Measurement and Analysis Interaction A1

1
Measurement and Analysis Interaction (A1)
YES
Are there new stratification factors to consider?
Study process and plan for measurement of Y
variables
Is there a stratification pattern?
NO
NO
YES
Hypothesize stratification factors
Hypothesis Generation for X (cause) variables
with no stratification
Any probable stratifications within the
re-focused data set?
NO
Measure
Hypothesis Generation for X (cause) variables
within stratification
Compute baseline metrics
YES
More measurements required?
NO
Measure
Data Exploration Find patterns related to
treatment differences (i.e. stratification)
YES
Verify Causes
Measure
2
Output of Measurement Selection (A1)

• Project Y variable(s) (CTQ) identified and linked
  to problem/goal
• At least one X variable (predictor) to help find
  cause of Y variable
• Start with stratification factor as initial type
  of X variable
• Plan for making sure you know
• Where to collect measurements
• Data is available
• It is feasible (time, money, personnel) to
  collect data
• Exercise CTQ Tree
• Exercise Measurement Assessment Tree

3
Output of Operational Definition (A2)

• Clear, concise, detailed, unambiguous
  description of what is being measured
• Definitions of key terms like defect, product and
  service
• Guidelines on how to interpret the routine and
  the unusual
• Initial data collection plan for what (sets up
  the when and how)
• Use Operational Definition Worksheet (pg. 169)

4
Output of Identifying Data Sources (A3)

• Identification of existing data sources that will
  meet some (or all) measurement needs. Criteria
  for acceptable existing data include
• Used the same operational definitions developed
  for the project collection efforts (especially in
  agreeing with customer definitions)
• Structured to support analysis stage (i.e. has
  required stratification factors)
• Identification of new data sources to needed to
  meet requirements
• Validating of ability to access and sort existing
  data

5
Prepare Data Collection and Sampling Plan (A4)

• Identify/confirm stratification factors
• Must begin with some idea of the end game
• Data exploration (analysis stage) lives or dies
  on decisions made here
• Develop sampling scheme
• Create data collection forms

6
Developing the Sampling Scheme (A4.2)

• Choice Population or Process sampling?
• Population sampling Large (essentially
  infinite), homogeneous pool of data
• Process sampling Sample taken from a running
  process stream
• Ref Tables 9-1 and 10-2 and Figures 10-6 to
  10-10
• Accounting for sampling bias
• Bad sampling processes convenience sampling and
  judgment sampling
• Good sampling processes systematic sampling,
  random sampling, stratified sampling
• Setting the Confidence Interval (CI) (Detailed
  discussion at end of Measure Stage of DMAIC
  model)
• Typical interval is set at 95 (this is Minitab
  default)
• Must know something about process to ballpark the
  sample size for a 95 CI
• Exercise Manual Sample size calculation (pg.
  171-172)

7
Creating Data Collection Forms (A4.3)

• Avoiding pitfalls
• KISS
• Good labeling
• Space for identifying data date, time, collector
• Have consistent structure
• Include key STRATIFICATION FACTORS
• Types of collection forms
• Check sheets
• Data sheets
• Travelers Excellent method to pair data when
  stratification factor and Y-variable measurement
  dont occur at same place and/or time

8
Output of Data Collection and Sampling Plan (A4)

• A list of stratification factors
• Completed sampling plan
• Data collection forms

9
Output of Implement/Refine Measurement Process
(A5)

• Review/finalize collection plan
• Perform Measurement System Analysis including
  Gage RR, bias assessment, stability and
  linearity testing, and calibration
• Prepare workplace Let all know whats going on
• Tested collection procedures
• KISS and trial run
• Validate collector training
• Collect data
• Monitor measurement accuracy and refine
• Exercise Gage RR Assessment (continuous and
  discrete)

10
Minitab Gage RR Example
11
Minitab Gage RR Session Window
Gage RR
Contribution Source VarComp
(of VarComp) Total Gage RR 0.0011386
98.84 Repeatability 0.0004267
37.04 Reproducibility 0.0007119
61.80 Operator 0.0006148
53.37 OperatorPart 0.0000972
8.44 Part-To-Part 0.0000133
1.16 Total Variation 0.0011519
100.00 Study Var
Study Var Source
StdDev (SD) (6  SD) (SV) Total Gage
RR 0.0337433 0.202460 99.42
Repeatability 0.0206559 0.123935
60.86 Reproducibility 0.0266823
0.160094 78.62 Operator
0.0247942 0.148765 73.05
OperatorPart 0.0098586 0.059151
29.05 Part-To-Part 0.0036515
0.021909 10.76 Total Variation
0.0339403 0.203642 100.00 Number of
Distinct Categories 1
12
Calculate Baseline Sigma Levels (B1)

• Key definitions
• Unit Item being processed (focus of the
  project)
• Defect Failure to meet customer expectation
• Defect Opportunity Chance for product/service
  to be defective
• Guidelines for defect opportunity definition
• Focus on defects that are important to the
  customer
• Should reflect number of places in the process
  where it can go wrong, NOT all the ways it can go
  wrong
• Focus on routine defects i.e. dont count the
  rare event
• Group similar defects in a single defect
  category
• Be consistent (within defect and across company)
• Dont change operation definition without
  compelling reason
• Simple 4-step process
• Exercise Sigma Calculation Worksheet (pg.
  178-179)

13
Calculate Final and First-Pass Yield (B2)

• Looks at the internal structure of the process
• Two different ways of looking at yield and
  process sigma final yield and first-pass yield
• Final yield
• How many defect-free items emerge at the end of
  the process including those that were
  successfully reworked
• Internal defects and their costs are hidden
• First-pass yield
• Number of items that make it through entire
  process without any rework included
• Same as Rolled Throughput Yield (RTY)

14
Measuring the Cost of Poor Quality (B3)

• Cost is connected to, but not the same as defect
  counts or sigma levels
• Translate defect data into Cost of Poor Quality
  (COPQ)

15
Output of Calculating the Performance Baseline

• Well defined units, defects and defect
  opportunity
• Calculated baseline sigma level
• Calculated final and/or first-pass yield for Y
  variable
• Identified labor and material rework costs
• Translated defects into dollars

16
Long-Term vs. Short-Term Variation

• Short-term variation is less than long-term
• Process shift adjustment of 1.5 sigma
• Short-term capability The best possible if
  process is centered
• Long-term capability Sustained reproducibility
  of the process
• The Z calculation and the Z table

17
Histogram and the Normal Distribution
Frequency of a Measurement Value
X
-S
-1.96S
-3S
S
1.96S
3S
(2S)
(-2S)
Measurement Value
18
The Z Table
Question 1 How many standard deviations are
there between the mean and the reference
measurement?
Distance from red dashed line to the mean
Z
One Standard Deviation
Z 1 std dev
Z
X
-S
-3S
S
3S
-1.96S
1.96S
Mean
(2S)
(-2S)
19
Z Table Examples
What percentage of measurements are to the left
of the red dashed line?
What percentage of measurements are to the right
of the solid yellow line?
Distance from red dashed line to the mean
Z
.5195
.5212
One Standard Deviation
.5212 - .5200
.5200 - .5195
Z
Z
.0010
.0010
Z 1.2
Z .5
P 11.51
P 30.85
X
-S
-2S
-3S
S
2S
3S
.5200
.5190
.5180
.5170
.5210
.5220
.5230
Inches
20
Z Table Exercise
What percentage of measurements are to the left
of the red dashed line?
What percentage of measurements are between the
dashed and solid lines?
Z
Z
.5192
.5208
Z
Z
X
-S
-2S
-3S
S
2S
3S
.5200
.5190
.5180
.5170
.5210
.5220
.5230
Inches