2. DMAIC
DEFINE
MEASURE
ANALYSE
IMPROVE
CONTROL
Identify Customer Problem
Translate to Practical
Problem
Translate to statistical
problem
Identify Statistical solution
Translate to practical
solution
1. What is the Project?
2. Define measurement system
3. Validate measure-ment
system
10. Validate measurement
system (X)
12. Implement process
controls
4. Actual process performance
5. Define statistical success
6. Identify causes of defects
7. Determine vital causes
8. Define optimal settings
9. Define tolerance limits
11. Determine new process
capability
3. X & Y
• Y is the outcome of the process.
• X are factors that influence Y
• Usually the project starts with an Y that is not very specific.
This is good for discussion but not for a Lean Six Sigma Project.
• Therefore the so called “external” Y needs to be translated to
an “internal” Y that is specific, concrete and measurable (slice
the project)
• Use the Voice of the Customer (VOC) to capture the
requirements
4. EXAMPLE
• Passengers are not satisfied with public transport.
• External Y = Public transport is not good
• You try to make this a bit more specific:
• Possible technique: Pareto, research
• The biggest problem is:
• Internal Y = the bus is often not on time 0.0
10.0
20.0
30.0
40.0
50.0
60.0
70.0
80.0
90.0
100.0
0
200
400
600
800
1000
1200
1400
Count
Cumulative %
5. STEP 1: PROJECT CHARTER (WHAT IS THE PROJECT?)
DEFINE
Business Case Process start and finish
[Short description of the process. How
should it work (which is the customer
requirement) and why is this process so
important?]
[Clearly define where the process you want
to improve starts and where it ends.]
Problem Statement In Scope and Out of Scope
[Where does the process deviate from the
customer requirement, and why is this a
problem?]
[Hint: don’t make it too big!]
Goal Statement Expected Benefits
[Short description of the improvement
target.
Examples:
Shorten throughput time with xx hour
Don’t do it if there is not enough benefit!
6. SIPOC DIAGRAM
Suppliers
• Raw materials
• Sources
• Manufacturers
• Suppliers
Inputs
• Manpower
• Resources
• Equipment
Process Outputs
• Product
• Timely
delivery
• Increased
quality
Customers
• Young people
• Students
• Service
holders
Requirement
s
• Customer
Satisfaction
• Expected
quality
• Reduced
Backlog
Look for
new
customer
segment
Find
customer
needs
Identify
critical
needs
Develop
prototype
Test
prototype
& go to
production
STEP 1: PROCESS DESCRIPTION OF THE PROCESS TO
IMPROVE
DEFINE
7. Champion: [Name sponsor]
Process owner: [Name]
Sr. employees: [Members project]
Financial Analyst: [Name]
Master Black Belt: [Name]
Black Belt: [Name]
STEP 2: PROJECT TEAM
DEFINE
8. STEP 2: MEASURABLE CCR +
SPECIFICATIONS
Unit: [Output of the Process.
Example: a cookie from a cookie factory]
Chance: [Number of possible defects
Example: 2 (see below)]
Defect: [What leads to an unhappy customer?
Example: a broken cookie, a cookie without a peanut]
Unit of Measure: [Unit in which the output is measured
Example: broken yes/no
peanut: yes/no]
MEASURE
9. STEP 2: MEASUREMENT SYSTEM
MEASURE
Definition of the data
How is the measurement unit
determined?
How is the data collected?
Automated or manual?
Who collects the data?
Frequency and collection
dates?
How will the data be used?
Test hypothesis
Root-cause analysis
How will the data be
presented?
Pareto, Histogram,
Control Chart, Probability plot,
Box plot, …
10. STEP 3: VALIDATION OF THE MEASUREMENT
SYSTEM
• [Gage R&R or Kappa test]
• or
• [The data is currently being used as source for management
info and both the Process Owner and the Campion recognize
this as representative for the process.]
MEASURE
Conclusion: We have a reliable Measurement System
11. STEP 4: CURRENT PERFORMANCE
• [Control Chart, example I-Chart (Control charts, individuals)]
• Probability plot (Graph)
Observation
IndividualValue
332925211713951
0,18
0,17
0,16
0,15
0,14
0,13
0,12
0,11
0,10
_
X=0,13555
UCL=0,16408
LCL=0,10703
1
1
I Chart of Proces X
ANALYZE
Conclusions: The data is representative for the process.
The data is normally distributed. (or not)
Proces X
Percent
0,180,170,160,150,140,130,120,110,100,09
99
95
90
80
70
60
50
40
30
20
10
5
1
Mean 0,1356
StDev 0,01326
N 35
AD 0,936
P-Value 0,016
Probability Plot of Proces X
Normal - 95% CI
12. STEP 4: CURRENT CAPABILITY OF THE
PROCESS
• Process capability (Quality tools, capability analysis, normal)
We have a target that currenty is not met. Most data should be between LSL and USL.
Conclusion: We have a problem.
ANALYZE
0,1620,1440,1260,1080,0900,072
LSL USL
LSL 0,06
Target *
USL 0,1
Sample Mean 0,135554
Sample N 35
StDev (Within) 0,0133565
StDev (O v erall) 0,0133565
Process Data
C p 0,50
C PL 1,89
C PU -0,89
C pk -0,89
Pp 0,50
PPL 1,89
PPU -0,89
Ppk -0,89
C pm *
O v erall C apability
Potential (Within) C apability
PPM < LSL 0,00
PPM > USL 1000000,00
PPM Total 1000000,00
O bserv ed Performance
PPM < LSL 0,01
PPM > USL 996115,04
PPM Total 996115,05
Exp. Within Performance
PPM < LSL 0,01
PPM > USL 996115,04
PPM Total 996115,05
Exp. O v erall Performance
Within
Overall
Process Capability of Proces X
Note: if the data is not
normally distributed, only
these parts are relevant.
13. STEP 5: STATISTIC SUCCESS
• Compare Process capability of department 1 (to be
improved)
• With Process capability of department 2 (is
performing better)
0,1620,1440,1260,1080,0900,072
LSL USL
LSL 0,06
Target *
USL 0,1
Sample Mean 0,135554
Sample N 35
StDev (Within) 0,0132909
StDev (O v erall) 0,0133565
Process Data
C p 0,50
C PL 1,89
C PU -0,89
C pk -0,89
Pp 0,50
PPL 1,89
PPU -0,89
Ppk -0,89
C pm *
O v erall C apability
Potential (Within) C apability
PPM < LSL 0,00
PPM > USL 1000000,00
PPM Total 1000000,00
O bserv ed Performance
PPM < LSL 0,01
PPM > USL 996264,09
PPM Total 996264,09
Exp. Within Performance
PPM < LSL 0,01
PPM > USL 996115,04
PPM Total 996115,05
Exp. O v erall Performance
Within
Overall
Process Capability of Amsterdam
ANALYZE
0,140,120,100,080,060,04
LSL USL
LSL 0,06
Target *
USL 0,1
Sample Mean 0,0899916
Sample N 35
StDev (Within) 0,0216307
StDev (O v erall) 0,0216307
Process Data
C p 0,31
C PL 0,46
C PU 0,15
C pk 0,15
Pp 0,31
PPL 0,46
PPU 0,15
Ppk 0,15
C pm *
O v erall C apability
Potential (Within) C apability
PPM < LSL 28571,43
PPM > USL 257142,86
PPM Total 285714,29
O bserv ed Performance
PPM < LSL 82792,73
PPM > USL 321791,34
PPM Total 404584,07
Exp. Within Performance
PPM < LSL 82792,73
PPM > USL 321791,34
PPM Total 404584,07
Exp. O v erall Performance
Within
Overall
Process Capability of Leeuwarden
Based on this analysis we hope to conclude that department is a
suitable benchmark
14. STEP 6: INVENTORY OF CAUSES
We have measured two (or more) different teams/species/types/whatever, and we
have measured a specific thing like: size, throughput time, number of failures,
etc.
Now we want to know which factors could explain the difference between these
teams/types/etc.
Example: Why is team A making far less mistakes than team B?
First make a long list of all causes (we call this a list of X’s). Use brainstorming
and/or data-analysis
Then shorten the list until max 8 major causes remain.
Use techniques like: common sense, knowledge, experience and hypothesis
testing.
ANALYZE
15. STEP 6: HYPOTHESIS TESTING,
X=SOME-CAUSE
ANALYZE
Hypothesis: For Some-cause there is no difference between X en Y.
Conclusion: Some-cause is a relevant X. (or not)
Show that there is or is not a statistically significant difference in distribution and
average between X and Y.
Thus you need a similar row of data of X and Y.
Verify that data (I-chart, probability plot).
Check which test needs to be done:
• For spread
• and for average.
Note: these tests depend on; is Y discrete or continuous? Is X discrete or
continuous? And do we have 2 or more groups of X?
16. STEP 6: SUMMARY OF CONCLUSIONS
Max 8 important X’s
1. PQR
2. STU
3. VWZ
4. ..
5. ..
6. ..
7. ..
8. ..
ANALYZE
Result:
data: Short list of possible causes, max 8
Process map: most critical process steps
(focus)
17. STEP 7: DETERMINE THE VITAL FEW ROOT
CAUSES
IMPROVE
From 8 important to (max) 3 root causes!!
using Lean techniques and hypothesis testing
18. HYPOTHESIS TESTING DECISION TREE
Internal Y
Discrete data
for Y
Continuous
data for X
Logistic
Regression
Discrete data
for X
Chi Square-
test
Continuous
data for Y
Continuous
data for X
Regression
Discrete data
for X
Mean
problem
1 group of
data for X
2 groups of
data for X
More groups
of data for X
Variance
problem
2 groups of
data for X
More groups
of data for X
Normal:
1-Way Anova
Not-normal:
Kruskal-Wallis test
Normal:
2-Sample t-test
Not-normal:
Mann-Whitney test
Normal:
F-test
Not-normal:
Levene’s test
Normal:
Bartlett’s test
Not-normal:
Levene’s test
Normal:
1-Sample t-test
Not-normal:
1-Sample Wilcoxon test
19. STEP 7: SUMMARY OF CONCLUSIONS
IMPROVE
3 root causes (X’s)
1. Root cause 1
2. Root cause 2
3. Root cause 3
20. STEP 8: DESIGN IMPROVEMENTS PER ROOT
CAUSE
Determine per root cause the optimal setting!
IMPROVE
Root Cause Optimal solution
21. STEP 9: DEVELOP PRACTICAL
SOLUTIONS PER ROOT CAUSE
IMPROVE
Root
cause
Practical
solution
Test
method
(step 11)
Effect costs conclusio
n
Pre
requisites
22. STEP 10: EVERYONE IN THE PROCESS
KNOWS THE NEW WAY OF WORKING,
AND IS CAPABLE OF DOING IT
CONTROL
Prove that everyone understands the new method using the new work
instructions or the Standard Operating Procedures (SOP).
We prove that now and in the future we can rely on our measurement
system by performing an analysis on the main X's.
LTL
corrected
Target UTL
corrected
Unit of
measurement
Gage
R&R in %
Procedure/
Sop nr
remarks
X1
X2
X3
X4
23. STEP 11: CALCULATE NEW “PROCESS
CAPABILITY”
CONTROL
Conclusion: success on short term has been proven!
0,150,140,130,120,110,10
LSL USL
LSL 0,1
Target *
USL 0,14
Sample Mean 0,131175
Sample N 35
StDev (Within) 0,00841327
StDev (O v erall) 0,00841327
Process Data
C p 0,79
C PL 1,24
C PU 0,35
C pk 0,35
Pp 0,79
PPL 1,24
PPU 0,35
Ppk 0,35
C pm *
O v erall C apability
Potential (Within) C apability
PPM < LSL 0,00
PPM > USL 200000,00
PPM Total 200000,00
O bserv ed Performance
PPM < LSL 105,51
PPM > USL 147103,66
PPM Total 147209,17
Exp. Within Performance
PPM < LSL 105,51
PPM > USL 147103,66
PPM Total 147209,17
Exp. O v erall Performance
Within
Overall
Process Capability of Amsterdam
Observation
IndividualValue
332925211713951
0,16
0,15
0,14
0,13
0,12
0,11
_
X=0,13118
UCL=0,15435
LCL=0,10800
I Chart of Amsterdam
Amsterdam
Percent
0,160,150,140,130,120,110,10
99
95
90
80
70
60
50
40
30
20
10
5
1
Mean 0,1312
StDev 0,008352
N 35
AD 0,285
P-Value 0,607
Probability Plot of Amsterdam
Normal - 95% CI
24. STEP 12: IMPLEMENTATION PROCESS
CONTROL
CONTROL
Conclusion: success on longer term is guaranteed!
Handover project + sign off
+ thanks team
Root cause Solution Assurance