factor at a timeD once all te factors ave 6een tested> include tat resulted into an
improvement in te final design
W ma-or dra46ac? , multiple factor interactions are not ta?en into account
9eplication, multiple o6servations performed 4it te different experimental units 4it te same factor settings
to measure experimental varia6ility> so 4e can decide 4eter te difference 6et4een responses is due to
te cange in factor levels /an induced special cause1 or to common cause varia6ilityD to see more clearly
4eter or not a factor is important
to o6tain t4o responses for eac set of experimental conditions W location> spread
replication provides te opportunity for factors tat are un?no4n or uncontrolla6le to 6alance out
along 4it randomiEation> replication acts as a 6ias decreasing effect
9epetition, multiple o6servations performed 4it te same experimental unit
9andomiEation, assign te order in 4ic te experimental trials 4ill 6e run using a random mecanism
averages te effect of any lur?ing varia6les over all of te factors in te experiment
elps validate statistical conclusions made from te experiment
.2.4.1. (nal$8in) a full factorial 9 replicated desi)n
#.2.".1.1. plot te ra4 data
o display run cart to see possi6le lur?ing varia6le
o display grapic statistic of te measures
cec? if data is normal
o dra4 6oxplot for te consolidated factor settings
o dra4 6oxplots for eac parameter to see 4ic parameters ave te igest impact
o compute te pooled standard deviation /square root of te average of te variances for eac factor setting1
tis is 4y 4e need replicated measures
o use +A0.+ one&4ay /for te consolidated factor settings1 to see 4at is te variation due to parameter
cange and te variation due to process3measurement
/to not forget to store te residuals> tey 4ill 6e used later1
4e must verify tat te parameters ave an impact /pZ=.=#> null ypotesis is tat te populations of eac
experiment are te same1
compute te residuals
#.2.".1.2. plot te residuals
o display run cart of te residuals
o plot te residuals against te average
o plot te residuals on a normal pro6a6ility scale
cec? tat tey are normal
#.2.".1.3. examine factor effects
o main effect plots
factor effect U average for te ig setting & average for te lo4 setting
a Factor <ffect is te cange in response due to a Mt4o&unitN /&1 to Y11 cange in te factor.
a Factor Coefficient is te cange in response due to a Mone&unitN /&1 to = or = to Y11 cange in te factor.
Factor Coefficient U Factor <ffect 3 2
o interaction plots
o perform Stat [ $0< [ Factorial 2lots to display
/if te 6uttons are grayed out> use Stat [ $0< [ Create Factorial $esign to create a $0< and ignore it1
te main effects
te interactions
/te cu6e ave no real interest1
#.2.".1.". confirm impressions 4it statistical procedures
o perform Stat [ $0< [ +nalyEe Factorial $esign
te significant parameters ave a p&value less tan =.=#
coded units, parameters are normaliEed in te S&1>1T range
o it is also possi6le to use Stat [ +A0.+ [ 7eneral *inear 8odel
tis so4s te importance of eac parameters> 6ut not te interactions> tey are in te error sum of square
to see te interactions> add te factors +V5> +VC> 5VC and +V5VC /it is not necessary to create te
corresponding columns> 8inita6 4ill andle tese products automatically1
te factors aving an impact are te ones 4it p Z =.=# and te sum of squares are te most significant /6ot
criteria sould 6e consistent> small p&value sould 6e ig sum of squares1
#.2.".1.#. summariEe conclusions
o *ist all te conclusions you ave made during te analysis.
o Interpret te meaning of tese results. For example> relate tem to ?no4n pysical properties> engineering
teories> or your o4n personal ?no4ledge.
o 8a?e recommendations.
o Formulate and 4rite conclusions in simple language.
at te 6ottom of te 9un CartD tere are four p&values> tey sould all 6e less tan =.=#.
.2.4.2. Compute Prediction Model
?eep only te significant factors
step 1, prepare a reduced design matrix
step 2, define custom design
step 3, run Factorial +nalysis using Standard $eviation as response
step ", 6uild prediction model using significant factors
step #, explore prediction models
optimiEe center /mean1 and spread /standard deviation1
step %, determine factor settings
.2.4.3. !educin) si8e of e:periments
alf factorial design, te last factor is te product of te oter ones
e.g. for four parameters> $ is equal to +5C> 6ut it 4ill 6e impossi6le to separate te effects of $ and +5C>
te effect of + and 5C$> +5 and C$> +5C$ and te mean...> 6ut effects of te com6ination of more tan 2
factors to not exist in reality
screening design
4e coose te next po4er of 2 and perform a full factorial for te first factors> te last ones are computes as
product of te first ones
2lac?et&5urman design
resolution
$etecta6le <ffect SiEe, smallest effect 4e can consistently detect 4it te current num6er of experimental runs
First 4e must identify te current level of varia6ility /standard deviation1 for eac response varia6le.
runs of num6er
stdev V (
effect siEe detecta6le =
(.. Step . % &sta'lish operating tolerance
!.3.1. Tolerancing
Step ( provided te experimental tecniques to esta6lis te relationsip 6et4een te measura6le B
caracteristics and te controlling ! factors.
Step )> tose relationsips> so&called transfer functions> 4ill 6e used to define te ?ey operating parameters and
tolerance to acieve te desired performance of te CTQs.
!.3.2. simulation
Simulation
advantages
no need to 6uild a prototype> ma?e an experiment
does not distur6 te system
no destruction in case of destructive test
disadvantages
requires a good model
metodology for modeling and simulation process
Specify & understand te pro6lem to 6e studied and o6-ective of doing simulation. $evelop a pro-ect plan
and get customer sign&off.
$evelop & descri6e model 6ased on expert intervie4s and o6servation of process.
Quantify & collect data needed to define process properties.
Implement & prepare soft4are model.
.erify & determine tat computer model executes properly.
.alidate & compare model output 4it real process /if it exists1.
2lan & esta6lis te experimental options to 6e simulated.
Conduct & execute options and collect performance measures.
+nalyEe & analyEe simulation results.
9ecommend & ma?e recommendations.
Crystal 5all
ttp,33sixsigmacafe.researc.ge.com
;. Control
06-ectives
to ma?e sure tat our process stays in control after te solution as 6een implemented.
to quic?ly detect te out of control state and determine te associated special causes so tat actions can 6e
ta?en to correct te pro6lem 6efore non&conformances are produced.
process entitlement, te 6est tat can 6e acieved 4it te current tecnology of our process
2rocess Control System,
$efines te actions> resources> and responsi6ilities needed to ma?e sure te pro6lem remains corrected and
te 6enefits from te solution continue to 6e realiEed.
2rovides te metods and tools needed to maintain te process improvement> independent of te current
team.
<nsures tat te improvements made ave 6een documented /often necessary to meet regulatory
requirements1.
Facilitates te solutionIs full&scale implementation 6y promoting a common understanding of te process
and planned improvements.
Pey Steps in $eveloping a 2rocess Control System
Complete an implementation plan,
2lan and implement te solution and develop a metod to control eac vital ! or ?ey sources of
variation
$efine all possi6le areas tat may require action in order to control te process ! and ten determine
te appropriate course of action to ta?e
$evelop a data collection plan to confirm tat your solution meets your improvement goals,
<sta6lis ongoing measurements needed for te pro-ect B and create a response plan to follo4 in case
process performance falls 6elo4 esta6lised standards
Communicate your strategy,
$ocument te process and control plan to ensure process standardiEation and te continuation of te
solutionIs 6enefits
Train personnel.
9un te ne4 process and collect te data to confirm your solution.
*.1. Step 1/ % define and )alidate measurement system on 0s in Actual
Application
define and validate te measurement system on !s in te actual application.
delivera6le, measurement system is adequate to measure !s.
4e use te same tool as te ones used in step 3 to measure te B
*.2. Step 11 % determine process capa'ility
determine process capa6ility.
delivera6le, determine post&improvement capa6ility and performance /Cst and Clt1.
Calculate post&improvement capa6ility of performance 6ased on te tecnique descri6ed in Step ".
Confirm improvement goal esta6lised in Step # as 6een realiEed on te B.
If not> go 6ac? to Step % to loo? for additional sources of variation.
*.. Step 12 % +mplement process control
quality plan
?ey questions
4y monitorH
4at sould I monitorH
o output measures
o process measures
o input measures
o4 muc data do I collectH
o4 can I detect canges in process variation or capa6ilityH
4at do I do if I detect a cangeH
if te process is in control and capa6le> are my customers still satisfiedH
#.3.1. ris9 management
1. Identify te ris? elements @ te ris? types.
Cost
Tecnology
Specification
8ar?eting
Installation
2. +ssign ris? ratings to te ris?s,
9ate pro6a6ility of occurrence /1 to #1,
9ate consequence of occurrence3impact /1 to #1,
9is? factor score O consequence x pro6a6ility U 1 to 2#.
3. 2rioritiEe te ris?s,
:I7: 9is? U 9<$ 9is? U Score of 1% to 2#
8<$I;8 9is? U B<**0G 9is? U Score of ) to 1#
*0G 9is? U 79<<A 9is? U Score of 1 to (
". Identify te ris? a6atement plans /ig and medium ris?s1.
#. Incorporate te ris? a6atement plan into te 4or? plans.
%. Trac? te ris? score reductions and a6atement actions vs. plan.
'. Continuously update for ne4 ris?s and reduction of old ris?s.
#.3.2. mista9e proofing ,po9a yo9e.
a tecnique for eliminating errors
ma?e it impossi6le to ma?e mista?es
defects is te result of an error
error is te cause of defects
2rinciples for 8ista?e 2roofing,
9espect te intelligence of 4or?ers.
Ta?e over repetitive tas?s or actions tat depend on constantly 6eing alert /vigilance1 or memory.
Free a 4or?erJs time and mind to pursue more creative and value&adding activities.
It is not accepta6le to produce even a small num6er of defects or defective products.
Te o6-ective is Eero defects.
Ten Types of :uman <rror
Forgetfulness /not concentrating1.
<rrors in mis&communications /-ump to conclusions1.
<rrors in identification /vie4 incorrectly...too far a4ay1.
<rrors made 6y untrained 4or?ers.
Gillful errors /ignore rules1.
Inadvertent errors /distraction> fatigue1.
<rrors due to slo4ness /delay in -udgment1.
<rrors due to lac? of standards /4ritten @ visual1.
Surprise errors /macine not capa6le> malfunctions1.
Intentional errors /sa6otage O least common1.
:uman <rror&2rovo?ing Conditions
+d-ustments.
Tooling3tooling cange.
$imensionality3specification3critical condition.
8any parts3mixed parts.
8ultiple steps.
Infrequent production.
*ac? of> or ineffective standards.
Symmetry.
+symmetry.
9apid repetition.
:ig volume3extremely ig volume.
<nvironmental conditions,
8aterial3process andling
:ouse?eeping
Foreign matter
2oor ligting
8ista?e 2roofing Tecniques
Tecnique 2rediction32revention $etection
S:;T$0GA Gen a mista?e is a6out to 6e made Gen a mista?e or defect as 6een made
C0AT90* <rrors are impossi6le $efective items can not move on to te next step
G+9AIA7 Tat someting is a6out to go 4rong Immediately 4en someting does go 4rong
8ista?e 2roofing Steps
1. Identify 2ro6lems
5rainstorming.
Customer returns.
$efective parts analyses.
<rror reports.
Failure 8ode and <ffects +nalysis /F8<+1.
2. 2rioritiEe 2ro6lems
Frequency.
Gasted materials.
9e4or? time.
$etection time.
$etection cost.
0verall cost.
3. See? out te 9oot Cause
$o not use mista?e proofing to cover&up pro6lems or to treat symptoms.
;se mista?e proofing to correct errors at teir source.
0ter metods to determine te root cause are,
+s? M4yN five times
Cause and effect diagrams
5rainstorming
Stratification
Scatterplot
". Create Solutions
8a?e it impossi6le to do it 4rong.
Cost36enefit analysis,
:o4 long 4ill it ta?e for te solution to pay for itselfH
Tin?ing outside of te 6ox.
#. 8easure te 9esults
:ave errors 6een eliminatedH
Gy or 4y notH
Gat is te financial impactH
#.3.3. control plan
+ good Control 2lan 4ill incorporate at least,
Customer&driven Critical&To&Quality /CTQs1.
Input @ 0utput varia6les.
+ppropriate tolerances /specifications for CTQs1.
$esignated control metods> tools and systems,
S2C
Cec?lists
8ista?e proofing systems
Standard 0perating 2rocedures
8anufacturing3Quality3<ngineering Standards
9eaction plan.
#.3.. control charts
Control carts,
+re used to monitor 6ot inputs to process> parameters of a process> or process outputs /!s and B1
+re used to recogniEe 4en a process as gone out of control
+re used for identifying te presence of special cause variation 4itin a process
$o not tell us if 4e meet specification limits
Aeiter identify nor remove special causes
3F&level control limits,
Created 6y Se4art to minimiEe t4o types of mista?es
2laced empirically 6ecause tey minimiEe te t4o types of mista?es
Calling a special cause of variation a common cause of variation
8issing an cance to identify a cange in te process1
Calling a common cause of variation a special cause of variation
Interfering 4it a sta6le process> 4asting resources loo?ing for special causes of variation tat do not
exist
+re not pro6a6ility limits
T4o Types of Control Cart
varia6le cart /continuous1
;ses measured values
7enerally one caracteristic per cart.
8ore expensive> 6ut more information.
attri6ute cart /discrete1
2ass3Fail> 7ood35ad> 7o3Ao&7o information.
Can 6e many caracteristics per cart.
*ess expensive> 6ut less information.
process focused cart
8onitors several parts from same process
8easures deviation from nominal3target
Typically an I @ 89 cart monitoring several caracteristic of several parts
Selecting te +ppropriate Control Cart
varia6le cart
ig volume
!&6ar and 9 carts
!&6ar, average of te sample
9, range of te sample /max of sample ^ min of sample1
if sample siEe is greater tan #> a S cart sould 6e used in place of te 9 cart
lo4 volume
individuals @ moving range carts
individual, te value of te sample
moving range, te different 6et4een te value of te sample and te value of te previous sample
attri6ute cart /defect U a single caracteristic tat does not meet requirementsD defective U a unit tat
contains one or more defects1
constant lot siEe
defects
C cart
defectives
A2 cart
varia6le lot siEe
defects
; cart
defectives
2 cart
Te 2rocess is M0ut&of&ControlN if...
Four Gestern <lectric 9ules
+ point is outside te control limits.
2 out of 3 consecutive points [ 2 F a4ay from te mean on te same side.
" out of # consecutive points [ 1 F a4ay from te mean on te same side.
) consecutive points are on one side of te mean.
8inita6 9ules
one point more tan 3 F from te center line
nine points in a ro4 on te same side of center line
six points in a ro4> all increasing or all decreasing
fourteen points in a ro4> alternating up and do4n
t4o out of tree points more tan 2 F from center line /same side1
four out of five points more tan 1 F from center line /same side1
fifteen points in a ro4 4itin 1 F sigma of center line /eiter side1
eigt points in a ro4 more tan 1 F from center line /eiter side1
process focused cart
1. $efine te process /general is 6etter tan specific1.
2. Identify te parameters tat measure performance.
3. 7ater data in production sequence.
". 9ecord varia6le data as a deviation from nominal3target.
#. +nalyEe for patterns.