Professional Documents
Culture Documents
DMAIC GB A TX PG V4.2.0
D M A I C
DMAIC GB A TX PG V4.2.0
D M A I C
30 Analyze 6–continued 6
15 Improve Preview 7,8,9
30 Improve 7 and 8–Design Of Experiment 7,8
15 Break
30 Improve 9–Statistical Tolerancing 9
15 Control Preview 10,11,12
60 Control 12–Control Charts 12
DMAIC GB A TX PG V4.2.0
D M A I C
DMAIC GB A TX PG V4.2.0
D M A I C
Assessment Strategy
Introductions
1
DMAIC GB B TX PG V 4.2.0
D M A I C
DMAIC GB B TX PG V 4.2.0
D M A I C
DMAIC GB B TX PG V 4.2.0
D M A I C
DMAIC GB B TX PG V 4.2.0
D M A I C
DMAIC GB B TX PG V 4.2.0
D M A I C
DMAIC GB B TX PG V 4.2.0
D M A I C
DMAIC GB B TX PG V 4.2.0
D M A I C
DMAIC GB B TX PG V 4.2.0
D M A I C
DMAIC GB B TX PG V 4.2.0
D M A I C
DMAIC GB B TX PG V 4.2.0
D M A I C
DMAIC GB B TX PG V 4.2.0
D M A I C
DMAIC GB B TX PG V 4.2.0
D M A I C
DMAIC GB B TX PG V 4.2.0
D M A I C
DMAIC GB B TX PG V 4.2.0
D M A I C
DMAIC GB B TX PG V 4.2.0
D M A I C
DMAIC GB B TX PG V 4.2.0
D M A I C
DMAIC GB B TX PG V 4.2.0
D M A I C
DMAIC GB B TX PG V 4.2.0
D M A I C
5 Relate and apply the Quality Function Deployment (QFD) process to Six
Sigma.
Explain the purpose of QFD.
Describe the phases of QFD.
Explain QFD flowdown for product and service applications.
Generate/build a House of Quality (Product Planning Chart).
Identify what the customer wants (the “what’s”).
Identify the functions or processes that impact customer wants (the how’s).
Evaluate the impact of each function/process on customer wants.
Calculate the overall magnitude of the impact each function/process has on
customer wants (prioritize actions).
Analyze and diagnose a completed House of Quality.
Describe other QFD applications.
Determine when QFD is appropriate to use.
Recognize QFD pitfalls.
Describe an example of QFD from GE Medical Systems.
DMAIC GB B TX PG V 4.2.0
D M A I C
DMAIC GB B TX PG V 4.2.0
D M A I C
DMAIC GB B TX PG V 4.2.0
D M A I C
DMAIC GB B TX PG V 4.2.0
D M A I C
11 Describe and define the use of gage R& R for discrete data.
Describe the use of the attribute R & R spreadsheet for discrete
data.
DMAIC GB B TX PG V 4.2.0
D M A I C
DMAIC GB B TX PG V 4.2.0
D M A I C
DMAIC GB B TX PG V 4.2.0
D M A I C
2 Set realistic and achievable defect reduction goals based on current baseline,
GE guidelines, benchmarking results, and the process entitlement.
Describe the methods used to set project goals.
Use GE Standards for defect reduction in combination with benchmark
results to determine project goals.
Use the process entitlement to validate the achievability of the project goals.
DMAIC GB B TX PG V 4.2.0
D M A I C
DMAIC GB B TX PG V 4.2.0
D M A I C
4 Use a Pareto Chart to separate the vital few from the trivial many
in a process to determine where to focus improvement efforts.
Recognize the purpose and benefits of a Pareto Chart.
Describe the Pareto Principle.
Describe the steps involved in building a Pareto Chart:
Collect data.
Total results and arrange data in descending order.
Draw and label a Pareto Chart.
Analyze results.
Compare before and after Pareto Chart to evaluate
improvement effectiveness.
DMAIC GB B TX PG V 4.2.0
D M A I C
DMAIC GB B TX PG V 4.2.0
D M A I C
DMAIC GB B TX PG V 4.2.0
D M A I C
DMAIC GB B TX PG V 4.2.0
D M A I C
DMAIC GB B TX PG V 4.2.0
D M A I C
DMAIC GB B TX PG V 4.2.0
D M A I C
DMAIC GB B TX PG V 4.2.0
D M A I C
DMAIC GB B TX PG V 4.2.0
D M A I C
DMAIC GB B TX PG V 4.2.0
D M A I C
DMAIC GB B TX PG V 4.2.0
D M A I C
DMAIC GB B TX PG V 4.2.0
D M A I C
DMAIC GB B TX PG V 4.2.0
D M A I C
DMAIC GB B TX PG V 4.2.0
D M A I C
DMAIC GB B TX PG V 4.2.0
D M A I C
DMAIC GB B TX PG V 4.2.0
D M A I C
DMAIC GB B TX PG V 4.2.0
D M A I C
DMAIC GB B TX PG V 4.2.0
D M A I C
DMAIC GB B TX PG V 4.2.0
D M A I C
DMAIC GB B TX PG V 4.2.0
D M A I C
DMAIC GB C TX PG V 4.2.0
D M A I C
Classroom Session–Objectives 2
DMAIC GB C TX PG V 4.2.0
D M A I C
Iterative Process 3
Steps A,B,C
Steps
10,11,12 Steps 1,2,3
The Phases of the 12-Step Process Phase 4 (Improvement). This phase is usually initiated by
selecting those product performance characteristics which
Phase 1 (Define). This phase defines the project. It must be improved to achieve the goal. Once this is done, the
identifies customer CTQ’s and ties them to business characteristics are diagnosed to reveal the major sources of
needs. Further, it defines a project charter and the variation. Next, the key process variables are identified by
business process bounded by the project. way of statistically designed experiments. For each process
variable which proves to be significant, performance
Phase 2 (Measure). This phase is concerned with specifications are established.
selecting one or more product characteristics; i.e.,
dependent variables, mapping the respective process, Phase 5 (Control). This phase is related to ensuring that the
making sure the measurement system is valid, making
new process conditions are documented and monitored via
the necessary measurements, and recording the results.
statistical process control methods. After a “settling in”
period, the process capability would be reassessed.
Phase 3 (Analyze). This phase entails estimating the Depending upon the outcomes of such a follow-up analysis,
short and long-term process capabilities and it may be necessary to revisit one or more of the preceding
benchmarking the key product performance metrics. phases.
Following this, a gap analysis is often undertaken to
identify the common factors of successful performance;
i.e., what factors explain best-in-class performance.
© GE Capital, Inc., 2000
DMAIC GB C TX PG V 4.2.0
D M A I C
Statistical Thinking 4
D M A I C
Practical Statistical Statistical Practical
Problem Problem Solution Solution
Problem Characterize the Root cause Verify critical
statement process analysis X’s and ƒ(x)
– Project Y – Stability – Critical X’s Change
– Magnitude – Shape Measure the process
– Impact – Center influence of the Control the
– Variation critical X’s on gains
Data Integrity the mean and – Risk
– MSA Capability variability analysis
– Brainstorm – ZBench ST & LT – Test – Control
potential X’s – Model plans
– Sampling plan – Estimate
Collect data
The Practical-To-Statistical-To-Practical
Transformation Process
DMAIC GB C TX PG V 4.2.0
D M A I C
DMAIC GB C TX PG V 4.2.0
D M A I C
DMAIC GB C TX PG V 4.2.0
D M A I C
Meeting Skills 7
Roles:
Leader or Facilitator
Timekeeper
Scribe
Note Taker
Tools:
Agenda
Desired Outcomes
Ground Rules
Decision-Making Process
Brainstorming
DMAIC GB C TX PG V 4.2.0
D M A I C
* N/3 is a prioritization technique used to narrow a large list down to the top
priorities. The number of ideas (N) is divided by 3 (N/3) and team members get
that number of votes to identify their top choices. All votes should not be put on
one choice.
DMAIC GB C TX PG V 4.2.0
D M A I C
DMAIC GB D TX PG V 4.2.0
D M A I C
2
First Things First (10 minutes)
DMAIC GB D TX PG V 4.2.0
D M A I C
Techn
ical St
rate g y
rat egy
iza tional St
an
al/Org
Cultur
Change Initiative
(Target)
Q x A = E
DMAIC GB D TX PG V 4.2.0
D M A I C
Cap Model 4
Leading Change
Creating
Creating A
A Shared
Shared Need
Need
Shaping A Vision
Mobilizing
Mobilizing Commitment
Commitment
Current Transition Improved
State State State
Making
Making Change
Change Last
Last
Monitoring Progress
DMAIC GB D TX PG V 4.2.0
D M A I C
Measure
Analyze
Control
DMAIC GB D TX PG V 4.2.0
D M A I C
Homework 6
Desired Outcome: Identify Cap Tools which you will use in your project
DMAIC GB D TX PG V 4.2.0
D M A I C
D M A I C
DEFINE
PHASE
OVERVIEW
DMAIC GB D TX PG V 4.2.0
D M A I C
DMAIC GB D TX PG V 4.2.0
D M A I C
DEFINE STEP
OVERVIEW
DMAIC GB D TX PG V 4.2.0
D M A I C
10
Define B–Define Team Charter
B.1 Define the business case B.5 Select project team and define roles
B.2 Develop problem statement B.6 Develop charter
B.3 Develop goal statement B.7 Get sign-off for team charter
B.4 Assess project scope
DMAIC GB D TX PG V 4.2.0
D M A I C
DEFINE STEP
OVERVIEW
DMAIC GB D TX PG V 4.2.0
D M A I C
Desired Outcome: Share elevator speech for the project with members of your
team
Elevator speeches will be shared with the entire class throughout the 3-day workshop.
DMAIC GB D TX PG V 4.2.0
D M A I C
Elevator Speech 13
This is a tool that helps the team members put together a It helps the team link the need for change with the vision
short “sell” pitch for the project. Being able to clearly and of the future. All the team members will be using a
simply state the need for change and describe the future common “sell” pitch. For teams that have thoroughly
state is essential for rallying the support and commitment debated and documented both the need and vision, this is
of key constituents. the synthesis event whereby they distill the essence of the
project. For teams struggling to get started in terms of
The metaphor of the elevator is useful in challenging the need and vision, it can be the place to begin to bring
team to be clear, precise and simple. some focus to the team’s more rambling discussion of
need and/or vision.
Imagine a chance meeting of a team member and a key
stakeholder in an empty elevator with a 90-second ride. The elevator speech should be able to deal with the
questions that will arise once the project is announced to
the broader constituent base.
Describe the need for change and the vision of the new
state, as one might respond to the question, “Why are we
doing this project?”
DMAIC GB D TX PG V 4.2.0
D M A I C
DMAIC Checklist–Define A 14
DMAIC GB D TX PG V 4.2.0
D M A I C
DMAIC Checklist–Define B 15
DMAIC GB D TX PG V 4.2.0
D M A I C
B.5 Roles
Are all functions represented?
Does each team member fully represent their manager’s input on the project?
DMAIC GB D TX PG V 4.2.0
D M A I C
DMAIC Checklist–Define C 17
DMAIC GB D TX PG V 4.2.0
D M A I C
Patent
Best Protection against independent development by others
Gives right to exclude others from making, using selling, the invention of 20
years from filing
Can wait 2-3 years for patent to issue
Right granted by the government for new, useful, non-obvious inventions
Trade Secret
Rights exist only so long as actually kept secret
Protects against theft or misappropriation, not independent development
Does not require inventiveness in the patent sense
Copyright
Protects original works of authorship fixed in a tangible medium
Does not protect against independent development
Trademark
Defines sources of goods or services
Arbitrary, fanciful marks get stronger protection
DMAIC GB D TX PG V 4.2.0
D M A I C
Note: When writing your Business Case be certain to consider and include IP
issues.
DMAIC GB D TX PG V 4.2.0
D M A I C
DMAIC GB D TX PG V 4.2.0
D M A I C
DMAIC GB E TX PG V 4.2.0
D M A I C
D M A I C
MEASURE
PHASE
OVERVIEW
Measure 3:
Measure 2: Establish Data
Measure 1:
Define Collection Plan,
Select CTQ
Performance Validate
Characteristic
Standards Measurement
System, & Collect
Data
DMAIC GB E TX PG V 4.2.0
D M A I C
3
The12-Step Process
DMAIC GB E TX PG V 4.2.0
D M A I C
DMAIC GB E TX PG V 4.2.0
D M A I C
5
Mapping Tools To Generation Of CTQ Components
Define A
Output
Output Unit
Unit
C-O-P-I-S
C-O-P-I-S
Define A
Output
Output
Characteristic
Characteristic C-O-P-I-S
C-O-P-I-S
(VOC) Operational
Operational QFD
QFD
C&E
Definition
Definition
C&E
C-O-P-I-S
C-O-P-I-S
Measure 1
QFD
QFD
CTQ
CTQ Project
Project YY
Measure
Process
ProcessMap
C&E
C&E
Map
Measure Pareto
Pareto
Measure 2
Specification
Specification VOC
Limits
Limits VOC
QFD
QFD
QFD Measure 2
Target
Target QFD
Pareto
Pareto
VOC
VOC
Measure 2
Defect
Defect FMEA
FMEA
C&E
C&E
##of
ofDefect
Defect Measure 2
FMEA
Opportunities
Opportunities FMEA
Process
ProcessMap
Map
Per
Per Unit
Unit
DMAIC GB E TX PG V 4.2.0
D M A I C
Statistical Thinking 6
D M A I C
Practical Statistical Statistical Practical
Problem Problem Solution Solution
Problem Characterize the Verify critical
statement process X’s and ƒ(x)
– Project Y – Stability Change
– Magnitude – Shape process
– Impact – Center Control the
– Variation gains
Data Integrity – Risk
– MSA Capability analysis
– Brainstorm – ZBench ST & LT – Control
potential X’s plans
– Sampling plan
Collect data
The Practical-To-Statistical-To-Practical
Transformation Process
DMAIC GB E TX PG V 4.2.0
D M A I C
1
Measure 1–Select CTQ Characteristics
MEASURE STEP
OVERVIEW
1.1 Identify measurable CTQ characteristics that will be improved (Project Y).
DMAIC GB F TX PG V 4.2.0
D M A I C
2
Mapping Tools To Generation Of CTQ Components
Define A
Output
Output Unit
Unit
C-O-P-I-S
C-O-P-I-S
Define A
Output
Output
Characteristic
Characteristic
C-O-P-I-S
C-O-P-I-S
(VOC) Operational
Operational QFD
QFD
C&E
Definition
Definition C&E
C-O-P-I-S
C-O-P-I-S
Measure 1
QFD
Project
Project YY QFD
CTQ
CTQ Measure
Measure
Process
ProcessMap
C&E
C&E
Map
Pareto
Pareto
Specification
Specification Measure 2
Limits
Limits VOC
VOC
QFD
QFD
QFD Measure 2
Target
Target QFD
Pareto
Pareto
VOC
VOC
Measure 2
Defect
Defect FMEA
FMEA
C&E
C&E
##of
ofDefect
Defect Measure 2
Opportunities
Opportunities FMEA
FMEA
Process
ProcessMap
Per
Per Unit
Unit Map
DMAIC GB F TX PG V 4.2.0
D M A I C
1
CTQ Tools Module Objectives
DMAIC GB G TX PG V 4.2.0
D M A I C
Process Mapping
DMAIC GB G TX PG V 4.2.0
D M A I C
3
Process Mapping–CTQ Tool #1
A Graphical Representation of
Steps, Events, Operations, and Relationships
of Resources Within a Process
DMAIC GB G TX PG V 4.2.0
D M A I C
4
Levels Of A Process
Acquire New
Core Business
Underwriters Customer/
Customer Service
Detailed
Subprocess
Map
Tasks Procedures
Earlier in Define, you developed a high-level or COPIS If the Project Y is a cost measure, which of the blocks
process map. By looking at a process from a “big picture” adds the most cost?
perspective, you evaluated customer needs and supplier
inputs, and determined initial measurement objectives. If the Project Y is a function measure, which block has
the most errors or problems?
Now, you will look in more detail at the subprocesses
defined in the COPIS map. Subprocess maps provide Like working with a puzzle, you begin to assemble the
specifics on the process flow that you can then analyze pieces of an area on which it makes sense to focus our
using several useful techniques. Choose what to efforts.
subprocess map by determining which of the major steps
in the COPIS have the biggest impact on the output (Y’s).
The block (or blocks) selected is the one on which you
create a subprocess map–using it to understand how and
why it impacts the output.
DMAIC GB G TX PG V 4.2.0
D M A I C
5
Versions Of A Process
What You Think It Is... What It Really Is.. What It Should Be... What It Could Be...
Subprocess mapping is simply drawing a picture of the Reconciling the map requires your team to observe the
process–documenting the flow of the process. But, there process–making sure you have recorded all the existing
are at least four major versions of a process map that you steps.
can draw.
As the team moves forward and does process analysis
First, you can document what individuals who touch the and problem-solving, eventually you will move toward
process think it is. Certainly with the daily contact, you the third version of the process map–the “should be”
would expect people to know how the process works. But map created as part of their Improvement plan. At this
many people can easily explain how things work when point, a check must be made as to whether customer
things go right. You need to know how the process works CTQ’s have been met or exceeded.
in all conditions, so you need to go further.
DMAIC GB G TX PG V 4.2.0
D M A I C
6
Outside-In Focus
DMAIC GB G TX PG V 4.2.0
D M A I C
7
Subprocess Mapping Techniques
Process flowchart
Alternate path method
Deployment or cross-functional map/flowchart
DMAIC GB G TX PG V 4.2.0
D M A I C
8
Process Mapping Symbols
DMAIC GB G TX PG V 4.2.0
D M A I C
9
Process Flowchart
Customer Call
Automated Touch- Rep
Customer Yes Chooses Placed In
Criteria Customer
Prepare Yes
System Tone Answers
Calls Routing Que, On
Met? LoanWaits?
Offer
Answers Phone Phone
Option Hold
No
Call Gets No
Routed To
Call
Voice-Activated
Ends
System
Key:
Start Review Or
Task Decision Direction
End
DMAIC GB G TX PG V 4.2.0
D M A I C
10
Alternate Path Flowchart
85 Customer Call 80
Automated Touch- Rep
Customer Yes Chooses Placed In
Criteria Customer
Prepare Yes
System Tone Answers
Calls Routing Que, On
Met? LoanWaits?
Offer
Answers Phone Phone
Option Hold
15
No
20
Call Gets No
Routed To
Call
Voice-Activated
Ends
System
Key:
60
40
Direction
Task Percentages Decision
DMAIC GB G TX PG V 4.2.0
D M A I C
11
Deployment Or Cross-Functional Flowchart
Sign File
Customer Negotiate
Contract Contract
Review Review
Attorney
Contract Contract
Write File
Dealer Contract Contract
DMAIC GB G TX PG V 4.2.0
D M A I C
12
Steps In Deployment Mapping
For Your Process/Subprocess: 3. Next, identify who receives the output of the second
step and what activity they perform. Write this activity
1. Identify the participants involved in the process, on a card and place it to the right of the second step on
write their names on cards, and pin or tape the cards the same row as the participant performing the activity.
in a vertical column on the wall with the customer at
4. Identify who receives the output of this activity and what
the top of the column.
activities that participant performs.
2. Identify the action that initiates the process, write it 5. Continue to create cards and position them in the same
on a card, and place it next to the appropriate manner until the process is complete (i.e., the customer
participant card. In many instances,the process is receives the final process output).
initiated by the customer but it is not always the case.
Also, identify the final step in the process and place in
the appropriate row at the end of the map. Here again,
the last step may be the customer’s action.
DMAIC GB G TX PG V 4.2.0
D M A I C
13
Tips In Subprocess Mapping
Here are some guidelines on building a subprocess map. Combine and Clarify–Make sure brainstormed steps
These are not absolute–but they should help you avoid are clear and don’t overlap.
some of the pitfalls of process mapping.
Organize in “Flow”–Creating the map is last. With all
Focus on “As Is”–To find out why problems are steps visible, it’s typically much easier to create a
occurring in a process, you need to concentrate on meaningful map without getting stuck on one or two
how it’s working now. minor issues.
Clarify Boundaries–If you’re working from a well-
Is the customer involved in your subprocess step?
done high-level map, this should be easy. If not, you’ll
If so, how?
need to clarify start and stop points.
Brainstorm Steps–It’s usually much easier to identify
the steps before you try to build the map. Starting each
step description with a verb (e.g., “Collate Orders”;
“Review Credit Data”) helps you focus on action in the
process. Who does the step is best left in parentheses
(or left out) –you want to avoid equating a person with
the process step.
DMAIC GB G TX PG V 4.2.0
D M A I C
14
Verifying CTQ Elements For Process Mapping
DMAIC GB G TX PG V 4.2.0
D M A I C
15
Subprocess Mapping–Breakout Activity (20 minutes)
DMAIC GB G TX PG V 4.2.0
D M A I C
16
Cause & Effect Diagrams–CTQ Tool #2
4. Brainstorm possible causes and attach them to Use the Affinity Diagram and brainstorm items into
appropriate branch. categories.
DMAIC GB G TX PG V 4.2.0
D M A I C
17
Cause & Effect Diagrams–Example
DMAIC GB G TX PG V 4.2.0
D M A I C
18
FMEA–CTQ Tool #3
DMAIC GB G TX PG V 4.2.0
D M A I C
19
Potential Failure Modes and Effects Analysis
Worksheet
Process/Product: Invoicing FMEA Date: (Original) 10/15/96
FMEA Team: Invoice Process Mgnt. Team (Revised) 5/25/97
Black Belt: E. Jones Page: 1 of 1
Occurrence
Occurrence
Responsibility
Detection
Detection
Severity
Severity
Potential Potential
Recommended And Target
RPN
RPN
Item/Process Potential Current Action
Effect(s) Of Cause(s) Of Completion
Step Failure Mode Controls Action Taken
Failure Failure Date
Enter
Amt Inaccurate Overbill 8 Wrong Ct 8 Rare 9 576 DMAIC Team E. Jones All DMAIC 8 4 2 64
Owed Wrong Price 5 “ 9 360 to Investigate 5/15/97 Tasks 8 3 2 48
Root Causes Complete
Underbill 5 Wrong Ct 8 Rare 9 360 of Count & 5 1 2 10
Wrong Price 5 “ 9 225 Price 5 1 3 15
Accessories
Missing No Payment 6 Sale Error 5 Reviewed 3 90 6 5 3 90
De Error 7 “ 3 126 6 7 3 126
De Error
Delay 3 Sale Error 5 Reviewed 3 45 3 5 3 45
De Error 7 “ 3 63 3 7 3 63
De Error
Delayed Late Bill 3 Sale “Too 7 Measured 4 84 3 7 4 84
Busy”
System Down 3 “ 4 36 3 3 4 36
Total Risk Priority Number 2,205 Resulting Risk Priority Number 821
DMAIC GB G TX PG V 4.2.0
D M A I C
20
Failure Mode And Effects Analysis
Failure Mode
Effects Controls
Severity Detectability
(1-10) (1-10)
Causes
Occurrence
(1-10)
RPN
RPN
Risk
Risk Priority Number
Priority Number
RPN
RPN = S x O x D == 11 to
= S x O x D to 1000
1000
Review the product, service, or process Assign Severity, Occurrence, and Detection Factors
Determine failure modes Calculate RPN.Prioritize RPNs from high to low.
List one or more potential effects for each failure Identify the top issues, those with high RPN’s.
mode. Answer the question: “If the failure occurs, Determine preventive or remedial actions, especially
what are the consequences?” for high-priority issues.
Identify potential causes Recalculate RPN after actions have been implemented
List current controls
DMAIC GB G TX PG V 4.2.0
D M A I C
21
FMEA: Calculating Risk Priority Number (RPN)
4
is likely to result in a complaint
Cause minor performance loss
Occurrence Scale
3 Cause a minor nuisance, but be Rating Time Period Probability
overcome with no performance loss
10 More than once per day > 30%
Good
2 Be unnoticed and have only minor
9 Once every 3-4 days ≤ 30%
Detection Scale
effect on performance
1 Be unnoticed and not affect the 8 Once per week ≤ 5% Rating Definition
performance 7 Once per month ≤ 1% 10 Defect caused by failure is not detectable
6 Once every 3 months ≤ .03% 9 Occasional units are checked for defect
8 Units are systematically sampled and inspected
5 Once every 6 months ≤ 1 Per 10,000
7 All units are manually inspected
4 Once per year ≤ 6 Per 100,000
6 Units are manually inspected with mistake-proofing
3 Once every 1-3 years ≤ 6 Per Million modifications
2 Once every 3-6 years ≤ 3 Per 10 Million 5 Process is monitored (SPC) and manually inspected
1 Once every 6-100 years ≤ 2 Per Billion 4 SPC is used with an immediate reaction to out-of-
control conditions
3 SPC as above with 100% inspection surrounding out -
of-control conditions
2 All units are automatically inspected
1 Defect is obvious and can be kept from affecting the
customer
DMAIC GB G TX PG V 4.2.0
D M A I C
22
Using A FMEA In The Measure Phase
DMAIC GB G TX PG V 4.2.0
D M A I C
23
Pareto Charts–CTQ Tool #4
C A E D B
Category of Defect
DMAIC GB G TX PG V 4.2.0
D M A I C
24
Pareto Charts (continued)
Cumulative Percentage
140 70
f* of D+B
120 60
Frequency
f* of D
100 50
LEGEND
80 40
A: Illegible
B: Bank Info Incomplete
60 30 C: Missing Signature
D: Personal Information
Incomplete
40 20 E: Employment History
Incomplete
20 10 F: Address Incomplete
0
Defect Type D B F A C E Other* Total
# Defects 105 40 20 10 5 3 25 208
% Defects 50.5 19.2 9.6 4.8 2.4 1.5 12.0 100
Cum % 50.5 69.7 79,3 84.1 86.5 88.0 100.0
Approximately 80% of Defects from Defects D + B + F
DMAIC GB G TX PG V 4.2.0
D M A I C
25
Using Pareto Charts In The Measure Phase
DMAIC GB G TX PG V 4.2.0
D M A I C
26
QFD–CTQ Tool #5
DMAIC GB G TX PG V 4.2.0
D M A I C
27
QFD Flowdown–Product Application
Functional
Requirements Part Characteristics
(HOW’s)
GE
Requirements
(HOW’s) Processes
House
Customer
(WHAT’s)
of Process
Requirements
Quality
(WHAT’s)
of
Characteristics
#1 (HOW’s)
Quality
(WHAT’s)
#2 Part House
Processes
of
(WHAT’s)
House
Quality
GE
of
Y #3
Quality
Y
#4
Key Functional
Requirements
Key Part
Characteristics X
Key
Manufacturing Key
Processes Process
Variables
DMAIC GB G TX PG V 4.2.0
D M A I C
28
QFD Flowdown–Services Application
Service
Requirements
(HOW’s) Service
Functions/Processes
(HOW’s)
Customer Wants
Process Controls
House (HOW’s)
(WHAT’s)
of
Requirements
Quality
House
(WHAT’s)
#1
Functions/Proceses
Service
of
Quality
(WHAT’s)
#2 House
Service
of
Quality
Y #3
Critical-to-Quality
Characteristics
(CTQ’s) Key X
Service
Processes Key
Process
Variables
DMAIC GB G TX PG V 4.2.0
D M A I C
29
House Of Quality Summary
Roof
Roof
1b. Correlation
Correlation Target
1b. Customer
Customer Target
Importance
Importance Direction
Conflict Direction
Conflict
How
How important
important Resolution
are Resolution
arethe
thecustomer
customer
wants
wants to
to the
the
Current Rating
Competitor #1
Competitor #2
customer?
customer?
3.Characteristics/measures
3.Characteristics/measures
HOW'S
HOW'S
How
How do
do you
you satisfy
satisfythe
thewants
wants??
1a.
1a. Customer
Customer H L L M
Needs
Needs H
WHAT'S
WHAT'S M M L 2.
2. Competitive
Competitive
H Assessment
Assessment
What
Whatdoes
doesthe
the
customer
customer
want?
L M Where
Whereare
arewe
competitors
weand
and
want? competitorsrelative
relativeto
to
M L H customer importance?
customer importance?
Voice
Voiceof
of the
the
Customer
Customer L M
4.4.Relationships Importance
Relationships ImportanceRatings
Ratings
HH Strong
Strong 99
MM Medium
Medium 3
3 6.
6. Target
Target values
values of
of HOW's
HOW's (units)
(units)
LL Weak
Weak 11
5.
5. Competitive
Competitive
Benchmarks
Benchmarks
How
How do
do competitors
competitors
perform
performrelative
relative to
to each
each
HOW?
HOW?
It is not necessary to develop every room for every house of To what extent will a HOW impact a WHAT? Using a 9, 3, and
quality you build. 1 scale forces a wider spread between the most important and
Room 1a least important items.
Customer Needs (in their language) Room 5
High-level wants identified by the customer Competitive Benchmarks
Room 1b How the competition performs relative to the
Customer Importance measures/characteristics.
Customer ranking of the wants. A typical Room 6
score is 1-5. Target and Specifications
1–least important Target values for the HOW’S. Setting measurable targets allows
5–most important the team to define what is required to achieve customer
Room 2 satisfaction.
Competitive Components Roof
Customer’s view of how GE compares with the competition Relationship
Room 3 Impact of the HOW’S on each other.
Characteristics/Measures ++ Strong positive
Measurable attributes that can be used to indicate how well the + Positive
customer’s needs are met. - Negative
Room 4 -- Strong Negative
Relationships
Strengths of interrelation between the WHAT’S and the
Importance Ratings (IR)
HOW’s. H–Strong (9)
IR = Σ (Customer Importance X Strength Of Relationship)
M–Medium (3)
L–Weak (1)
© GE Capital, Inc., 2000
DMAIC GB G TX PG V 4.2.0
D M A I C
Cost of ingredients
Customer Needs Weight of portion
PB&J Sandwich
Time to prepare
Instant Soup
Importance
Fast Food
Direction for improvement
Fills us up 5 H M M 2 3 4
Is nutritious 4 M H L L 3 3 2
Tastes good 3 H L 2 3 4
Is easy to make 4 H H 2 1 4
Is easy to clean up 2 M H 2 3 5
Sticks with us 4 H M 2 2 4
Is inexpensive 1 L L L M 4 3 1
Is clean 2 M M H 2 3 4
DMAIC GB G TX PG V 4.2.0
D M A I C
31
Analyzing A House Of Quality
Empty
Empty columns
columns in in Strong
Strong negative
negative relationship
relationship
Room
Room 33 -- Perhaps
Perhaps an an in
unnecessary
unnecessary in roof –Identify
roof –Identify trade-offs
trade-offs
measure
measure
characteristic
characteristic
signaling
signaling itit does
does not
not
affect
affect customer
customer
Current Rating
Competitor #1
Competitor #2
wants
wants
L L M
H Highest
Highest score
score on
on
competitive
competitive
M M L comparison
comparison -- AbleAble to
to
lead
lead the
the market
market with
with
existing
existing
L M product/service
product/service
L H
L M
Empty Low
Low score
score on
on competitive
competitive comparison,
comparison, but
but
Emptyrows
rowsininRoom
Room33- - high
Unaddressed customer high score
score on
on competitive
competitive benchmarks
benchmarks --
Unaddressed customer Market
Market technical
technical advantages
advantages to
to improve
improve
want
wantwould
wouldbebeaamajor
major
problem
customer
customer perception
perception
problem
Low
Low competitive
competitive benchmarks
benchmarks -- Poor
Poor long-term
long-term market
market performance
performance
DMAIC GB G TX PG V 4.2.0
D M A I C
Measure 1–Select CTQ Characteristics
Define A
Here’s our next step: Output
Output Unit
Unit AA
Lunch
LunchPortion
Portion
Output
Output Define A
Characteristic
Characteristic Healthy
HealthyLunch
Lunch
Project
Project YY Measure 1
Operational Total
TotalWeight
Weight
Operational In
Definition InOunces
OuncesPer
PerServing
Serving
High-Level Definition
Need (VOC)
Measure 1
Project
Project YY
Measure
Measure Weight
Weight
Prepared
Prepared
Lunch
CTQ
Lunch
Measure 2
Specification LSL=14
LSL=14oz.
oz.
Specification USL=18
Limits* USL=18oz.
oz.
Limits*
Measure 2
Target*
Target* 16
16oz.
oz.
DMAIC GB G TX PG V 4.2.0
D M A I C
Methodology Step: Measure 1–Select CTQ Characteristics
33
CTQ And Performance Standard Worksheet
Define A
Output
Output Unit
Unit AA
Lunch
LunchPortion
Portion
Define A
Output
Output
Characteristic Healthy
Healthy
Characteristic
Lunch
Lunch
Project
Project YY Measure 1
Operational
Operational %
%Carbohydrate
Carbohydrate
High-Level Definition
Definition Per
PerServing
ServingPer
PerUSDA
USDA
Need (VOC)
Measure 1
Project
Project YY
Measure
Measure %
%Carbohydrate
CarbohydrateContent
Prepared
Prepared Content
Lunch
CTQ
Lunch
Measure 2
Specification
Specification
Limits*
Limits* >>20%
20%
Measure 2
Target*
Target*
25%
25%
Measure 2
Defect
Defect**
You may not know Spec, Target
or Defect at this time (in Step 1). <<20%
20%
If you know it now, write it down.
This information will come from
the VOC. ##of
ofDefect
Defect Measure 2
Opportunities
Opportunities
Per
Per Unit
Unit 11
DMAIC GB G TX PG V 4.2.0
D M A I C
34
Another Example
# Tools Used
# Completed
Assessment
Customer
Exam Pass
Importance
Knowledge
Project Per
In Project
Needs
Quarter
Score
Rate
Learn The
Material
5 9 3 3 1
Pass The
Exam
4 3 9 1 1
Apply This
Material To Project
2 1 1 9 3
Importance
Rating 59 53 37 15
Key
High = 9
Medium = 3
Low = 1
DMAIC GB G TX PG V 4.2.0
D M A I C
35
Points To Remember
DMAIC GB G TX PG V 4.2.0
D M A I C
Methodology Step: Measure 1–Select CTQ Characteristics
36
CTQ And Performance Standard Worksheet
Define A
Output
Output Unit
Unit Knowledge
Knowledge
Assessment
Assessment
Define A
Output
Output GB
GBKnowledge
Knowledge
Characteristic
Characteristic Level
Level
Project
Project YY %
%Correct
Correct== Measure 1
Operational
Operational Total
TotalCorrect
Correct
High-Level Definition
Definition
Questions
QuestionsDivided
DividedBy
By100
100
Need (VOC)
Knowledge
Knowledge Measure 1
Project
Project YY Assessment
AssessmentScore
Score
Measure
Measure (Percentage
(PercentageCorrect)
Correct)
Learn
Learn
Six
CTQ
Six Sigma
Sigma
Measure 2
Specification
Specification ≥≥80%
80%
Limits*
Limits*
Measure 2
Target*
Target* 85%
85%
Measure 2
You may not know Spec, Target Defect
Defect**
or Defect at this time (in Step 1). <<80%
80%
If you know it now, write it down.
This information will come from
the VOC. ##of
ofDefect
Defect Measure 2
Opportunities
Opportunities 11
Per
Per Unit
Unit
© GE Capital, Inc., 2000
DMAIC GB G TX PG V 4.2.0
D M A I C
37
Common QFD Pitfalls
QFD on everything
– Set the “right” granularity
– Don’t apply to every last project
Lack of teamwork
– Wrong participants
– Lack of team skills
– Lack of support or commitment
Too much “chart focus”
“Hurry up and get done”
Failure to integrate and implement QFD
DMAIC GB G TX PG V 4.2.0
D M A I C
38
QFD Activity
Background:
Capital Logistics has established a customer help desk to handle
customer inquiries regarding deliverables. However, customers are
dissatisfied with the help desk.
DMAIC GB G TX PG V 4.2.0
D M A I C
DMAIC GB G TX PG V 4.2.0
D M A I C
40
QFD Activity
HOWS
Service Characteristics
or good on customer
Rep (CSR) excellent
resolved on first call
% customers rating
Number of hours in
Customer Service
resolve problems
survey of (CSR).
courteousness
Cycle time to
% problems
Importance
available
Quick problem resolution 5
WHATS
Customer Wants
Courteous Service 3
Importance rating
Units % Hrs.
Hrs. %
days
Use L=1
M=3
H=9
DMAIC GB G TX PG V 4.2.0
D M A I C
41
CTQ Tools: Summary
DMAIC GB G TX PG V 4.2.0
D M A I C
DMAIC GB G TX PG V 4.2.0
D M A I C
MEASURE STEP
OVERVIEW
DMAIC GB H TX PG V 4.2.0
D M A I C
Operational Definitions 2
Purpose
To remove ambiguity
Everyone has a consistent understanding
DMAIC GB H TX PG V 4.2.0
D M A I C
Features
What: Must have specific and concrete criteria
How: Must have a method to measure criteria
Must be useful to both you and the customer
(the “wing-to-wing” concept)
The clock stops when the phone caller notes time of completed
application decision notification in desk log
4
Why Start Our Measurement Work With The Project Y Metric?
The Project Y is the key metric for the project and is defined
from the customer’s perspective
A customer-focused Project Y measure allows us to assess the
process from the customer’s perspective
Changes in the values of a good Project Y measure will predict
changes in the degree to which the process meets the customer
CTQ’s
A Project Y measure expresses the “voice of the process”–
changes in input or process conditions drive changes in true
output measures
DMAIC GB H TX PG V 4.2.0
D M A I C
Hold Time
(Project Y Measure) Established in Measure, Step 1
Customer Need
In Measure, Step 2, we are completing our CTQ Element Specification: the variation around the target the
Tree. Use the template on the next page or the one in customer will accept.
your GB Workbook.
Defect: failure to meet specification. Anything
that results in customer dissatisfaction.
Definitions:
Characteristic: a word or phrase that describe some
aspect of the product or service.
DMAIC GB H TX PG V 4.2.0
D M A I C
6
Mapping Tools To Generation Of CTQ Components
Define A
Output
OutputUnit
Unit C-O-P-I-S
C-O-P-I-S
AACall
Call
Define A
Output
Output Timely
TimelyCall
CallPick-up
Pick-up
Characteristic
Characteristic C-O-P-I-S
C-O-P-I-S
Definition Pareto
Pareto Customer
CustomerService
Service
Definition Pick-Up
Pick-UpInInSecs.
Secs.
Customer
Customer Measure 1
waits
waits QFD
QFD
on CTQ
CTQ Project
ProjectYY
Process
ProcessMap
Map HOLD
HOLDTIME
TIME
on Hold
Hold Measure
Measure
C&E
C&E
Pareto
Pareto
too
too Long
Long
Measure 2
Specification
Specification VOC
VOC USL=20
USL=20Seconds
Seconds
Limits
Limits QFD
QFD
QFD
QFD Measure 2
Target
Target Pareto
Pareto 10
10Seconds
Seconds
VOC
VOC
Measure 2
Defect
Defect FMEA
FMEA >>20
20Seconds
Seconds
C&E
C&E
##of
ofDefect
Defect Measure 2
Opportunities
Opportunities FMEA
FMEA 11
Process
ProcessMap
Per
PerUnit
Unit Map
DMAIC GB H TX PG V 4.2.0
D M A I C
A performance standard
translates customer needs
into quantified requirements
for our product or process Output
Output Timely Resolution
Characteristic
Characteristic of a Call
Quick
Quick
Answer CTQ
CTQ Target
Target
5 Minutes or
Answer less
Specification/
Specification/ Not Greater
tolerance
tolerance Than 60
limit(s)
limit(s) (USL/USL)
(USL/USL) Minutes (USL)
Any Response
Defect
Defect Taking More Than
60 Minutes
DMAIC GB H TX PG V 4.2.0
D M A I C
DMAIC GB H TX PG V 4.2.0
D M A I C
Once the Project Y metric has been established, the target and
specification limits can be established.
Output
Output
Characteristic Timely response
Characteristic
Specification/
Specification/ Variance not
tolerance
tolerance greater than
limit(s)
limit(s) one day
Any response
Defect
Defect greater than
1 day
Target/Nominal Value
Optimal value for Y
Specification/Tolerance Limit(s)
Window or range of acceptable output values for Y
Also called “performance standards”
Information about where to set target and specification
limits comes from the customer. This information can
be collected not only from VOC but also from:
– Contractual agreements
– Industry standards
– Customer observations
– Process data
DMAIC GB H TX PG V 4.2.0
D M A I C
Project Y 10
DMAIC GB H TX PG V 4.2.0
D M A I C
Project Y Alignment 11
Project Y
Key project metric defined
from the customer’s
perspective
X1 X2 X3
DMAIC GB H TX PG V 4.2.0
D M A I C
Desired Outcome
Practice defining units and defect for your CTQ
Define units For your process output, agree on the Facilitator 5 mins.
and defect unit and defect for your Project Y. Use
the worksheet on page 17 of your GB
Workbook.
DMAIC GB H TX PG V 4.2.0
D M A I C
1
Minitab and Graphical Analysis Module Objectives
DMAIC GB I TX PG V 4.2.0
D M A I C
2
Minitab Windows
Menu
Menu Bar
Bar
Session
Session Window:
Window:
•• Analytical
Analytical Output
Output
Data
Data Window:
Window:
•• A
A Worksheet,
Worksheet, not
not aa
Spreadsheet
Spreadsheet
Graph
Graph Window
Window
Data Window–Stores data to be used for analysis. You can move through the windows using Ctrl-Tab, the
Discussed in more detail on next page. windows menu, or the following shortcuts.
DMAIC GB I TX PG V 4.2.0
D M A I C
3
Data Window
Column Names
Are Entered Here
Data is Entered
Here
Scroll Bars
Data window is a worksheet, not a spreadsheet. If the value you enter into the first row is something
Variables are usually entered in columns, observations other than a number or date, Minitab assumes that the
in rows. entire column will be text (T).
Each column has a title area which is 31-characters or
less and this title must be unique.
Everything in a column is considered to be the same
variable.
Column names are above the column, not in the first
row.
Minitab cells contain values only, not formulas, like
Excel.
To perform mathematical functions, use
Calc > Calculator.
The orientation of the data arrow determines whether
the enter key will take you to the next row or the next
column.
DMAIC GB I TX PG V 4.2.0
D M A I C
4
Minitab Menus–Summary
File Menu Print and save the window that is currently active
File menu changes depending on the window that is currently active
Allows open, close, and save
Edit Menu Similar to the edit menu in most standard Windows applications
Manip. Menu Sort, code or manipulate data
Six Sigma* Product & process report can be used for determining performance levels
Menu (Calculating Sigma)
Save Project As–When you save a project, you save all If you use the icon, only projects will be shown. You
the worksheet, session, info, history and graph windows must use the file > open worksheet to view all Minitab
that are currently open. Minitab will use the file extension worksheets.
of .mpj.
*Six Sigma add-on in Minitab can be obtained via a link
Save Current Worksheet As–When you save as a in support central.
worksheet, you save only the worksheet portion (you will
lose session window and graphs). Minitab will use the
file extension of .mtw.
DMAIC GB I TX PG V 4.2.0
D M A I C
5
Using Minitab: A Typical Session
1. Enter data
2. Select menu command (for desired statistical/graphical
function)
3. Enter command parameters in the dialog window
4. View results in session window or graph window
5. Copy output to another application
6. Print output
7. Save file
X Y
Factor Response
Independent Dependent
Input variable
By variable variables
graph variables
DMAIC GB I TX PG V 4.2.0
D M A I C
6
Using Minitab: A Typical Session
DMAIC GB I TX PG V 4.2.0
D M A I C
7
Using Minitab: A Typical Session
Type data
directly into
worksheet
DMAIC GB I TX PG V 4.2.0
D M A I C
8
Using Minitab: A Typical Session
Importing Data: When importing from Excel, an Excel file can contain
only one worksheet and the information contained in
To determine whether a file can be opened by Minitab, Row 1 will become the titles in Minitab.
choose Open Worksheet from the File menu. Notice that
the area labeled Files of Type contains a drop-down Note: During import, Minitab will convert percentages,
menu. This is indicated by the arrow (triangle) at the right dollars, etc.
of the Files of Type selection box.
DMAIC GB I TX PG V 4.2.0
D M A I C
9
Using Minitab: A Typical Session
3.Go to Minitab:
ALT > Tab
Insertion point
DMAIC GB I TX PG V 4.2.0
D M A I C
10
Using Minitab: A Typical Session
DMAIC GB I TX PG V 4.2.0
D M A I C
11
Using Minitab: A Typical Session
DMAIC GB I TX PG V 4.2.0
D M A I C
12
Using Minitab: A Typical Session
Above is a set of data that has been captured and entered Note: For additional practice, see the Green Belt Training
into an Excel spreadsheet. Support Central Site.
To make the data suitable for use in Minitab:
DMAIC GB I TX PG V 4.2.0
D M A I C
13
Using Minitab: A Typical Session
DMAIC GB I TX PG V 4.2.0
D M A I C
14
Using Minitab: A Typical Session
DMAIC GB I TX PG V 4.2.0
D M A I C
15
Using Minitab: A Typical Session
DMAIC GB I TX PG V 4.2.0
D M A I C
16
Using Minitab: A Typical Session
Copying graphs:
1. Make sure the graph window is active in Minitab
2. Click right mouse button and select “Copy Graph”
3. Open the application into which you want to copy the graph or table,
e.g., Microsoft Word
4. Paste graph using Paste from the toolbar or Ctrl-V
Note: You can also copy graphs from the window >
manage graphs.
DMAIC GB I TX PG V 4.2.0
D M A I C
17
Using Minitab: A Typical Session
DMAIC GB I TX PG V 4.2.0
D M A I C
18
Recruitment Cycle Time Case Study
We will use the following case study to practice basic graphical analysis in Minitab.
You are working on a project to reduce cycle time for filling temporary positions in the
Information Technology department within the company. Currently, you work with one
agency who finds temporary contractors to fill positions on request by the IT department.
The performance standard for cycle time to fill these positions is 20 (working) days
maximum. This is measured from the time the agency receives the request to the time
the position is filled.
You have completed the Define and Measure phases. In the Measure phase, you collected
cycle time data for a sample of positions filled within the last 6 months.
The data is contained in file fill_time.mtw. The column labels are as follows:
C1: yrs exp–years experience of the person who filled the position
C2: agent–specifies the agent who processed the request and found the resource
C3: site–specifies the company site from which the request originated (there are three sites in
the area who request temporary resources)
C4: type of resource–specifies the type of resource requested (DBA, Programmer,
Systems Analyst)
C5: cycle time–time to fill position (working) days
DMAIC GB I TX PG V 4.2.0
D M A I C
19
Graphical Analysis Of Data
Key Questions:
How is my data distributed (variation)?
What relationships exist between the
Y variable and X variables?
DMAIC GB I TX PG V 4.2.0
D M A I C
20
Review–Variation
DMAIC GB I TX PG V 4.2.0
D M A I C
21
Using Data To Understand Variation
Measurement
Frequency
Measurement Time
For Continuous Data For Discrete Data For Continuous Data For Discrete Data
DMAIC GB I TX PG V 4.2.0
D M A I C
22
Review–Continuous vs. Discrete Data
OR
–The exact time it took to process each application?
© GE Capital, Inc., 2000
DMAIC GB I TX PG V 4.2.0
D M A I C
23
5 M’s & 1 P
Sources Of Variation
Machines P
Methods R
Materials O
Measurements
C
E
Mother Nature
S
People
S
Machines–The various appliances used in the transformation In a training session, failure to regulate the thermostat
from inputs into outputs. For example, a PC can turn various can result in a non-conducive learning environment.
sources of information into an organized manual that then
relates to a training service. People–The staffing that influences customer needs
and requirements. While often dominant in the
Methods–The procedures, formal or otherwise, that transform service industry, this is an area still too often blamed
inputs into outputs. For example, there is a standard procedure for failure to meet or exceed customer requirements.
for billing collections in the GE Capital businesses.
Sometimes in a service environment, the following
Materials–The components, tangible or otherwise, that are categories are used:
transformed from inputs into outputs. For example, paper stock
and ink quality will affect a product brochure’s quality. — Policies–Higher level decision rules or
management practices.
Measurements–The tools that monitor a process’sperformance. — Procedures–The way in which tasks are performed.
For example, a doctor’s blood pressure reading of a patient — Plant–The building, equipment, work space, and
would determine subsequent activity related to treatment. environmental factors that affect performance.
— People–The human element.
Mother Nature–The environmental elements within a process
that influence a customer need or requirement.
— Parts–Systems, documents, and other supplies that
are needed to perform the service.
DMAIC GB I TX PG V 4.2.0
D M A I C
24
Two Types Of Variation
Always Present
Common Cause Expected
Predictable
Normal
There are two types of variation: The distinction between common and special causes is
important to determine the basic strategy for process
Common causes
improvement and control.
Special causes
DMAIC GB I TX PG V 4.2.0
D M A I C
25
Describing Variation For A Period Of Time: Data Distributions
Key Questions:
What is the shape of the distribution–symmetrical, skewed,
twin peaks, flat?
What is the central tendency (“center”)
of the distribution?
What is the variation (“spread”) of the distribution–wide or
narrow?
DMAIC GB I TX PG V 4.2.0
D M A I C
26
Statistics
DMAIC GB I TX PG V 4.2.0
D M A I C
27
Some Distributions
Uniform Distribution
P(X) Single roll of a die
Triangular Distribution
Sums of a pair of dice
P(X)
Rapidly descending
P(X), no tails
X
DMAIC GB I TX PG V 4.2.0
D M A I C
28
Distributions
Normal Distribution
P(X)
Process/repair times
Exponential Distribution
Time between arrivals
P(X)
Time between random
(unrelated) failures
DMAIC GB I TX PG V 4.2.0
D M A I C
29
Shape
7 Center 20
6
Frequency
5
Frequency
4
10
3
2
1
0 0
12 13 14 15 16 17 18 19 0 10 20 30 40 50 60 70 80 90
Bimodal Distribution
Spread Spread
6 Center Center
5
Frequency
4
3
2
1
0
7 9 11 13 15 17 19 21 23
DMAIC GB I TX PG V 4.2.0
D M A I C
30
The Normal Curve
Definition:
A probability distribution is where the most frequently occurring value is in the
middle and other probabilities tail off symmetrically in both directions.
Characteristics:
The curve does not reach zero
The curve can be divided in half with equal pieces falling either side of the most
frequently occurring value
The peak of the curve represents the center of the process
The area under the curve represents 100% of the product the process is capable of
producing
DMAIC GB I TX PG V 4.2.0
D M A I C
31
The Normal Curve (continued)
Specific Characteristics
34.13% 34.13%
13.60% 13.60%
2.14% 2.14%
0.13% 0.13%
95.46%
99.73%
The normal curve can be divided into a series of Most common statistical terms and analysis tools are
segments. Each segment is mathematically called a based on a normal data distribution. If we use these tools
standard deviation from the mean. It is also noted by the with a data set that is not normal, the accuracy of the
small s. tools may be compromised.
DMAIC GB I TX PG V 4.2.0
D M A I C
32
Normal Probability Plot
.999
.99
.95
Probability
.80
.50
.20
.05
.01
.001
2 12 22 32
Cycle Time
Average: 16.3921 Anderson-Darling Normality Test
StDev: 5.61675 A-Squared: 0.208
N: 240 P-Value: 0.864
A normality test is a statistical process used to determine There are two occasions when you should use a
if a sample, or any group of data, fits a standard normal normality test:
distribution. A normality test can be done mathematically — When you are first trying to characterize raw data,
or graphically. normality testing is used in conjunction with
graphical tools such as histograms and box plots.
A normality test can be thought of as a “litmus test” for — When you are analyzing your data, and you need to
determining if a distribution is a normal distribution. calculate basic statistics such as Z values or to
employ statistical tests that assume normality, such
The Y axis is the cumulative % of data points which fall as t-Test and ANOVA.
below the value on the X axis.
Interpreting A Normal Probability Plot
Why Is A Normality Test Useful?
— When plotted data follows a straight line, the
Many statistical tests (tests of means and tests of Anderson-Darling p-value will exceed 0.05 and will
variances) assume that the data being tested is normally increase, I have normally distributed data.
distributed. A normality test is used to determine if that
assumption is valid. — Rule: p > .05 indicates data is normal.
© GE Capital, Inc., 2000
DMAIC GB I TX PG V 4.2.0
D M A I C
33
Normal Probability Plot (continued)
Distribution Type:
Normal Bimodal curve Skewed Long-Tailed
6 5 5
Frequency
5
Frequency
Frequency
4
Frequency
4
3 10
3 3
2 2
2
1 1
1
0
0
12 13 14 15 16 17 18 19 0 0
7 9 11 13 15 17 19 21 23
0 10 20 30 40 50 60 70 80 90 7 9 11 13 15 17 19 21 23
Normal Probability Plot for a Normal Probability Plot for a Normal Probability Plot for a Normal Probability Plot for
Normal Distribution Bimodal Distribution Skewed Distribution Long-Tailed Distribution
99 99
ML Estimates ML Estimates
ML Estimates
Mean: 14.6382
Mean: Mean: 15.0790
95 95
StDev: 5.47084
StDev: 90 StDev: 12.6232 90
80 80
70
Percent
Percent
70
Percent
60 60
Percent
50 50
40 40
30 30
20 20
10 10
5 5
1 1
0 10 20 30 0 10 20 30
In the top diagram on the left, we see that all of the data
points roughly form a straight line and fall within the
limits. We would conclude that there is no serious
departure from normality in this data. In the other
diagrams, we see that many of the data points fall outside
the limits and do not form a straight line. We would
therefore conclude that these data sets significantly depart
from the normal distribution.
DMAIC GB I TX PG V 4.2.0
D M A I C
34
What If Your Data Is Not Normal?
DMAIC GB I TX PG V 4.2.0
D M A I C
35
Minitab–Histogram
Frequency
20
Tool: Histogram
10
Data Type: Continuous. The X axis is cycle time,
the Y axis is the number of times
cycle time fell within the range of each 0
bar or interval in the histogram. The Y 0 10 20 30
axis is calculated for you.
Cycle Time
Data A single column of data for each
Structure: histogram
DMAIC GB I TX PG V 4.2.0
D M A I C
36
Minitab–Normal Probability Plot
Datafile: Fill_time.mtw
.999
.99
Question: Are the cycle time data .95
normally distributed?
Probability
.80
.50
.20
Tool: Normality Test
.05
.01
2 12 22 32
DMAIC GB I TX PG V 4.2.0
D M A I C
37
“Center” Or Central Tendency
Mean (X)
x
Median (“middle” data point)
Long-tailed Distribution
Quartile Values (Q1, Q3)
Skewed Distributions
Q1 Q3
DMAIC GB I TX PG V 4.2.0
D M A I C
38
“Center” Or Central Tendency (continued)
X=
∑X i
1. The sum of all data
values
DMAIC GB I TX PG V 4.2.0
D M A I C
39
Variation
p=.01 p=.99
Skewed Distributions
Q1 Q3 Q1 Q3
DMAIC GB I TX PG V 4.2.0
D M A I C
Variation (continued) 40
∑ (X − X )
square root of
2 B- square each
the result
i difference Q Q
1 3
n −1 C- sum the
squared
differences
D- divide by 1 less than the
number of data values
Q More Q
1 1 Less
0 Variatio 1 Variatio
Q n Q
3 3 n
DMAIC GB I TX PG V 4.2.0
D M A I C
41
The Computational Equations
Population N
Mean ∑ Xi
i =1
µ=
N
Population N
∑ (Xi − µ)
2
Variance 2 i =1
σ =
N
Population
Standard N
∑ (Xi − µ)
2
Deviation σ= i =1
N
n
Sample
∧
∑X i
Mean µ=X= i =1
Sample ∑ (X
n
− X)
2
Variance ∧ i =1
i
σ 2 = s2 =
n−1
Sample
∑ (X − X)
n
Standard 2
∧ i =1
i
Deviation σ = s=
n−1
DMAIC GB I TX PG V 4.2.0
D M A I C
42
Deviation
Deviation==Measures
Measures Of Variation
Of Variation
Random data (“1” to “99”, n = 10 samples):
n Xi Xbar Dev =(Xi-Xbar) (Xi-Xbar)^2
1 20 35.7 -15.70 246.49
2 1 35.7 -34.70 1204.09
3 6 35.7 -29.70 882.09
4 35 35.7 -0.70 0.49
5 54 35.7 18.30 334.89
6 67 35.7 31.30 979.69
7 43 35.7 7.30 53.29
8 99 35.7 63.30 4006.89
9 5 35.7 -30.70 942.49
10 27 35.7 -8.70 75.69
----- ------ ---------- --------------
357 0.0 8726.10
DMAIC GB I TX PG V 4.2.0
D M A I C
43
Summary–Measures
Summary–Measures Of Variation
Of Variation:
Note that for any given set of numbers (Xis), the value
for SST is fixed, unchanging. The only way to change
SST is to change one or more of the numbers–which
means that the spread around the mean will likely be
changed.
DMAIC GB I TX PG V 4.2.0
D M A I C
44
Variation For A Period Of Time
Descriptive Statistics Summary
Standard
Percent
ML Estimates
Mean: 15.0790
StDev:12.6232
Q1 or Q3 (SF)
skewed
Normal Probability Plot for a Long-Tailed Distribution
99 ML Estimates
Mean: 14.6382
( )
95
StDev:5.47084
~
90
Percent
80
70
60
50 Median X Span
40
30
20
10
5
long-tailed 1
0 10 20 30
Normal Probability Plot for a Bimodal Distribution
99 ML Estimates
80
70
60
50
40
before descriptive statistics
can be calculated.
30
20
10
5
bimodal 1
0 10 20 30
DMAIC GB I TX PG V 4.2.0
D M A I C
45
Displaying Variation For A Period Of Time
Histogram
Measurements Graphical Display
Time Time Histogram of Time Estimates
Estimates Estimates
Round 1 Round 2
16.48 13.84 # of Occurrences 10
18.89 13.50
13.18 15.41
11.11 14.35
14.67 14.37
16.53 14.63 5
14.79 13.58
18.06 14.75
14.48 11.95
14.89 14.36 0
15.63 16.17
10 11 12 13 14 15 16 17 18 19
13.95 15.15
13.74 12.48 Time Estimates (in seconds)
17.67 14.12
10.23 19.00
13.67 13.81
11.35 12.97
15.03 14.19 Illustrates
Shape (pattern) of the data
Central tendency (center) of the data
Variation (spread) of the data
DMAIC GB I TX PG V 4.2.0
D M A I C
46
Displaying Variation For A Period Of Time (continued)
Box Plots
* Outlier
Highest Value
Lowest Value
Outlier calculations
High: Q3 + 1.5 (Q3-Q1)
Low: Q1 - 1.5 (Q3-Q1)
DMAIC GB I TX PG V 4.2.0
D M A I C
47
Summary–Variation For A Period Of Time
10
Data
28 23 13 34 24 29
21 16 24 11 49 21 Histogram
21 25 26 27 27 29
30 29 30 20 10 30 5
12 11 27 23 24 28
17 9 30 29 29 28
10 11 12 13 14 15 16 17 18 19
DMAIC GB I TX PG V 4.2.0
D M A I C
48
Minitab–Graphical Summary
To Make A Stat > Basic Statistics > Display 16.0 16.5 17.0 17.5
15.6779 17.1063
95% Confidence Interval for Sigma
Histogram: Descriptive Statistics: 5.1552 6.1698
95% Confidence Interval for Median
This graphical summary is one of the most widely used Skewness–A measure of asymmetry. A value more
charts in Minitab. than or less than zero indicates skewness in the data.
Useful Items: But, a zero does not necessarily indicate symmetry.
Anderson-Darling normality TEST–We use the p- Kurtosis–A measure of how different a distribution is
value to determine whether our distribution is normal from the normal distribution. A negative value
or non-normal. If p > 0.05, the distribution is normal. typically indicates a distribution more peaked than the
(we will discuss p-value & hypothesis testing in normal. A positive value typically indicates a
Analyze, Step 6.) distribution flatter than normal.
Mean–Average of the samples. N–sample size.
StDev–Standard deviation of the samples. Confidence Interval for Mu–Tells us the interval in
which the true population mean lies (with 95%
Variance–(Standard Deviation Squared.) A measure of confidence.)
how far the data are spread about the mean.
DMAIC GB I TX PG V 4.2.0
D M A I C
49
Variation Over Time
Run Chart
A graphical tool to monitor the “stability of Project Y
Allows observation of time order properties such as trend
Should be used before any detailed data analysis
Example of a Run
Chart
Median
DMAIC GB I TX PG V 4.2.0
D M A I C
50
Run Charts–Special Cause Patterns
If p < 0.05, then there is significant statistical evidence to show that one of the
trends below exists.
Mixture Cluster
Oscillating Trend
The Minitab Run Chart tests for two kinds of Runs: Oscillation represents more than expected number of
(A) Consecutive points on the same side of the median. runs of type (B).
(Thus a new run starts when the median is crossed).
(B) Consecutive points in the same direction up or down. Trend Pattern
(Thus, a new run starts when the direction is changed).
A trend represents less than expected number of runs
of type (B).
Mixture Pattern
At this point in time, our biggest concern is whether
A mixture represents more than expected number of runs the process is stable or not. We use the above-
of type A (above). mentioned checks to indicate stability. Minitab will do
this for us.
Cluster Pattern
DMAIC GB I TX PG V 4.2.0
D M A I C
51
Minitab–Run Chart
Cycle Time
22
40 140 240
Observation
Data Structure: A single column of data is all that Number of runs about median: 124.000 Number of runs up or dow n: 151.000
is necessary. It is assumed that Expected number of runs:
Longest run about median:
121.000
9.000
Expected number of runs:
Longest run up or dow n:
159.667
4.000
the data is entered in the order in Approx P-Value for Clustering:
Approx P-Value for Mixtures:
0.651
0.349
Approx P-Value for Trends:
Approx P-Value for Oscillation:
0.091
0.909
which it is collected (time
order).
DMAIC GB I TX PG V 4.2.0
D M A I C
52
Two Types Of Variation
DMAIC GB I TX PG V 4.2.0
D M A I C
53
Two Types Of Variation (continued)
Why is it important to know the source of variation and treat it Treating special causes as common causes is essentially under-
according to the appropriate strategy? Because not reacting reacting to individual data points. The problem with making
appropriately to the type of variation present in a process can this type of mistake is missing an opportunity to prevent defects
seriously impact customer satisfaction and the amount of and reduce process variation. An example of this type of
variation and defects, and it can increase costs. mistake would be failing to correct the steering on a car
heading for a ditch, and ending up in the ditch.
When interpreting variation, you can make two types of Appropriately reacting to the source of variation in a process
mistakes: provides the correct economic balance between over-reacting
Treating common causes as special causes and under-reacting to variation from a process.
Treating special causes as common causes
DMAIC GB I TX PG V 4.2.0
D M A I C
54
Graphical Analysis Tools
Continuous Y Discrete Y
Boxplot
Pareto Chart
Scatterplot
DMAIC GB I TX PG V 4.2.0
D M A I C
55
Box Plots
30
10
DMAIC GB I TX PG V 4.2.0
D M A I C
56
Minitab–Box Plots
DMAIC GB I TX PG V 4.2.0
D M A I C
57
Scatter Diagrams–Analyzing Relationships
40
Cycle Time (Days) (Y)
35
30
25
20
15
10
5
1K 2K 3K 4K 5K 6K 7K 8K 9K 10K
DMAIC GB I TX PG V 4.2.0
D M A I C
58
Warning! Correlation Does Not Imply Causation
70 70
Population
(In Thousands)
60 60
50 50
100 200 300
Number Of Storks
Source: Box, Hunter, Hunter. Statistics For Experimenters. New York, NY: John Wiley & Sons. 1978
DMAIC GB I TX PG V 4.2.0
D M A I C
59
Common Construction Mistakes
Examples
Problem Variable Axis Variable Axis
# Of Errors In
Bank
Size Of Loan Closing Cost
Errors
Estimate
Length Of Time To
Loan Delays # Of Changes
Approve Loan
DMAIC GB I TX PG V 4.2.0
D M A I C
60
Interpretation Of A Scatter Diagram
Look For:
Common patterns in the data
Range of the predictor variable (X)
Irregularities in the data pattern
DMAIC GB I TX PG V 4.2.0
D M A I C
61
Interpreting A Scatter Diagram
2 4 6
DMAIC GB I TX PG V 4.2.0
D M A I C
62
Minitab–Scatter Plot
Tool: Plot
Cycle Time 20
Experience Click OK
DMAIC GB I TX PG V 4.2.0
D M A I C
63
Pairs Exercise–Analyze Patterns
With A Partner:
Review scatter diagrams on following
pages as assigned
Determine the type of relationship you
observe (e.g., negative, positive, strong, weak, etc.)
Describe an example from your business
of this type of relationship
DMAIC GB I TX PG V 4.2.0
D M A I C
Strong,
Plot +/- Other Weak, Other
Example
1
Effect
Potential Cause
2
Effect
Potential Cause
3
Effect
Potential Cause
DMAIC GB I TX PG V 4.2.0
D M A I C
Strong, Weak,
Plot +/- Other Other
Example
4
Effect
Potential Cause
5
Effect
Potential Cause
6
Effect
Potential Cause
DMAIC GB I TX PG V 4.2.0
D M A I C
7
Effect
Potential Cause
8
Effect
Potential Cause
9
Effect
Potential Cause
DMAIC GB I TX PG V 4.2.0
D M A I C
10
Effect
Potential Cause
11
Effect
Potential Cause
12
Effect
Potential Cause
DMAIC GB I TX PG V 4.2.0
D M A I C
68
Pareto Charts
Frequency
C A E D B
Category of Defect
DMAIC GB I TX PG V 4.2.0
D M A I C
69
Minitab–Pareto Chart
Percent
Count
100
A tally table in two columns. 80
One column contains the Type 60 100
80
of Defect, the second column 40 60
contains the frequency of the defect. 20 40
20
0 0
To Make A Stat>Quality Tools>Pareto Chart Defect
Pareto Chart: Chart Defects Data In: Agent
4 1 2 3
To find out which site has the most defects: Follow steps Alternate method (subsetting a worksheet)
above except, place “Site” in “Chart Defects Data In” Box. MANIP > SUBSET WORKSHEET
Name: No Transactions
Alternate Method (Splitting A Worksheet)
Include or Exclude: CLICK ON “Specify which rows
to include”
MANIP > SPLIT WORKSHEET
Specify which rows to include:
By variable: Cycle_Time_Defect CLICK ON “Rows that Match”
Click: OK
CLICK ON “Condition” Button
Now you have 2 worksheets-one with “N” transactions Condition: Double Click on C6 then = “N”
and one with “Y” transactions (‘cycle_time_defect’) = “N”
Activate the correct worksheet and then follow the steps Click “OK” twice
in the slide above
Now you have a worksheet with “N” transactions
DMAIC GB I TX PG V 4.2.0
D M A I C
DMAIC GB I TX PG V 4.2.0
D M A I C
1
Measure
Define 3–Plan
ReviewData Collection And Collect Data
Objectives 1
MEASURE STEP
OVERVIEW
Data Collection
Define Review Plan Worksheet
Objectives
2
How to measure:
Sampling Plan (Scheme, Frequency, Size)
Measurement Procedure
Refer to more detailed documentation, if necessary.
Measurement System
Define Review Analysis (MSA) Worksheet
Objectives
3
Part p-value Y N
Oper p-value Y N
Oper & Part p-value Y N
Pass?
2) % Tolerance [ <30% ] Y N
3) % Contribution [ <8% ] Y N
4) % Study [ <30% ] Y N
5) # Distinct Categories [ >4 ] Y N
Develop
Establish Ensure Data Collect Data
1 Data Collection
Goals
2 Operational
Definitions And
Procedures
3 Consistency
And Stability 4 And Monitor
Consistency
Clarify purpose Write and pilot Test and validate Train data
of data collection operational definitions measurement collectors
Identify what systems
Develop and pilot Pilot process
data to collect
data collection forms and make
and procedures adjustments
DMAIC GB K TX PG V 4.2.0
D M A I C
Step 1: Establish
Define Review Data Collection Goals
Objectives
2
DMAIC GB K TX PG V 4.2.0
D M A I C
Segmentation:
– An analysis technique that involves temporarily dividing a large group of
data into smaller logical categories to look for areas of very good vs. very
poor performance
– Can be used to understand which X’s drive variation of the Project Y
Collect Project Y data to identify patterns and performance
trends and establish current baseline defect rate
Collect segmentation factor data to be used for later analysis
Think ahead about how you plan to analyze the For example, in deal businesses where the Project Y
data. Collect additional information is the cycle time of closing the deal, a potential
corresponding to the Project Y that may be segmentation factor is client size (large, medium, or
helpful in subsequent analysis. small).
– Include possible external factors that may .
be useful for segmenting the total Y data
set.
– Consider how to collect continuous data
instead of discrete data.
Thinking about segmentation factors now allows
you to gather these conditions along with data
on your Project Y so they don’t have to be
reconstructed after the measurement is
completed.
Segmentation factors include external variables
outside our control that may explain patterns in
the Project Y measures.
DMAIC GB K TX PG V 4.2.0
D M A I C
Common Segmentation
Define Review Factors
Objectives
4
DMAIC GB K TX PG V 4.2.0
D M A I C
Common Segmentation
Define Review Factors
Objectives
5
Other Categories
Factor Example
What type Complaints, defects, problems
When Year, month, week, day
Where Country, region, city, work site
Who Business, department, individual,
customer type, market segment
Tip: Begin with factors “outside” the process box–often these are factors that were not considered
when the process was first designed
DMAIC GB K TX PG V 4.2.0
D M A I C
How To Collect
Define Review Data For Segmentation
Objectives
6
DMAIC GB K TX PG V 4.2.0
D M A I C
Step 1: Breakout
Define Review Activity (5 Minutes)
Objectives
7
Time 5 Minutes
DMAIC GB K TX PG V 4.2.0
D M A I C
Step 2: Develop
Define ReviewOperation Definitions and Procedures
Objectives
8
DMAIC GB K TX PG V 4.2.0
D M A I C
Operational Definitions
Define Review Objectives
9
DMAIC GB K TX PG V 4.2.0
D M A I C
Operational Definitions–Scale
Define Review Objectives Of Scrutiny
10
DMAIC GB K TX PG V 4.2.0
D M A I C
1
Sampling
Define Review Objectives
1
x x x
x x
x x
x x x x x x
x x x x x
x
x x x
N = 5,000 n = 100
DMAIC GB L TX PG V 4.2.0
D M A I C
When To Review
Define Sample Objectives 2
When to sample
Collecting all the data is impractical or too costly
Data collection can be a destructive process
When measuring a high-volume process
When not to sample
A subset of data may not accurately depict the process, leading
to a wrong conclusion (every unit is unique – e.g., structured
deals)
DMAIC GB L TX PG V 4.2.0
D M A I C
Goal Of AReview
Define Useful Sample
Objectives
3
Representative Samples
Representative sample:
All parts of the target population are represented
(i.e., selected for measurement) equally
The customer’s view is captured
How to guarantee a representative sample:
Design a sampling strategy
Understand special characteristics of the population before
sampling
DMAIC GB L TX PG V 4.2.0
D M A I C
Sample
DefineSize For Continuous
Review ObjectivesData
4
Consider the following example: Therefore, from the statistical theory we can answer
We want to estimate average call length in handling according to the formula:
customer inquiries, and we want our estimate to be
accurate to within + 1 minute. 1.96 × σ
2
n=
∆
Based on a small random sample of 30 inquiries we
know that the variation in call length, as measured
by a statistic called the standard deviation, is 5 Where n = sample size, u = standard deviation and
minutes. ∆ = degree of precision required. In our example,
the required sample size is:
We want to have 95% confidence that the estimate n = [(1.96x5)/1] 2 = 96.04
will be in the range of specified accuracy – i.e., + 1 or 96 samples
minute.
DMAIC GB L TX PG V 4.2.0
D M A I C
Common Segmentation
Define Review Factors
Objectives
5
Other Categories
Factor Example
What type Complaints, defects, problems
When Year, month, week, day
Where Country, region, city, work site
Who Business, department, individual,
customer type, market segment
Tip: Begin with factors “outside” the process box–often these are factors that were not considered
when the process was first designed
DMAIC GB L TX PG V 4.2.0
D M A I C
Calculating A Sample
Define Review Size
Objectives
6
– σ Standard error
CI –X X ± Zα/2 * n
–
Given X ± ∆
σ
Solve for n ∆ = Zα /2 n
Z α /2 *σ
n =
∆ 2
n = (∆ )
Zα /2 *σ
n = sample size
∆ = precision of the estimate. How much error is
ok? The smaller ∆, larger the sample size. May be a
business decision.
CI = Confidence Interval
Zα/2 = Z score we usually set at 1.96 for 96%
confidence.
σ = estimated standard deviation
DMAIC GB L TX PG V 4.2.0
D M A I C
How To Estimate
Define Review Standard Deviation When It Is Unknown
Objectives
7
3 Ways To Estimate σ
Use an existing Xbar or R chart
σˆ = R /d2 where d2 is control chart factor
(
σˆ = UCL − X /3 )
Collect a small pre-sample & calculate s (n = 30)
Ask subject matter experts to take an educated guess at the
plausible range of data
n d2
2 1.128
3 1.693
4 2.059 σˆ = R /d 2
5 2.326
6 2.534
DMAIC GB L TX PG V 4.2.0
D M A I C
Sample
DefineSize For Continuous
Review ObjectivesData (continued)
8
Sample
n
Values
∆ ∆
⌧-∆ ⌧ ⌧+∆
Calculate average (⌧)
Conclusion:
Whenever samples are taken to estimate a Confidence intervals are important when precise
population there will be differences between the estimates of populations are required, and the
“true” population values and the sample values. degree of precision in the estimate needs to be
known.
Through statistical theory we can determine the
amount of variation we can expect in our estimates. For those statistically inclined, an important
This is known as a confidence interval. Confidence statistical theory supporting confidence intervals is
intervals are stated in terms of an interval and a the “central limits theorem”. The central limit
confidence level. theorem states that regardless the shape of the
population, sample averages will always be
For example, when estimating average inquiry “normally” distributed and inversely proportioned
resolution time from samples, we may obtain a 95% to the square root of the sample size.
confidence interval for the average 1.2 - 2.3 days.
This means that if repeated samples were taken
from the same population, 95 times out of 100 we
would expect the sample average to be between 1.2
and 2.3 days.
DMAIC GB L TX PG V 4.2.0
D M A I C
Finite Population
Define Review Correction
Objectives
9
3. Calculate n finite
nfinite = n/(1+n/N)
DMAIC GB L TX PG V 4.2.0
D M A I C
Sample
DefineSize For Discrete
Review Data
Objectives
10
DMAIC GB L TX PG V 4.2.0
D M A I C
Sample
DefineSize For Discrete
Review Data (continued)
Objectives
11
Sample
n
Values
Conclusion:
I know with 95% confidence that the population
proportion defective is P + ∆
Calculate Proportion
Defective (P)
Recalculate n* based on the calculated P. If the new required sample size (n*) is more than the number
of samples taken, take (n*-n) samples and recalculate P based on the full sample size. If it is not
practical to take more samples, then use the actual n and P to recalculate the actual precision (∆).
2
1.96
n= P(1 − P )
∆
Example
We want to estimate the defect rate (P) within +
0.02 (I.e., ∆ = 0.02). We expect P to be
approximately 0.05.
2
1.96
n= 0.05(1 − 0.05 ) = 456
0. 2
DMAIC GB L TX PG V 4.2.0
D M A I C
How To Estimate
Define Review P and ∆
Objectives
12
Estimate P
Take a small pre-sample of data (n = 100) and calculate P
Use X from an existing control chart
Set P = .5 as a worst case (largest n)
Estimate ∆
See estimation of ∆ for continuous data
To determine ∆ for a given sample size
Px (1− P )
∆ = Ζ α/2
n
0 .5 1
P
DMAIC GB L TX PG V 4.2.0
D M A I C
Sample
DefineSize Considerations
Review Objectives
13
DMAIC GB L TX PG V 4.2.0
D M A I C
Sampling ExerciseObjectives
Define Review
14
Time 10 Minutes
DMAIC GB L TX PG V 4.2.0
D M A I C
DMAIC GB M TX PG V 4.2.0
D M A I C
Observations
Inputs Process Outputs Inputs Measureme
nt Outputs Measuremen
ts
Process
Data
Accuracy
Accuracy Precision Discriminatio
Precision Discriminatio
(Bias)
(Bias) (Measureme nn
(Measureme (Resolution)
nt Error) (Resolution)
nt Error)
We will be evaluating the variation in our measurement Data is only as good as the process that
systems. We will include the variation due to the operator measures it
and gage (Accuracy, Precision and Discrimination). MSA identifies how much variation is present in
the measurement process
When collecting data, we are seeing the process through Understanding measurement variation is
the lens of our measurement system. We never really see necessary for identifying “true” process variation
the actual process variation-instead, we see the actual and maximizing true Y improvements
process variation and the measurement process variation. Without MSA, you run the risk of making
Therefore, the total observed variation can be broken decisions based on an inaccurate picture of your
down into 2 parts: the actual process variation and the process
variation (or measurement error) created by the
MSA helps direct efforts aimed at decreasing
measurement process.
measurement variation
Excessive measurement variation distorts our
understanding of what the customer feels
DMAIC GB M TX PG V 4.2.0
D M A I C
Measurement Error vs Process Variation make process improvements. The Minitab results
obtained from your Gage R&R will indicate where the
When conducting a Gage R&R, you will be looking at the largest sources of measurement variation are hiding.
Measurement Error and determining if it is at an Your goal will be to fix or reduce this variation and then
acceptable level. We want to minimize the measurement to re-run the Gage R&R. You cannot collect and/or use
error (or measurement process variation). By conducting your data until the Gage R&R passes.
a Gage R&R, you will know how much of the total
variation is due to the measurement process itself. In
other words, by knowing the Measurement Error, we can
assume that the variation we are observing in the process
is mainly due to the process variation and not due to the
measurement error. If the measurement error is found to
be unacceptable, the measurement process must be fixed
prior to collecting and/or analyzing data and trying to
DMAIC GB M TX PG V 4.2.0
D M A I C
Questions to ask:
What are you measuring?
Who is measuring?
What do you use to measure (gage)?
How are you measuring? Do you always use the same
procedure?
Under what conditions are you measuring (6 m’s)?
What is the resolution of your measurement system? Does the
customer measure the same way?
Which analysis will you use (continuous, Attribute and/or
Destructive)?
DMAIC GB M TX PG V 4.2.0
D M A I C
Performing An MSA 5
DMAIC GB M TX PG V 4.2.0
D M A I C
Reproducibility–variation when two or more people measure the same unit with
the same measuring equipment
Stability–variation obtained when the same person measures the same unit
with the same equipment over an extended period of time
DMAIC GB M TX PG V 4.2.0
D M A I C
Conduct MSA
Yes
Continue Process
Improvement
For Attribute R&R:
DMAIC GB M TX PG V 4.2.0
D M A I C
Attribute = Discrete
DMAIC GB M TX PG V 4.2.0
D M A I C
Guidelines for determining how to design an AR&R study: The examples that follow balance all
1. For accuracy: The sample size is the total number three characteristics of an AR&R study.
of measurements. (40 parts x 3 repeated measures
x 3 operators = 360 measures.)
DMAIC GB M TX PG V 4.2.0
D M A I C
DMAIC GB M TX PG V 4.2.0
D M A I C
This data sheet uses 3 operators, 3 trials, and 40 units to An extra column is added to define the true state of the
measure repeatability, reproducibility and accuracy. standard according to a subject matter expert. In this
Repeatability of the measurement system is assessed example, N = Non-defective and D = Defective.
by determining the proportion of times each operator
matches on (e.g., three) repeated measures of one unit.
For this example, there are 120 opportunities for a
repeatability match.
Reproducibility of the measurement system is assessed
by determining the proportion of times all operators
match on (e.g., nine) repeated measures of one unit.
For this example, there are 40 opportunities for a
reproducibility match.
Accuracy of the measurement system is assessed by
determining the number of times each individual
measure matches a standard. For this example, there
are 360 opportunities for an accuracy match.
DMAIC GB M TX PG V 4.2.0
D M A I C
DMAIC GB M TX PG V 4.2.0
D M A I C
DMAIC GB M TX PG V 4.2.0
D M A I C
DMAIC GB M TX PG V 4.2.0
D M A I C
DMAIC GB M TX PG V 4.2.0
D M A I C
√
Attribute Gage R&R
1) % Repeatability [ >90% ] 89.17% OK
2) % Reproducibility [ >90% ] 52.5% Need to improve. See plan below
3) % Accuracy [ >90% ] 88.89% OK
Gage R&R Pass? Y √ N, If NO:
Plan for improvement: Will work on Reproducibility problem by conducting a team meeting
with the 3 operators and discussing the differences in their measurements. Suspect that additional
training is needed for less experienced operators because the data showed less experienced
operators answering differently. Will also investigate if there is anything unique with each of the
samples where reproducibility was an issue.
DMAIC GB M TX PG V 4.2.0
D M A I C
If you fail the Gage R&R, here are some factors that could
cause measurement process variation (measurement error).
Repeatability
Operational definitions
Maintain stability
Reproducibility
Operational definitions
Consistent use of gage
Training
Varying work environment
AUDIT-follow-up on training
Human/physical characteristics
Performance measures
Unclear requirements
DMAIC GB M TX PG V 4.2.0
D M A I C
Accuracy
Error in master
Instrument used improperly by appraiser
Operational definition
Standard not understood
DMAIC GB M TX PG V 4.2.0
D M A I C
Other Considerations 19
Temporal Effects
Gage R&R depends upon our ability to measure things multiple
times
If the item is an event, it may not happen the same way twice.
For this reason, Gage R&R may be impossible.
If the event is recorded, it may be possible to conduct a Gage
R&R study
If the event cannot be recorded, but multiple judges can
observe at once, reproducibility can be estimated, but
repeatability cannot
DMAIC GB M TX PG V 4.2.0
D M A I C
Other Considerations 20
DMAIC GB M TX PG V 4.2.0
D M A I C
Other Considerations 21
DMAIC GB M TX PG V 4.2.0
D M A I C
Other Considerations 22
DMAIC GB M TX PG V 4.2.0
D M A I C
Test-Retest Study
DMAIC GB M TX PG V 4.2.0
D M A I C
Test–Retest Study 24
Best Practice Hint: Do Test-Retest before Gage R&R–a quick look at the
situation.
How:
Repeatedly measure the same item
Same conditions, operator, device, and location on item–same, same,
same
Completely mount and dismount item for each measurement–exercise gage
through full range of normal use
DMAIC GB M TX PG V 4.2.0
D M A I C
DMAIC GB M TX PG V 4.2.0
D M A I C
Conduct MSA
Yes
Continue Process
Improvement
DMAIC GB M TX PG V 4.2.0
D M A I C
Target Analogy
== Xbar
==
True Value
==
True Value
DMAIC GB M TX PG V 4.2.0
D M A I C
Measurement System 28
Observations
Inputs Outputs Inputs Measurement Outputs Measurements
Process Process Data
#1
Measurement
System Variability
- Investigated
through “R&R
Study”
DMAIC GB M TX PG V 4.2.0
D M A I C
DMAIC GB M TX PG V 4.2.0
D M A I C
R&R
EV AV
The variation due to the equipment and appraiser do not Reproducibility: variation when two or more
directly add up to determine the total (R&R) variation. people measure the same unit with the same
There is an overlap between EV and AV. You can use the measuring equipment.
Pythagorean theorem (right angles) to add to the
influence of each EV and AV to calculate the total
variation.
σ 2= variance
DMAIC GB M TX PG V 4.2.0
D M A I C
99%
R&R
(5.15σ)
DMAIC GB M TX PG V 4.2.0
D M A I C
DMAIC GB M TX PG V 4.2.0
D M A I C
DMAIC GB M TX PG V 4.2.0
D M A I C
DMAIC GB M TX PG V 4.2.0
D M A I C
DMAIC GB M TX PG V 4.2.0
D M A I C
Use Minitab file: Gage R&R–Continuous Data.mtw Columns 1-3 in the Minitab worksheet represent
the data collected by the dispatchers as each
truck driver called on their delivery times.
DMAIC GB M TX PG V 4.2.0
D M A I C
DMAIC GB M TX PG V 4.2.0
D M A I C
DMAIC GB M TX PG V 4.2.0
D M A I C
Number
Numberof of R&R
R&R%%Study
Study
R&R
R&R Distinct
Distinct
R&R
R&R Variation
Variation
%Contribution
%Contribution Categories
Categories
%Tolerance
%Tolerance (Ratio
(Ratioof
ofStdDev’s
StdDev’s--
(Ratio
(Ratioof
ofVariances)
Variances) (Discrimination
(Discrimination --AIAG)
AIAG)
Index)
Index)
Red
30%
30% 8%
8% 44 30%
30%
Yellow
10%
10% 2%
2% 10
10 15%
15%
Green
CAUTION:
The magnitude of these
%’s are exaggerations
If the Gage R&R (as a percent of tolerance) is less than If the discrimination index (number of distinct
10%, the MS is green, or acceptable; if Gage R&R is categories) is greater than 10, the measurement system
10-30% , the MS is yellow, or marginal; if Gage R&R (MS) is green, or acceptable; if the discrimination
is greater than 30%, the MS is red, or unacceptable. index is 4-10, the MS is yellow, or marginal; if the
This estimate may be appropriate for evaluating how discrimination index is less than 4, the MS is red, or
well the measurement system can perform with respect unacceptable. The discrimination index is just a
to specifications. We can use it to look into the future transformation of %R&R contribution (for ease of
and ask “Will my measurement system variation be too interpretation.) It addresses the same issues and
large when I reach the 6 Sigma goal (with small contains the same information.
process spread, good capability)?” If Gage R&R (as a percent study) is less than 15%, the
If Gage R&R (as a percent contribution) is less than MS is green, or acceptable; if Gage R&R is 15-30%,
2%, the MS is green, or acceptable; if the Gage R&R is the MS is yellow, or marginal; if Gage R&R is greater
2-8%, the MS is yellow, or marginal; if Gage R&R is than 30%, the MS is red, or unacceptable.
greater than 8%, the MS is red, or unacceptable.
Note: the limits shown above can be derived from one
another. Some rounding was done for ease of
interpretation.
DMAIC GB M TX PG V 4.2.0
D M A I C
A. R&R% of Tolerance
1. R&R less than 10%–Measurement System “acceptable”
RULES OF THUMB:
DMAIC GB M TX PG V 4.2.0
D M A I C
DMAIC GB M TX PG V 4.2.0
D M A I C
D. R&R % Study
1. % Study less than 15%–measurement system acceptable.
2. % Study 15% to 30%–may be acceptable (marginal)
3. % Study over 30%–not acceptable
DMAIC GB M TX PG V 4.2.0
D M A I C
% Tolerance is 171.53% which fails, based on our % Study, probably the least important, (because %
criteria.This shows us the % of the Tolerance which is contribution tell us roughly the same thing) also fails
being taken up by the gage. In this case, 171.53% of based on our criteria. This gives us a measure of gage
our tolerance is taken up by the variation of the gage. variation compared to 99% of a normal distribution.
Note: You must put in the tolerance in Minitab in
order to get this data. If you do not enter a tolerance,
these figures will not appear. If you do not have a Number of distinct categories passes! This tells us the
tolerance, then % Contribution would be more number of divisions the measurement system can
important in your analysis accurately measure across the seen process variation.
.
% Contribution is 10.67% which fails our criteria. % For a more detailed explanation of output, refer to the
Contribution is based on overall measure of gage end of this section.
variation (regardless of the specification).
DMAIC GB M TX PG V 4.2.0
D M A I C
Gage name:
Date of study :
Gage R&R (ANOVA) for Meas Reported by :
Tolerance:
Misc:
90 3.0SL=87.96 90 2
Average
80 X=80.75 80
1 70 -3.0SL=73.54
70
3
4
60
60
50
40 50
30 40
0 Part/Truck 1 2 3 4 5 6 7 8 9 10
3.0SL=12.52 100
10 90
80
70
5
R=3.833 60
2 50 5
0 -3.0SL=0.00E+00 40
0 Oper 1 2 3
300 %Toler 80
3 200
70
60 6
100 50
0 40
Gage R&R Repeat Reprod Part-to-Part Part/Truck 1 2 3 4 5 6 7 8 9 10
Chart 4: Chart 5:
The chart shows the 10 parts and the average This chart shows the average reading for all parts by
measurement of those parts broken-out by operator. operator. The red circle is the average. This could show
Ideally, all 3 lines should be on top of each other. We can us if one operator is measuring higher/lower (on average)
see that there is some measurement variation in parts 4, 8 than the others.
and 10. We should investigate why.
DMAIC GB M TX PG V 4.2.0
D M A I C
DMAIC GB M TX PG V 4.2.0
D M A I C
If you fail the Gage R&R, here are some factors that could
cause measurement process variation (measurement error).
Repeatability
Operational definitions
Calibrating too often
Granularity of the measure (e.g., nearest hour, minute, second)
Maintain stability
Reproducibility
Operational definitions
Consistent use of gage
Training
Varying work environment
AUDIT–follow-up on training
Human/physical characteristics
Performance measures
Unclear requirements
DMAIC GB M TX PG V 4.2.0
D M A I C
Accuracy
Error in master
Instrument not calibrated
Worn components/trend or drift in master
Instrument used improperly by appraiser
Calibrating too often
Operational definition
Range of the master
DMAIC GB M TX PG V 4.2.0
D M A I C
Linearity
Instrument calibrated incorrectly
Error in master
Worn instrument
Stability
Error in master
Worn instrument
Instrument measuring wrong characteristic
Instrument not calibrated properly
Instrument used improperly by appraiser
DMAIC GB M TX PG V 4.2.0
D M A I C
Explanation of Bullets: The GB also noticed that all the data was positive,
The Dispatchers are using a clock on the wall that is (indicating that the deliveries were always late). Upon
not digital. Therefore, the GB implemented a new investigation, the GB found that the dispatchers did not
procedure to use the computer clock. All computers know that they should enter both positive and negative
were synchronized and a procedure was implemented numbers. A new spreadsheet was created to fix this
to have the system automatically synchronize clocks issue
each day with the Atomic Clock.
The GB also found that the Truck Drivers were
inconsistent in calling in their arrival times. After
going into the field and interviewing some truck
drivers, the GB found that a new call-in process needed
to be implemented. Part of the problem was not with
the gage but with “when” the truck driver called in the
delivery time. In many cases, the truck driver was
within the 10 minute window but had forgotten to call
in the arrival. A standardized process was
implemented to address this issue.
DMAIC GB M TX PG V 4.2.0
D M A I C
Step 5 (continued)
The % Contribution also failed (10.67%). From the Minitab data, the GB
found Reproducibility to be the larger issue and the interaction between
certain trucks and certain operators was the issue. To address these issues,
the GB has done the following:
– The data indicates that the measuring of the arrival time for trucks 4 & 10 seem to
vary from operator to operator. By going back and reviewing the data and talking
to the operators, the following was found:
There was a typo on the recording of time from Operator 1 on truck 4. He accidentally
typed in the wrong number. To address this problem, a new spreadsheet was
developed and now the Dispatcher is asked to verify his data prior to saving it in the
spreadsheet.
For Truck 10, the Operators had a very difficult time hearing the recording because the
Truck Driver was using a cell phone with a poor connection. The GB changed the
procedure so that each Dispatcher was now required to repeat-back the time given by
the truck driver. The Truck Driver then confirms that the correct time is being recorded.
– The data also indicated that Operator 2 seems to measure slightly lower than
Operators 1 & 3. The GB found that Operator 2 was sitting on an angle and when
he read the clock his time was distorted. By converting to the computer clock,
this issue should be resolved.
DMAIC GB M TX PG V 4.2.0
D M A I C
DMAIC GB M TX PG V 4.2.0
D M A I C
DMAIC GB M TX PG V 4.2.0
D M A I C
Rules of Thumb
1. R&R (% Tolerance)
Less than 10%–Measurement System is acceptable
10% to 30%–maybe acceptable–make decision based on classification of
characteristic, hardware application, customer input, etc.
Over 30%–Measurement System is not acceptable. Find problem, re-visit
the fishbone diagram, remove root causes. Validate Measurement System
again
2. % Contribution (or Gage R&R StdDev) :
GR&R Variance should be “small” compared to Part-to-Part
Variance–applies in cases where Tolerance Width is not
meaningful, and % Tolerance is unavailable–Such as one-sided
specs.
1.% Contribution < 2%-Measurement System “acceptable”
2.% Contribution 2%-8%-Measurement System “marginal”
3.% Contribution > 8%-Measurement System “unacceptable”
DMAIC GB M TX PG V 4.2.0
D M A I C
4. R&R % Study
1. % Study less than 15%–measurement system acceptable
2. % Study 15% to 30%–may be acceptable (marginal)
3. % Study over 30%–not acceptable
DMAIC GB M TX PG V 4.2.0
D M A I C
DMAIC GB M TX PG V 4.2.0
D M A I C
Desired Outcome: Outline of a Data Collection Plan and MSA for your
GB project
2 . MSA* 15 mins.
Develop an MSA for your project Y
data
List factors that might cause
measurement system variation and
how you would reduce the impact
of those factors
DMAIC GB M TX PG V 4.2.0
D M A I C
DMAIC GB M TX PG V 4.2.0
D M A I C
Graphical Output
•ANOVA
•ANOVAsimplysimplystands
standsfor
for““Analysis
AnalysisofofVariance”
Variance”
•It•Itisisaatool
tool used to analyzethe
used to analyze thetotal
totalvariability
variabilityamong
amongdifferent
differentsources.
sources.
•It•Itpartitions
partitions the total variation into “buckets”, and then allocateseach
the total variation into “buckets”, and then allocates eachbucket
buckettotoeach
eachsource
source
(part,
(part,operator,
operator,oper*part
oper*partinteraction).
interaction).
Gage name:
Date of study :
Gage R&R (ANOVA) for Meas Reported by :
Tolerance:
Misc:
90 3.0SL=87.96 90
1 2
4
Average
80 X=80.75 80
-3.0SL=73.54 3
70
70
60
60
50
40 50
30 40
0 Part/Truck 1 2 3 4 5 6 7 8 9 10
3.0SL=12.52 100
10 90
80
70
5
2 R=3.833 60
50
5
0 -3.0SL=0.00E+00 40
0 Oper 1 2 3
300 %Toler 80
70
3 200
60 6
100 50
0 40
Gage R&R Repeat Reprod Part-to-Part Part/Truck 1 2 3 4 5 6 7 8 9 10
What do the graphs tell you about the measurement 5 Scatter of individual measures should be equal by
system? Is it acceptable? operator.
1 > 50% of points should be outside control limits. 6 Scatter of individual measures should be equal by part
(truck.)
2 Range of measures by operator should be in control
(this is showing repeatability.)
DMAIC GB M TX PG V 4.2.0
D M A I C
Average of 2 measurements
taken by operator 1. Gage name:
Date of study :
Gage R&R (ANOVA) for Meas Reported by :
Tolerance:
Misc:
90 3.0SL=87.96 90 2
Average
80 X=80.75 80
-3.0SL=73.54 3
70
70
60
60
50
40 50
30 40
0 Oper 1 Oper 2 Oper 3 Part/Truck 1 2 3 4 5 6 7 8 9 10
3.0SL=12.52
10 90
80
70
5
R=3.833 60
50
0 -3.0SL=0.00E+00 40
0 Oper 1 2 3
Xbar chart by Operator: We want the R chart to be “in control”. If it is not, then
Plots the average for each part measured by each this may be an indication that repeatability is poor. Our
operator (average of 2-3 #’s). In our example, it is the data in our session window will confirm if we have a
average of 2 measurements. Repeatability issue. At this point, make a mental note.
Look for any points outside the Upper Control Limit
We want the parts to be “out of control” If they are not (verify the measurement is not a typo!)
out of control, we can’t tell one part from another!
Remember , we’ve chosen parts that are different from In this case, the R Chart looks OK. The chart is in
each other. Usually, we want 50% of our points to be control indicating that Repeatability is probably not an
outside the control limits. issue.
R Chart by Operator:
Plots the range of each average point in the Xbar chart.
This is the difference between the highest and lowest
measurement for all 10 parts for each operator.
© GE Capital, Inc., 2000
DMAIC GB M TX PG V 4.2.0
D M A I C
Gage name:
Date of study :
Gage R&R (ANOVA) for Meas Reported by :
Tolerance:
Misc:
90 3.0SL=87.96 90 2
Average
80 X=80.75 80
-3.0SL=73.54 3
70
70
60
60
50
40 50
30 40
0 Part/Truck 1 2 3 4 5 6 7 8 9 10
3.0SL=12.52
10 90
80
70
5
R=3.833 60
50
0 -3.0SL=0.00E+00 40
0 Oper 1 2 3
300 %Toler 80
70
200
60
100 50
0 40
Gage R&R Repeat Reprod Part-to-Part Part/Truck 1 2 3 4 5 6 7 8 9 10
A B C D
DMAIC GB M TX PG V 4.2.0
D M A I C
Gage name:
Date of study :
Gage R&R (ANOVA) for Meas Reported by :
Tolerance:
Misc:
90 3.0SL=87.96 90 2
Average
80 X=80.75 80
-3.0SL=73.54 3
70
70
60
60
50
40 50
30 40
0 Part/Truck 1 2 3 4 5 6 7 8 9 10
3.0SL=12.52
10 90
80
70
5
R=3.833 60
50
0 -3.0SL=0.00E+00 40
0 Oper 1 2 3
300 %Toler 80
70
200
60
100 50
0 40
Gage R&R Repeat Reprod Part-to-Part Part/Truck 1 2 3 4 5 6 7 8 9 10
Operator by Part Interaction: Shows interaction By Operator: Main effect plot for Operator
between the operator and the part
We want a flat horizontal line. This would indicate that it
We want the lines to be on top of each other. Crossed is unlikely the operator is having an effect...confirm with
lines could indicate the potential for interaction. Confirm the p-value from ANOVA. This chart shows the
suspected interactions with the p-value in ANOVA Table. measurements, by operator. Some dots may be on top of
We want all operators measuring the parts the very each other. Visually, you can see each operator’s
similarly (ie, getting the same measurement). Use this measurements.
chart to help you look at particular parts which operators
may be measuring differently. By reviewing this chart, By Part: Main effect plot for Part.
we notice that the operators are measuring Part 4 & Part
10 differently. We would want to investigate why. This
We want to see lots of changes in slope. This would
chart would also show us if one operator measures higher
indicate that it is likely that the parts are having an effect.
or lower than the others. It may indicate a
Confirm with the p-value from Anova. (Remember, we
Reproducibility issue. It appears that Operator 2 is
set up our study this way. We wanted differences in the
measuring slightly lower than Operators 1 & 3. The
parts.) More importantly, look for parts that have varying
Minitab Session Window can also tell us more.
measurements. For instance, part 10 shows more
variation in measurement than some of the other parts.
DMAIC GB M TX PG V 4.2.0
D M A I C
ANOVA TABLE
Two-Way ANOVA Table With Interaction
Source DF SS MS F P
For this gage... parts, operators and the interaction between parts & operators
are statistically significant sources of variation!
There are 4 major outputs other than the graphical Oper*Part/Truck: p=0.00016; This is the interaction
summary. We will discuss each separately. between the operator and the parts (eg, when operators
measure/record certain travel times). Since this is
Explanation of output for ANOVA Table: statistically significant, it is indicating that there may be
some issues when some operators measure/record certain
Part/Truck: p=0.00000; I would expect the parts to be runs. We will look further into the Minitab output to find
significant because we chose truck travel routes that out which run and which operators.
were really different from each other (short runs and long
runs).
DMAIC GB M TX PG V 4.2.0
D M A I C
% Comparisons Table
Source %Contribution %Study Var %Tolerance
P/T
P/TRatio
Ratio
(Precision-to-Tolerance Total Gage R&R 10.67 32.66 171.53
(Precision-to-ToleranceRatio)
Ratio) Repeatability 3.10 17.62 92.54
Reproducibility 7.56 27.50 144.43
Oper 2.19 14.81 77.76
Oper*Part/Truck 5.37 23.17 121.70
Part-To-Part 89.33 94.52 496.41
Total Variation 100.00 100.00 525.21
“R&R
“R&Rasasaa%
%ofof
Tolerance”
Tolerance”
% Tolerance
Acceptable Ranges
Not Acceptable
30%
Questionable
10%
Acceptable
This shows us the % of the Tolerance which is being If % Tolerance is unacceptable now, then once you
taken up by the gage. For instance, in this case, 171.53% reach a Six Sigma Process, the measurement variation
of our tolerance is taken up by the variation of the gage. will be too large compared to the process variation.
Refer to the next page for a visual explanation of % In this case, the gage fails % Tolerance.
Tolerance. Note: You must put in the tolerance in
Minitab in order to get this data. If you do not enter a
tolerance, these figures will not appear. If you do not
have a tolerance, then % Contribution would be more
important in your analysis.
DMAIC GB M TX PG V 4.2.0
D M A I C
%Tolerance: Graphically
LSL USL
%Tolerance = 10%
%Tolerance = 30%
%Tolerance = 70%
DMAIC GB M TX PG V 4.2.0
D M A I C
Source %Contribution
% Contribution = a good indicator for Present Status Note: If you don’t have a tolerance, then, %
Contribution becomes a very critical measurement for
The measurement system (gage) is contributing to
Gage R&R. It gives you a good indication of the
10.67% of the total variation and the parts are
measurement variation, now, compared to a realistic
contributing 89.33%.
estimate of current process spread.
This Gage Fails the Criteria for % Contribution, but it
is close to passing.
If you wanted to calculate %R&R Contribution by hand:
σ 2
% R&R R & R
=
( M e a s . S y s t .)
× 1 0 0 %
Contribution σ 2
T o t a l
DMAIC GB M TX PG V 4.2.0
D M A I C
Discrimination Index
Number of Distinct Categories = 4
Discrimination
DiscriminationIndex
Index
For this gage: the Discrimination Index = 4
• This says that the gage has reached the minimum acceptable
level.
Discrimination Index:
Discrimination Index:
•Provides the number of divisions that the Measurement System can accurately
•Provides the number of divisions that the Measurement System can accurately
measure
measureacross
acrossthe
theprocess
processvariation.
variation.
“Number
“Numberof
ofDistinct
Distinct •Indicates
•Indicateshow
how well
wellaagage
gagecan
candetect
detectpart-to-part
part-to-partvariation
variation---- process
processshifts
shiftsand
and
Categories”
Categories” improvement.
improvement.
Discrimination
DiscriminationIndex
IndexAcceptable
AcceptableValues
Values
•Less than 2, inadequate
•Equal to 2, equal to a go/nogo gauge.
•Minimum acceptable value = 4.
•Optimal = 10 or more.
If you wanted to calculate
Discrimination Index by hand:
σ2
Std Dev (Parts) D is c rim = 2 • 2 T o ta l −1
x 1.41 σ R &R
Std Dev (GRR) ( M e a s .S y s t . )
DMAIC GB M TX PG V 4.2.0
D M A I C
33Key
KeyIndices
Indicesfor
forMeasurement
MeasurementSystem
SystemCapability
Capability
11 22 33
P/T Discrimination
DiscriminationIndex
P/TRatio
Ratio %R&R
%R&RContribution
Contribution Index
(Precision-to-Tolerance
(Precision-to-ToleranceRatio)
Ratio)
“R&R
“R&Rasasaa%
%ofof “R&R
“R&Ras
asaa%%of
ofTotal
Total “Number
“Numberof
ofDistinct
Distinct
Tolerance”
Tolerance” Measure Variation”
Measure Variation” Categories”
Categories”
The next question to be answered: What do I do now We also know that Operators and the
that the % Tolerance & % Contribution has failed? Operator*Part/Truck Interaction is significant
(from p-value). This is an indication that the
Well, this is what we have learned so far about the break- operators are measuring differently. It also tells me
down of % Tolerance and % Contribution: there is some variation when certain operators
For % Tolerance and % Contribution, we know the measure certain trucks. Again, the graphs indicate
bigger issue is reproducibility. And, with which trucks and which operators.
Reproducibility, the bigger issue is the interaction
between parts and operators. By looking at the I would investigate those areas mentioned above.
graphical summary, it appears that trucks 4 and 10 Depending on your findings, you may need to implement
may be the issue. Investigate why there is a a new training program with your operators on how to
difference. Is there something unique about those measure/record the time. You also may need to do
truck routes? Is it harder to measure the time for something like having the computer measure the time
Trucks 4 & 10 due to time length? Should there be an when we measure parts similar to Parts 4 & 10. Once
Operational Definition or Procedure on how to you’ve done those items, go back and re-run the Gage
measure? Was there a problem with calculating the R&R using your implemented changes.
time from target?
DMAIC GB M TX PG V 4.2.0
D M A I C
Acceptable
Acceptable Ranges–ANOVA
Ranges–ANOVA METHOD
METHOD
This tells us the ratio of each
This tells us how much of the component of variation (the
This tells us how much
tolerance is being taken up by standard deviation for each
variation the gage itself is
the variation the gage. component divided by the total
contributing to the variation.
standard deviation).
Not Acceptable
8.0% 30% 4 30%
Questionable
2.0% 10% 10 15%
Acceptable
• Provides the number of divisions that the Measurement System can accurately
measure across the process variation.
• Indicates how well a gage can detect part-to-part variation--process shifts and
improvement.
DMAIC GB M TX PG V 4.2.0
D M A I C
DMAIC GB M TX PG V 4.2.0
D M A I C
DMAIC GB M TX PG V 4.2.0
D M A I C
1
Analyze Module Objectives
You will learn the concepts and tools for the Analyze
phase during this classroom session. After the classroom
session, you can use the Six Sigma training CD to
review how the Analyze phase is applied.
DMAIC GB N TX PG V 4.2.0
D M A I C
2
Analyze Phase Flowchart
D M A I C
ANALYZE
PHASE
OVERVIEW
DMAIC GB N TX PG V 4.2.0
D M A I C
3
The12-Step Process
DMAIC GB N TX PG V 4.2.0
D M A I C
4
Analyze Phase Overview
DMAIC GB N TX PG V 4.2.0
D M A I C
5
Statistical Thinking
D M A I C
Practical Statistical Statistical Practical
Problem Problem Solution Solution
Problem Characterize the Root cause Verify critical
statement process analysis X’s and ƒ(x)
– Project Y – Stability – Critical X’s Change
– Magnitude – Shape Measure the process
– Impact – Center influence of the Control the
critical X’s on gains
– Variation
Data Integrity the mean and – Risk
– MSA Capability variability analysis
– Brainstorm – ZBench ST & LT – Test – Control
potential X’s – Model plans
– Sampling plan – Estimate
Collect data
The Practical-To-Statistical-To-Practical
Transformation Process
DMAIC GB N TX PG V 4.2.0
D M A I C
DMAIC GB N TX PG V 4.2.0
D M A I C
ANALYZE STEP
OVERVIEW
DMAIC GB O TX PG V 4.2.0
D M A I C
DMAIC GB O TX PG V 4.2.0
D M A I C
DMAIC GB O TX PG V 4.2.0
D M A I C
DMAIC GB O TX PG V 4.2.0
D M A I C
DMAIC GB O TX PG V 4.2.0
D M A I C
DMAIC GB O TX PG V 4.2.0
D M A I C
Process Performance
Purchase Order 98% accuracy
Generation
Accounts Receivable 33 days average aging
Customer Service 82% rated 4 or 5 on responsiveness
Supplier Delivery 95% on-time delivery
DMAIC GB O TX PG V 4.2.0
D M A I C
e.g Light On
Discrete
DiscreteData
Data Continuous
Continuous
Data
Data
Defects per Six Sigma Process
Opportunity Report:
Defects per Million ZLong term
Opportunities
ZShort term = ZBench =
Six Sigma Product reported yield
Report
ZShift
DMAIC GB O TX PG V 4.2.0
D M A I C
DMAIC GB O TX PG V 4.2.0
D M A I C
X− µ
Z=
σ
µ 1σ 2σ 3σ 4σ
DMAIC GB O TX PG V 4.2.0
D M A I C
Calculating Capability 11
X = 8.5
s = 0.1
xx
Probability
Probability Probability
Probability
ofofaadefect
defect LSL
LSL USL
USL ofofaadefect
defect
less
lessthan
than greater
ZUSL greater
LSL
LSL ZLSL than
thanUSL
USL
Standard Deviations -4 -3 -2 -1 0 1 2 3 4
Units of Measure 8.1 8.2 8.3 8.4 8.5 8.6 8.7 8.8 8.9
ZST = 3.4
DMAIC GB O TX PG V 4.2.0
D M A I C
Yield
Probability
Of A Defective
Example = .028066
Specification
Limit
Z-score
Units of Measure
0.20 .420740315 1.71 .043632958 3.22 .000640954 4.73 .000001153
0.25 .401293634 1.76 .039203955 3.27 .000537758 4.78 .000000903
.382088486 1.81 .035147973 3.32 .000450127 4.83 .000000705
1.86 .031442864 3.37 .000375899 4.88 .000000550
1.91 .028066724 3.42 .000313179 4.93 .000000428
1.96 .024998022 3.47 .000260317 4.98 .000000332
2.01 .022215724 3.52 .000215873 5.03 .000000258
2.06
2.11 .019699396 3.57 .000178601 5.08 .000000199
2.16 .017429293 3.62 .000147419 5.13 .000000154
To the
2.21 .015386434 3.67 .000121399 5.18 .000000118
2.26 .013552660 3.72 .000099739 5.23 .000000091
2.31
2.36 .011910681 3.77 .000081753 5.28 .000000070
right of the
.010444106 3.82 .000066855 5.33 .000000053
.009137469 3.87 .000054545 5.38 .000000041
2.41 .007976235 3.92 .000044399 5.43 .000000031
2.46 .006946800 3.97 .000036057 5.48 .000000024
Z-score Z = 1.91
2.51
2.56
2.61
.006036485
.005233515
.004527002
4.02
4.07
4.12
.000029215
.000023617
.000019047
5.53
5.58
5.63
.000000018
.000000014
.000000010
DMAIC GB O TX PG V 4.2.0
D M A I C z
(Values of Z from 0.00 to 4.99)
Single–Tail Z Table 13
Z 0.00 0.01 0.02 0.03 0.04 0.05 0.06 0.07 0.08 0.09
0.00 5.00e-001 4.96e-001 4.92e-001 4.88e-001 4.84e-001 4.80e-001 4.76e-001 4.72e-001 4.68e-001 4.64e-001
0.10 4.60e-001 4.56e-001 4.52e-001 4.48e-001 4.44e-001 4.40e-001 4.36e-001 4.33e-001 4.29e-001 4.25e-001
0.20 4.21e-001 4.17e-001 4.13e-001 4.09e-001 4.05e-001 4.01e-001 3.97e-001 3.94e-001 3.90e-001 3.86e-001
0.30 3.82e-001 3.78e-001 3.74e-001 3.71e-001 3.67e-001 3.63e-001 3.59e-001 3.56e-001 3.52e-001 3.48e-001
0.40 3.45e-001 3.41e-001 3.37e-001 3.34e-001 3.30e-001 3.26e-001 3.23e-001 3.19e-001 3.16e-001 3.12e-001
0.50 3.09e-001 3.05e-001 3.02e-001 2.98e-001 2.95e-001 2.91e-001 2.88e-001 2.84e-001 2.81e-001 2.78e-001
0.60 2.74e-001 2.71e-001 2.68e-001 2.64e-001 2.61e-001 2.58e-001 2.55e-001 2.51e-001 2.48e-001 2.45e-001
0.70 2.42e-001 2.39e-001 2.36e-001 2.33e-001 2.30e-001 2.27e-001 2.24e-001 2.21e-001 2.18e-001 2.15e-001
0.80 2.12e-001 2.09e-001 2.06e-001 2.03e-001 2.00e-001 1.98e-001 1.95e-001 1.92e-001 1.89e-001 1.87e-001
0.90 1.84e-001 1.81e-001 1.79e-001 1.76e-001 1.74e-001 1.71e-001 1.69e-001 1.66e-001 1.64e-001 1.61e-001
1.00 1.59e-001 1.56e-001 1.54e-001 1.52e-001 1.49e-001 1.47e-001 1.45e-001 1.42e-001 1.40e-001 1.38e-001
1.10 1.36e-001 1.33e-001 1.31e-001 1.29e-001 1.27e-001 1.25e-001 1.23e-001 1.21e-001 1.19e-001 1.17e-001
1.20 1.15e-001 1.13e-001 1.11e-001 1.09e-001 1.07e-001 1.06e-001 1.04e-001 1.02e-001 1.00e-001 9.85e-002
1.30 9.68e-002 9.51e-002 9.34e-002 9.18e-002 9.01e-002 8.85e-002 8.69e-002 8.53e-002 8.38e-002 8.23e-002
1.40 8.08e-002 7.93e-002 7.78e-002 7.64e-002 7.49e-002 7.35e-002 7.21e-002 7.08e-002 6.94e-002 6.81e-002
1.50 6.68e-002 6.55e-002 6.43e-002 6.30e-002 6.18e-002 6.06e-002 5.94e-002 5.82e-002 5.71e-002 5.59e-002
1.60 5.48e-002 5.37e-002 5.26e-002 5.16e-002 5.05e-002 4.95e-002 4.85e-002 4.75e-002 4.65e-002 4.55e-002
1.70 4.46e-002 4.36e-002 4.27e-002 4.18e-002 4.09e-002 4.01e-002 3.92e-002 3.84e-002 3.75e-002 3.67e-002
1.80 3.59e-002 3.51e-002 3.44e-002 3.36e-002 3.29e-002 3.22e-002 3.14e-002 3.07e-002 3.01e-002 2.94e-002
1.90 2.87e-002 2.81e-002 2.74e-002 2.68e-002 2.62e-002 2.56e-002 2.50e-002 2.44e-002 2.39e-002 2.33e-002
2.00 2.28e-002 2.22e-002 2.17e-002 2.12e-002 2.07e-002 2.02e-002 1.97e-002 1.92e-002 1.88e-002 1.83e-002
2.10 1.79e-002 1.74e-002 1.70e-002 1.66e-002 1.62e-002 1.58e-002 1.54e-002 1.50e-002 1.46e-002 1.43e-002
2.20 1.39e-002 1.36e-002 1.32e-002 1.29e-002 1.25e-002 1.22e-002 1.19e-002 1.16e-002 1.13e-002 1.10e-002
2.30 1.07e-002 1.04e-002 1.02e-002 9.90e-003 9.64e-003 9.39e-003 9.14e-003 8.89e-003 8.66e-003 8.42e-003
2.40 8.20e-003 7.98e-003 7.76e-003 7.55e-003 7.34e-003 7.14e-003 6.95e-003 6.76e-003 6.57e-003 6.39e-003
2.50 6.21e-003 6.04e-003 5.87e-003 5.70e-003 5.54e-003 5.39e-003 5.23e-003 5.08e-003 4.94e-003 4.80e-003
2.60 4.66e-003 4.53e-003 4.40e-003 4.27e-003 4.15e-003 4.02e-003 3.91e-003 3.79e-003 3.68e-003 3.57e-003
2.70 3.47e-003 3.36e-003 3.26e-003 3.17e-003 3.07e-003 2.98e-003 2.89e-003 2.80e-003 2.72e-003 2.64e-003
2.80 2.56e-003 2.48e-003 2.40e-003 2.33e-003 2.26e-003 2.19e-003 2.12e-003 2.05e-003 1.99e-003 1.93e-003
2.90 1.87e-003 1.81e-003 1.75e-003 1.69e-003 1.64e-003 1.59e-003 1.54e-003 1.49e-003 1.44e-003 1.39e-003
3.00 1.35e-003 1.31e-003 1.26e-003 1.22e-003 1.18e-003 1.14e-003 1.11e-003 1.07e-003 1.04e-003 1.00e-003
3.10 9.68e-004 9.35e-004 9.04e-004 8.74e-004 8.45e-004 8.16e-004 7.89e-004 7.62e-004 7.36e-004 7.11e-004
3.20 6.87e-004 6.64e-004 6.41e-004 6.19e-004 5.98e-004 5.77e-004 5.57e-004 5.38e-004 5.19e-004 5.01e-004
3.30 4.83e-004 4.66e-004 4.50e-004 4.34e-004 4.19e-004 4.04e-004 3.90e-004 3.76e-004 3.62e-004 3.49e-004
3.40 3.37e-004 3.25e-004 3.13e-004 3.02e-004 2.91e-004 2.80e-004 2.70e-004 2.60e-004 2.51e-004 2.42e-004
3.50 2.33e-004 2.24e-004 2.16e-004 2.08e-004 2.00e-004 1.93e-004 1.85e-004 1.78e-004 1.72e-004 1.65e-004
3.60 1.59e-004 1.53e-004 1.47e-004 1.42e-004 1.36e-004 1.31e-004 1.26e-004 1.21e-004 1.17e-004 1.12e-004
3.70 1.08e-004 1.04e-004 9.96e-005 9.57e-005 9.20e-005 8.84e-005 8.50e-005 8.16e-005 7.84e-005 7.53e-005
3.80 7.23e-005 6.95e-005 6.67e-005 6.41e-005 6.15e-005 5.91e-005 5.67e-005 5.44e-005 5.22e-005 5.01e-005
3.90 4.81e-005 4.61e-005 4.43e-005 4.25e-005 4.07e-005 3.91e-005 3.75e-005 3.59e-005 3.45e-005 3.30e-005
4.00 3.17e-005 3.04e-005 2.91e-005 2.79e-005 2.67e-005 2.56e-005 2.45e-005 2.35e-005 2.25e-005 2.16e-005
4.10 2.07e-005 1.98e-005 1.89e-005 1.81e-005 1.74e-005 1.66e-005 1.59e-005 1.52e-005 1.46e-005 1.39e-005
4.20 1.33e-005 1.28e-005 1.22e-005 1.17e-005 1.12e-005 1.07e-005 1.02e-005 9.77e-006 9.34e-006 8.93e-006
4.30 8.54e-006 8.16e-006 7.80e-006 7.46e-006 7.12e-006 6.81e-006 6.50e-006 6.21e-006 5.93e-006 5.67e-006
4.40 5.41e-006 5.17e-006 4.94e-006 4.71e-006 4.50e-006 4.29e-006 4.10e-006 3.91e-006 3.73e-006 3.56e-006
4.50 3.40e-006 3.24e-006 3.09e-006 2.95e-006 2.81e-006 2.68e-006 2.56e-006 2.44e-006 2.32e-006 2.22e-006
4.60 2.11e-006 2.01e-006 1.92e-006 1.83e-006 1.74e-006 1.66e-006 1.58e-006 1.51e-006 1.43e-006 1.37e-006
4.70 1.30e-006 1.24e-006 1.18e-006 1.12e-006 1.07e-006 1.02e-006 9.68e-007 9.21e-007 8.76e-007 8.34e-007
4.80 7.93e-007 7.55e-007 7.18e-007 6.83e-007 6.49e-007 6.17e-007 5.87e-007 5.58e-007 5.30e-007 5.04e-007
4.90 4.79e-007 4.55e-007 4.33e-007 4.11e-007 3.91e-007 3.71e-007 3.52e-007 3.35e-007 3.18e-007 3.02e-007
DMAIC GB O TX PG V 4.2.0
D M A I C z
(Values of Z from 5.00 to 9.99)
Single–Tail Z Table 14
Z 0.00 0.01 0.02 0.03 0.04 0.05 0.06 0.07 0.08 0.09
5.00 2.87e-007 2.72e-007 2.58e-007 2.45e-007 2.33e-007 2.21e-007 2.10e-007 1.99e-007 1.89e-007 1.79e-007
5.10 1.70e-007 1.61e-007 1.53e-007 1.45e-007 1.37e-007 1.30e-007 1.23e-007 1.17e-007 1.11e-007 1.05e-007
5.20 9.96e-008 9.44e-008 8.95e-008 8.48e-008 8.03e-008 7.60e-008 7.20e-008 6.82e-008 6.46e-008 6.12e-008
5.30 5.79e-008 5.48e-008 5.19e-008 4.91e-008 4.65e-008 4.40e-008 4.16e-008 3.94e-008 3.72e-008 3.52e-008
5.40 3.33e-008 3.15e-008 2.98e-008 2.82e-008 2.66e-008 2.52e-008 2.38e-008 2.25e-008 2.13e-008 2.01e-008
5.50 1.90e-008 1.79e-008 1.69e-008 1.60e-008 1.51e-008 1.43e-008 1.35e-008 1.27e-008 1.20e-008 1.14e-008
5.60 1.07e-008 1.01e-008 9.55e-009 9.01e-009 8.50e-009 8.02e-009 7.57e-009 7.14e-009 6.73e-009 6.35e-009
5.70 5.99e-009 5.65e-009 5.33e-009 5.02e-009 4.73e-009 4.46e-009 4.21e-009 3.96e-009 3.74e-009 3.52e-009
5.80 3.32e-009 3.12e-009 2.94e-009 2.77e-009 2.61e-009 2.46e-009 2.31e-009 2.18e-009 2.05e-009 1.93e-009
5.90 1.82e-009 1.71e-009 1.61e-009 1.51e-009 1.43e-009 1.34e-009 1.26e-009 1.19e-009 1.12e-009 1.05e-009
6.00 9.87e-010 9.28e-010 8.72e-010 8.20e-010 7.71e-010 7.24e-010 6.81e-010 6.40e-010 6.01e-010 5.65e-010
6.10 5.30e-010 4.98e-010 4.68e-010 4.39e-010 4.13e-010 3.87e-010 3.64e-010 3.41e-010 3.21e-010 3.01e-010
6.20 2.82e-010 2.65e-010 2.49e-010 2.33e-010 2.19e-010 2.05e-010 1.92e-010 1.81e-010 1.69e-010 1.59e-010
6.30 1.49e-010 1.40e-010 1.31e-010 1.23e-010 1.15e-010 1.08e-010 1.01e-010 9.45e-011 8.85e-011 8.29e-011
6.40 7.77e-011 7.28e-011 6.81e-011 6.38e-011 5.97e-011 5.59e-011 5.24e-011 4.90e-011 4.59e-011 4.29e-011
6.50 4.02e-011 3.76e-011 3.52e-011 3.29e-011 3.08e-011 2.88e-011 2.69e-011 2.52e-011 2.35e-011 2.20e-011
6.60 2.06e-011 1.92e-011 1.80e-011 1.68e-011 1.57e-011 1.47e-011 1.37e-011 1.28e-011 1.19e-011 1.12e-011
6.70 1.04e-011 9.73e-012 9.09e-012 8.48e-012 7.92e-012 7.39e-012 6.90e-012 6.44e-012 6.01e-012 5.61e-012
6.80 5.23e-012 4.88e-012 4.55e-012 4.25e-012 3.96e-012 3.69e-012 3.44e-012 3.21e-012 2.99e-012 2.79e-012
6.90 2.60e-012 2.42e-012 2.26e-012 2.10e-012 1.96e-012 1.83e-012 1.70e-012 1.58e-012 1.48e-012 1.37e-012
7.00 1.28e-012 1.19e-012 1.11e-012 1.03e-012 9.61e-013 8.95e-013 8.33e-013 7.75e-013 7.21e-013 6.71e-013
7.10 6.24e-013 5.80e-013 5.40e-013 5.02e-013 4.67e-013 4.34e-013 4.03e-013 3.75e-013 3.49e-013 3.24e-013
7.20 3.01e-013 2.80e-013 2.60e-013 2.41e-013 2.24e-013 2.08e-013 1.94e-013 1.80e-013 1.67e-013 1.55e-013
7.30 1.44e-013 1.34e-013 1.24e-013 1.15e-013 1.07e-013 9.91e-014 9.20e-014 8.53e-014 7.91e-014 7.34e-014
7.40 6.81e-014 6.31e-014 5.86e-014 5.43e-014 5.03e-014 4.67e-014 4.33e-014 4.01e-014 3.72e-014 3.44e-014
7.50 3.19e-014 2.96e-014 2.74e-014 2.54e-014 2.35e-014 2.18e-014 2.02e-014 1.87e-014 1.73e-014 1.60e-014
7.60 1.48e-014 1.37e-014 1.27e-014 1.17e-014 1.09e-014 1.00e-014 9.30e-015 8.60e-015 7.95e-015 7.36e-015
7.70 6.80e-015 6.29e-015 5.82e-015 5.38e-015 4.97e-015 4.59e-015 4.25e-015 3.92e-015 3.63e-015 3.35e-015
7.80 3.10e-015 2.86e-015 2.64e-015 2.44e-015 2.25e-015 2.08e-015 1.92e-015 1.77e-015 1.64e-015 1.51e-015
7.90 1.39e-015 1.29e-015 1.19e-015 1.10e-015 1.01e-015 9.33e-016 8.60e-016 7.93e-016 7.32e-016 6.75e-016
8.00 6.22e-016 5.74e-016 5.29e-016 4.87e-016 4.49e-016 4.14e-016 3.81e-016 3.51e-016 3.24e-016 2.98e-016
8.10 2.75e-016 2.53e-016 2.33e-016 2.15e-016 1.98e-016 1.82e-016 1.68e-016 1.54e-016 1.42e-016 1.31e-016
8.20 1.20e-016 1.11e-016 1.02e-016 9.36e-017 8.61e-017 7.92e-017 7.28e-017 6.70e-017 6.16e-017 5.66e-017
8.30 5.21e-017 4.79e-017 4.40e-017 4.04e-017 3.71e-017 3.41e-017 3.14e-017 2.88e-017 2.65e-017 2.43e-017
8.40 2.23e-017 2.05e-017 1.88e-017 1.73e-017 1.59e-017 1.46e-017 1.34e-017 1.23e-017 1.13e-017 1.03e-017
8.50 9.48e-018 8.70e-018 7.98e-018 7.32e-018 6.71e-018 6.15e-018 5.64e-018 5.17e-018 4.74e-018 4.35e-018
8.60 3.99e-018 3.65e-018 3.35e-018 3.07e-018 2.81e-018 2.57e-018 2.36e-018 2.16e-018 1.98e-018 1.81e-018
8.70 1.66e-018 1.52e-018 1.39e-018 1.27e-018 1.17e-018 1.07e-018 9.76e-019 8.93e-019 8.17e-019 7.48e-019
8.80 6.84e-019 6.26e-019 5.72e-019 5.23e-019 4.79e-019 4.38e-019 4.00e-019 3.66e-019 3.34e-019 3.06e-019
8.90 2.79e-019 2.55e-019 2.33e-019 2.13e-019 1.95e-019 1.78e-019 1.62e-019 1.48e-019 1.35e-019 1.24e-019
9.00 1.13e-019 1.03e-019 9.40e-020 8.58e-020 7.83e-020 7.15e-020 6.52e-020 5.95e-020 5.43e-020 4.95e-020
9.10 4.52e-020 4.12e-020 3.76e-020 3.42e-020 3.12e-020 2.85e-020 2.59e-020 2.37e-020 2.16e-020 1.96e-020
9.20 1.79e-020 1.63e-020 1.49e-020 1.35e-020 1.23e-020 1.12e-020 1.02e-020 9.31e-021 8.47e-021 7.71e-021
9.30 7.02e-021 6.39e-021 5.82e-021 5.29e-021 4.82e-021 4.38e-021 3.99e-021 3.63e-021 3.30e-021 3.00e-021
9.40 2.73e-021 2.48e-021 2.26e-021 2.05e-021 1.86e-021 1.69e-021 1.54e-021 1.40e-021 1.27e-021 1.16e-021
9.50 1.05e-021 9.53e-022 8.66e-022 7.86e-022 7.14e-022 6.48e-022 5.89e-022 5.35e-022 4.85e-022 4.40e-022
9.60 4.00e-022 3.63e-022 3.29e-022 2.99e-022 2.71e-022 2.46e-022 2.23e-022 2.02e-022 1.83e-022 1.66e-022
9.70 1.51e-022 1.37e-022 1.24e-022 1.12e-022 1.02e-022 9.22e-023 8.36e-023 7.57e-023 6.86e-023 6.21e-023
9.80 5.63e-023 5.10e-023 4.62e-023 4.18e-023 3.79e-023 3.43e-023 3.10e-023 2.81e-023 2.54e-023 2.30e-023
9.90 2.08e-023 1.88e-023 1.70e-023 1.54e-023 1.39e-023 1.26e-023 1.14e-023 1.03e-023 9.32e-024 8.43e-024
DMAIC GB O TX PG V 4.2.0
D M A I C
ST Process Defects Per Defects Per Defects Per Defects Per Defects Per
Long-Term Yield
Sigma 1,000,000 100,000 10,000 1,000 100
99.99966% 6.0 3.4 0.34 0.034 0.0034 0.00034
99.9995% 5.9 5 0.5 0.05 0.005 0.0005
99.9992% 5.8 8 0.8 0.08 0.008 0.0008
99.9990% 5.7 10 1 0.1 0.01 0.001
99.9980% 5.6 20 2 0.2 0.02 0.002
99.9970% 5.5 30 3 0.3 0.03 0.003
99.9960% 5.4 40 4 0.4 0.04 0.004
99.9930% 5.3 70 7 0.7 0.07 0.007
99.9900% 5.2 100 10 1.0 0.1 0.01
99.9850% 5.1 150 15 1.5 0.15 0.015
99.9770% 5.0 230 23 2.3 0.23 0.023
99.9670% 4.9 330 33 3.3 0.33 0.033
99.9520% 4.8 480 48 4.8 0.48 0.048
99.9320% 4.7 680 68 6.8 0.68 0.068
99.9040% 4.6 960 96 9.6 0.96 0.096
99.8650% 4.5 1,350 135 13.5 1.35 0.135
99.8140% 4.4 1,860 186 18.6 1.86 0.186
99.7450% 4.3 2,550 255 25.5 2.55 0.255
99.6540% 4.2 3,460 346 34.6 3.46 0.346
99.5340% 4.1 4,660 466 46.6 4.66 0.466
99.3790% 4.0 6,210 621 62.1 6.21 0.621
99.1810% 3.9 8,190 819 81.9 8.19 0.819
98.930% 3.8 10,700 1,070 107 10.7 1.07
98.610% 3.7 13,900 1,390 139 13.9 1.39
98.220% 3.6 17,800 1,780 178 17.8 1.78
97.730% 3.5 22,700 2,270 227 22.7 2.27
97.130% 3.4 28,700 2,870 287 28.7 2.87
96.410% 3.3 35,900 3,590 359 35.9 3.59
95.540% 3.2 44,600 4,460 446 44.6 4.46
94.520% 3.1 54,800 5,480 548 54.8 5.48
93.320% 3.0 66,800 6,680 668 66.8 6.68
91.920% 2.9 80,800 8,080 808 80.8 8.08
90.320% 2.8 96,800 9,680 968 96.8 9.68
88.50% 2.7 115,000 11,500 1,150 115 11.5
86.50% 2.6 135,000 13,500 1,350 135 13.5
84.20% 2.5 158,000 15,800 1,580 158 15.8
81.60% 2.4 184,000 18,400 1,840 184 18.4
78.80% 2.3 212,000 21,200 2,120 212 21.2
75.80% 2.2 242,000 24,200 2,420 242 24.2
72.60% 2.1 274,000 27,400 2,740 274 27.4
69.20% 2.0 308,000 30,800 3,080 308 30.8
65.60% 1.9 344,000 34,400 3,440 344 34.4
61.80% 1.8 382,000 38,200 3,820 382 38.2
58.00% 1.7 420,000 42,000 4,200 420 42
54.00% 1.6 460,000 46,000 4,600 460 46
50% 1.5 500,000 50,000 5,000 500 50
46% 1.4 540,000 54,000 5,400 540 54
43% 1.3 570,000 57,000 5,700 570 57
39% 1.2 610,000 61,000 6,100 610 61
35% 1.1 650,000 65,000 6,500 650 65
31% 1.0 690,000 69,000 6,900 690 69
28% 0.9 720,000 72,000 7,200 720 72
25% 0.8 750,000 75,000 7,500 750 75
22% 0.7 780,000 78,000 7,800 780 78
19% 0.6 810,000 81,000 8,100 810 81
16% 0.5 840,000 84,000 8,400 840 84
14% 0.4 860,000 86,000 8,600 860 86
12% 0.3 880,000 88,000 8,800 880 88
10% 0.2 900,000 90,000 9,000 900 90
8% 0.1 920,000 92,000 9,200 920 92
Note: Subtract
1.5 to get long-
term
Sigma level
DMAIC GB O TX PG V 4.2.0
D M A I C
Long-Term Data
Y
(Continuous)
Short-Term
Data
Time
How much variation we observe in a process is For now, and as a default, treat all data as long-term data.
influenced by whether we are looking at long-term data The convention for Six Sigma is to report short-term, as
or short-term data. you will see later.
Data collected over a long enough period of time and We want to standardize reporting of the best a process
over diverse enough conditions such that it is likely to can do, given its current capability. Therefore,
contain some process shifts and other special causes. Short-Term Sigma tells us this.
Short-Term:
DMAIC GB O TX PG V 4.2.0
D M A I C
It is possible to calculate short-term, but for most Green Special Cause = Between subgroup variation due to
Belt projects we do not attempt to do so; instead we use assignable causes, non-random influences.
the 1.5 sigma shift rule.
Common Cause = Within subgroup variation,
In order to truly calculate short-term sigma, you have to inherent in a process, random influences.
use several rational subgroups. If you did not use a
rational sub-grouping sampling scheme for data To ensure you are sampling* properly, these 3 items must
collection in the Measure phase, it will not be possible to be true:
calculate actual short-term sigma. (Therefore, add 1.5 to – Representative Data
long-term sigma.)
– Enough Data (at least 20 subgroups)
– Measured with something that passes the Gage R&R
DMAIC GB O TX PG V 4.2.0
D M A I C
DMAIC GB O TX PG V 4.2.0
D M A I C
3.4 ppm
TT
LSL USL
4.5σ
DMAIC GB O TX PG V 4.2.0
D M A I C
SL =USL
LSL λ= T (Target)
µ (Mean)
Z =
st (short-term)
lt (long-term) σst
σlt
SL − T
ZST =
σ ST
DMAIC GB O TX PG V 4.2.0
D M A I C
Z-Bench 21
Short-Term
Long-Term
P(d)LSL P(d)USL
_
LSL T x USL
Z-Long-Term Z-Short-Term
SL - µ SL - T
Zlt = Zst =
σ lt σst
Z-Bench-Long-Term Z-Bench-Short-Term
USL- µ µ - LSL USL- T T - LSL
ZUSL= ZLSL= ZUSL= ZLSL=
σ lt σ lt σ st σ st
DMAIC GB O TX PG V 4.2.0
D M A I C
Mean = 5
Standard Deviation = 2
Upper Spec. Limit = 9
You can use the hand calculation or you can open the
spreadsheet: SIGMACAL.xls.
Enter:
Xbar: 5
S: 2
USL: 9
Note: This gives you ZST (the 1.5 shift was added to ZLT).
DMAIC GB O TX PG V 4.2.0
D M A I C
DMAIC GB O TX PG V 4.2.0
D M A I C
Solution–Run Chart 24
DMAIC GB O TX PG V 4.2.0
D M A I C
Solution–Run Chart 25
1. Double click on
“C5”
3. Click ”OK”
DMAIC GB O TX PG V 4.2.0
D M A I C
Solution–Run Chart 26
100
time
0 Median
2 7 12 17
Subgroup Number
Number of runs about median: 10.0000 Number of runs up or dow n: 9.0000
Expected number of runs: 9.4706 Expected number of runs: 11.0000
Longest run about median: 4.0000 Longest run up or dow n: 4.0000
Approx P-Value for Clustering: 0.6050 Approx P-Value for Trends: 0.1118
Spread
Approx P-Value for Mixtures: 0.3950 Approx P-Value for Oscillation: 0.8882
This Run Chart indicates that the process is stable over time.
DMAIC GB O TX PG V 4.2.0
D M A I C
Solution–Normality Testing 27
.999
.99
.95
Probability
.80
.50
.20
.05
.01
.001
-100 0 100
time
Average: -4.40316 Anderson-Darling Normality Test
StDev: 43.1714 A-Squared: 0.130
N: 506 P-Value: 0.983
DMAIC GB O TX PG V 4.2.0
D M A I C
Descriptive Statistics
Variable: time This
data is
normal
Anderson-Darling Normality Test
.
A-Squared: 0.130
P-Value: 0.983
Mean -4.4032
StDev 43.1714
Variance 1863.77
Skewness 7.65E-02
Kurtosis -1.2E-01
-100 -60 -20 20 60 100 N 506
Minimum -114.021
1st Quartile -35.085
Median -4.956
3rd Quartile 23.421
95% Confidence Interval for Mu Maximum 120.046
95% Confidence Interval for Mu
-8.174 -0.633
-10 -5 0 95% Confidence Interval for Sigma
40.665 46.009
95% Confidence Interval for Median
95% Confidence Interval for Median
-9.369 0.532
Variable = Time
Click on “Graphs” & choose “Graphical Summary”
DMAIC GB O TX PG V 4.2.0
D M A I C
1. Double
Click C5 time
2. Double
click C1 to
use Date1
as the
Subgroup
size
3. Type in
Lower and
Upper
specs
4. Click OK
Now let’s look at how to calculate process capability for The Six Sigma Process Report is used to calculate the
continuous data in Minitab, using the Six Sigma Process long-term and short-term z-values of your process. In
Report. order to be accurate in the calculation of the z-values, it is
important that we enter all of the information available.
The subgroup size can be entered in two different ways: If a target is available, then enter the target value for the
Size of each subgroup –10 (subgroup size must be calculation of the z-values. If the upper and lower
constant) specification limits are entered and the target is left
blank, Minitab approximates the target value as the
ID column–oper 50 (subgroup size may be constant or midpoint of the specific range. If only one specification
not) limit is entered and the target is left blank, then Minitab
approximates the target as the mean of the data. Hence,
Warning: If you choose to use a subgroup size of 1, the when the target is available, always input the given value
short-term values calculated in the Six Sigma Process to avoid approximations made by Minitab.
Report will be invalid since there is not variation within
the groups.
DMAIC GB O TX PG V 4.2.0
D M A I C
10,000
Process Benchmarks
1000
Actual (LT) Potential (ST)
100
Sigma 0.97 0.98
(Z.Bench)
10
PPM 166781 164350
1
0 5 10 15
Report 1 contains the essence: Upper graph: show short (dotted) and long-term (full
Sigma potential (ST) = ZBench-ST: if you were to use line) distributions. By definition, the short-term
just one number to indicate the capability of your distribution is shown to be on target. In this case, the
process, it is this one. If a firm XY claims to be a 5σ long-term distribution is slightly below the target.
company, they talk about this value. If you only see Lower graph: shows cumulative PPM’s.
ZST, it means ZBench-ST. It is a convention to benchmark
other companies using this value. However, it is a
Note: When reporting ZST, determine if you will be using
good practice to report out both ZBench values, for the
the shift calculated from your data or the standard 1.5
short and long-term.
shift. If using the 1.5 shift, you will need to manually add
Sigma actual (LT) = ZBench-LT: this value corresponds the 1.5 to ZLT to get ZST. Minitab cannot accommodate a
to the actual performance of your process, and is easier predetermined shift. (When in doubt, use the 1.5 shift
to comprehend with PPM. and/or consult your BB/MBB).
PPM actual = DPMO actual: the actual number of
defects per million opportunities that the process
produces.
PPM potential (not so important): corresponds to
ZBench-ST, not commonly used.
DMAIC GB O TX PG V 4.2.0
D M A I C
Report 2 is more detailed: Note: When reporting ZST, determine if you will be using
The table to the right side contains all the details of the the shift calculated from your data or the standard 1.5
shift. If using the 1.5 shift, you will need to manually add
calculation of ZBench for short and long-term.
the 1.5 to ZLT to get ZST. Minitab cannot accommodate a
The graphs to the left side are called control charts. predetermined shift. (When in doubt, use the 1.5 shift
This subject will be covered in the “control” session of and/or consult your BB/MBB).
your training. In order to use Minitab’s calculated shift
(right column in the table), the “S” chart must be in
control (minimize variation within subgroup)
In this example, S chart is out-of-control (one point
outside control limits). It does not meet the condition
mentioned in the 2nd bullet, therefore, report ZLT =
0.9670. To get ZST: ZST = ZLT + ZShift
ZST = 0.9670 + 1.5 = 2.467
Refer to the next 2 pages for rules on using this
Process Capability Report.
DMAIC GB O TX PG V 4.2.0
D M A I C
DMAIC GB O TX PG V 4.2.0
D M A I C
Is your
Process Yes Do you Yes Is the Yes Use Six Sigma> Process Report
Stable (use have Distribution (Input Data, Specification Limits
Run Chart)? Continuous Normal? and Target as appropriate)
Data?
No No No
Do you
Analyze for Special Six Sigma>Product Report, No Use Z Bench-LT only
have
Causes- DPMO, L1 (Input D, U and O) and add 1.5 Shift
Rational
to get Z Bench-ST
Should not •Sigma Conversion Table gives ZST Subgroups?
calculate process
•Z LT = Z ST - Z Shift (1.5)
capability until the
process is stable. •Product Report and L1 give Z Bench Yes
Unstable process •Z Bench = Z ST
data can only be
•Z LT = Z Bench - Z Shift (1.5) Is the S- No Use Z Bench-LT only
used to describe and add 1.5 Shift
Chart in- to get Z Bench-ST
the capability of the
control?
current data set, not
the future capability
of the process. Yes
DMAIC GB O TX PG V 4.2.0
D M A I C
High
Probability High
of Defective Probability of
Defective
Good Design
Capability
LSL USL
Low Low
Probability Probability
of Defective of Defective
LSL USL
DMAIC GB O TX PG V 4.2.0
D M A I C
Summary–Z-Value 35
Note on Capability:
Capability can be described as any of the following:
– ZBenchST = ZST = σST = Sigma CapabilityST =
SigmaST = ZBench
– ZBenchLT = ZLT = σLT = Sigma CapabilityLT =
SigmaLT
DMAIC GB O TX PG V 4.2.0
D M A I C
DMAIC GB O TX PG V 4.2.0
D M A I C
DMAIC GB O TX PG V 4.2.0
D M A I C
Definitions 38
Unit (U)
The number of parts, sub-assemblies, assemblies, or systems
inspected or tested
– Squares: 4 units
Opportunity (OP)
A characteristic you inspect or test
– Circles: 5 opportunities per unit
Defect (D)
Anything that results in customer dissatisfaction. Anything that
results in a non-conformance.
– Black circles: 9 defects
Clarification
Defect–Any single non-conformance (defect=black dot)
DMAIC GB O TX PG V 4.2.0
D M A I C
Formulas 39
Total Opportunities
TOP = U*OP
4*5 = 20
DMAIC GB O TX PG V 4.2.0
D M A I C
ConvertingDPMO
Converting DPMOTo
ToZZST 40
DMAIC GB O TX PG V 4.2.0
D M A I C
DPMO Calculation 41
Example
Errors Detected
28 – Wrong amount
14 – Wrong address
12 – Improper accounting code
30 Invoices
Mailed Late
Customer CTQs
Invoice mailed on date specified
Invoice is error free
– Correct address
– Correct amount
DMAIC GB O TX PG V 4.2.0
D M A I C
42
Calculating Process Sigma Discrete Data Method
DMAIC GB O TX PG V 4.2.0
D M A I C
43
Exercise–Calculating Process Sigma Discrete Data
D ( )
DPO = = =
U∗OP ( )( )
3. Click “OK”
DMAIC GB O TX PG V 4.2.0
D M A I C
44
Activity–Calculating Process Capability–Discrete Data
DMAIC GB O TX PG V 4.2.0
D M A I C
45
Six Sigma Product Report
DMAIC GB O TX PG V 4.2.0
D M A I C
46
Discrete Data Examples
2. Select OK.
DMAIC GB O TX PG V 4.2.0
D M A I C
47
Six Sigma Product Report Output
Characteristic Defs Units Opps TotOpps DPU DPO PPM ZShift ZBench
U DPMO ZST
D
OP
DMAIC GB O TX PG V 4.2.0
D M A I C
48
Yield
DMAIC GB O TX PG V 4.2.0
D M A I C
49
Types Of Yield
Classical Yield:
YC = ¾
# parts making it through final inspections = 3
Total # Parts = 4
First-Time Yield
YFT = ¼
# Parts that are acceptable after first inspection = 1
Total # Parts = 4
DMAIC GB O TX PG V 4.2.0
D M A I C
DMAIC GB O TX PG V 4.2.0
D M A I C
Right
First
YRT = .9953*.9713 *.9452 = 91.38% Time
DMAIC GB O TX PG V 4.2.0
D M A I C
Quantification Of Defects 52
Knowing the observed defect level and the test effectiveness, we can estimate
submitted defects and escaping defects.
For example:
Inspection
E = 0.8
Submitted Escaping
DPU Level DPU Level
(DPUS ) (DPUE )
DMAIC GB O TX PG V 4.2.0
D M A I C
Given:
Observed Defects = DPU of 0.25 and E = .8
Then:
Submitted Defects = DPUS = DPUO /E = (0.25/0.8)
= DPUS = 0.31
Escaping Defects = DPUE = DPUS - DPUO = (0.31 - 0.25)
= DPUE = 0.06
Inspection
E = 0.8
Submitted Escaping
DPUS Level DPUE Level
= 0.31 (calc) = 0.06 (calc)
DMAIC GB O TX PG V 4.2.0
D M A I C
DMAIC GB O TX PG V 4.2.0
D M A I C
Exercise (continued) 55
Inspection
E = 0.95
Submitted Escaping
DPUS Level DPUE Level
DMAIC GB O TX PG V 4.2.0
D M A I C
Answers To Exercise 56
1. DPUE = 1 = 0.002
500
2. Maximum acceptable submitted defects =
DMAIC GB O TX PG V 4.2.0
D M A I C
DMAIC GB O TX PG V 4.2.0
D M A I C
Summary–Z-Shift 58
DMAIC GB O TX PG V 4.2.0
D M A I C
ANALYZE STEP
OVERVIEW
DMAIC GB P TX PG V 4.2.0
D M A I C
Use benchmarking
Use other sources
DMAIC GB P TX PG V 4.2.0
D M A I C
DMAIC GB P TX PG V 4.2.0
D M A I C
DMAIC GB P TX PG V 4.2.0
D M A I C
DMAIC GB P TX PG V 4.2.0
D M A I C
DMAIC GB P TX PG V 4.2.0
D M A I C
DMAIC GB P TX PG V 4.2.0
D M A I C
DMAIC GB P TX PG V 4.2.0
D M A I C
Sources Of Information 9
DMAIC GB P TX PG V 4.2.0
D M A I C
Project Goal(s)
Reduction of deliveries that are too early or too late
Benchmarking Process
Identified 20 similar 3rd party Logistics providers from Logistician Magazine
Selected 3PL (2) that provide supply chain solutions to the petroleum industry
(Non-competitors–Capital Logistics strategically avoids supporting petroleum
industry because of risks)
Polled logistics managers (telephone polls) to determine delivery practices
Results
Identified one viable new delivery method…The Dynamic Routing System
– Computer system that selects routes based on individual driver performance for
load types and regional deliveries
Invited Petro-hauz to Stamford, CT for product demo with Capital Logistics
Engineering and Technology Teams
DMAIC GB P TX PG V 4.2.0
D M A I C
Benchmarking Example 11
Characteristic Defs Units Opps TotOpps DPU DPO PPM ZShift ZBench
DMAIC GB P TX PG V 4.2.0
D M A I C
Summary–Analyze 5 12
DMAIC GB P TX PG V 4.2.0
D M A I C
ANALYZE STEP
OVERVIEW
DMAIC GB Q TX PG V 4.2.0
D M A I C
Transfer Function
Y = f (X)
DMAIC GB Q TX PG V 4.2.0
D M A I C
Process
Input
Measures Outputs
(X’s) (Y’s)
X X X X
Process Measures
(X’s)
Definitions: Process
Project Y (Dependent Variable) Provides context for improvement
Product or service produced or delivered by the Links outputs (Y’s) to inputs (X’s) and
process process variables (X’s)
Process X’s (Independent Variable)
Those variables that influence the output and are In Analyze 6 we will be analyzing upstream
generally controllable by those who operate the variables, or input and process variables (X’s), to
process determine how they affect output variables (Y’s)
and to what extent.
Input X’s (Independent Variable)
Materials and information used by the process to create Remember Y=f (X1, X2, ..., Xn)
the outputs. Inputs are often outside the control of the
process operator.
or in words….
The results we get (the Y’s) are a function of the
process and input variables (the X’s). In Analyze,
we will study the process and the data in order to
gain understanding of how it all fits together.
DMAIC GB Q TX PG V 4.2.0
D M A I C
DMAIC GB Q TX PG V 4.2.0
D M A I C
Problem
Statement
DMAIC GB Q TX PG V 4.2.0
D M A I C
Continuous Y Discrete Y
Boxplot Pareto Chart
Scatterplot
Histogram
DMAIC GB Q TX PG V 4.2.0
D M A I C
For each scenario described below, which tool would you use to investigate? For each
scenario, what pattern would you look for to make your conclusions?
1. (a) A Six Sigma project is being conducted in the field to improve the cycle time for
warranty repair returns. The warranty return cycle time was measured for a period of 6
weeks for 4 regions. The Green Belt wants to investigate whether there is any
difference in average warranty repair cycle time between each of the regions.
(b) The Green Belt would also like to know if there is any difference in the variance of
the cycle time between each of the regions.
2. Checks Are Us is a payroll processing firm. Timecard errors are routinely monitored
and recorded. A Black Belt investigating the errors wishes to determine if there are
any differences in the number of errors between five of its major customers. The
number of errors contained in a sample of 150 employees was recorded for five
weeks. How would you proceed with stratifying the data to make this investigation?
3. Tungsten steel erosion shields are fitted to the low pressure blading in steam turbines.
The most important feature of a shield is its resistance to wear. Resistance to wear
can be measured by abrasion loss, which is thought to be associated with the
hardness of steel. How would you investigate the relationship between hardness (the
X) and abrasion loss (the Y)?
DMAIC GB Q TX PG V 4.2.0
D M A I C
Load types: D/H Drop & Hook–A driver unhooks their full trailer, and then hooks into an empty to take
back. Unload–The driver opens the trailer and allows the store to unload its freight, and brings back
the trailer.
To create Boxplots:
Stat > Basic Statistics: > Display Descriptive Statistics
variable: time
click: by variable: load type or driver
click Graph > Choose Boxplots
DMAIC GB Q TX PG V 4.2.0
D M A I C
Types Of Analysis
Moments of truth–What does the customer feel?
Nature of work–What is of value to the customer?
Flow of work–What makes the customer wait?
The purpose of Process Map Analysis is to: The second, Nature of Work, is concerned with value
View our process and its performance from the analysis: understanding whether or not the work in your
customer’s perspective process is valued by the customer or by your internal
culture.
Identify current process problems and opportunities
Create a shared sense of urgency for improving the The third, Flow of Work, characterizes the work along
the dimension of time. The two latter components are
process clearly identify and document value-
focused on the “Profitability” piece of the GE Capital
added/nonvalue-added activities and process flow
Quality Vision.
measures and characteristics.
DMAIC GB Q TX PG V 4.2.0
D M A I C
Moments Of Truth 10
Customer Moments
of Truth
Intermediaries
Or Front Line
Backroom
Activity
Support Or
Suppliers
A moment of truth is defined as anytime a customer “If you’re a service person and you get it wrong at your
draws a critical judgment, positive or negative, about the point in the customer’s chain of experience, you are very
service, based upon a service experience (or lack likely erasing from the customer’s mind all the memories
thereof). Certainly every time we “touch” the customer is of the good treatment he or she may have had up until
a moment of truth, but it isn’t limited to these cases. now.
But if you get it right, you have a chance to undo all the
From the customer’s view, the process is a set of service wrongs that may have happened before the customer got
encounters. In this case, the customers believe the steps to you. You really are the moment of truth.”
should be simple: apply for the loan and receive the
money.
DMAIC GB Q TX PG V 4.2.0
D M A I C
Value-Enabling Work
DMAIC GB Q TX PG V 4.2.0
D M A I C
Control/Inspection Move
DMAIC GB Q TX PG V 4.2.0
D M A I C
Flow Of Work 13
Process Time
+ Delay Time
Cycle Time
Delay Time
Total time in which a unit of work is waiting to have
something done to it.
DMAIC GB Q TX PG V 4.2.0
D M A I C
Gaps
Redundancies
Implicit or unclear requirements
Inefficient hand-offs
Conflicting objectives
Common problem areas
Process disconnects can often be found where there are Inefficient hand-offs: no check is in place to assure the
delays in the process. These are points in the process process is continuing without delays. For example,
where work can be disrupted or delayed, or where defects Department A sends something to Department B but
neither has a way to know if it was properly received.
can be created. They include:
Conflicting objectives: the goals of one group cause
Gaps: responsibility for a given step in the process is problems or errors for another. For example, one group
unclear, or the process seems to go off track. is focused on process speed while another is oriented to
Redundancies: duplication of efforts such as when two error reduction–the result may be that neither group
people or groups approve a document. Redundancies accomplishes its objectives.
occur when different groups take action that they are Common problem areas: occurs when steps are
unaware is being done somewhere else in the process. repeated in a variety of places in process. Noting these
areas may provide insight into potential solutions.
Implicit or unclear requirements: operational
definitions do not exist or clear differences exist in
perspective or interpretation. Your team should encode these “disconnects” and
highlight them directly on your process map.
DMAIC GB Q TX PG V 4.2.0
D M A I C
Unclear
requirements
1. Receive
Are we Yes
application in 11. Make loan extending
mail and open decision
Redundancy loan?
envelope
No 12. Generate
2. Place 10. Review for
Inefficient loan packet
application in completeness
hand-off 19. Generate
mail slot and make
turndown letter
decision
Unclear
4. Place requirements Inefficient 14. Move to 17. Place in
8. Score hand-off
application in mailroom outbound mail
application
in-box basket
5. Retrieve 7. Enter
Yes 15. Wait for 16. Post package
application and Is application application
postage or letter
review for complete? to computer
completeness system
No
6. Call to obtain
necessary
information
Cycle time is the total time taken from the point at which Constructing a time line helps identify areas of potential
the customer requests a good or service until the good or improvement by illustrating the time required for each
service is delivered to the customer. It is the sum of the step in a process.
process time, move time, inspect time, delay time, and
storage time. Remember:
Time data may not be available for all steps in a process.
Cycle time analysis often provides eye-opening insights. Some possible sources include:
People seldom think of work in this context. When an Time, schedule, and activity logs
Improvement Team presents cycle time data, Files, studies, and reports
management may become defensive. For example, when
one manager at a commercial insurer found out it was If data is not available:
taking 100 days, on average, to process the commercial
line of a new business, she resisted the analysis. After
Estimate and ten track to determine actual average
heated “discussion,” she suggested that the period
value
sampled did not reflect normal conditions. She
recommended looking at another period of time to get a
“better number.” The team obliged. The next sample Use minimum of 30 data points to determine an average
averaged 103 days! value when calculating averages, also record the highest
and lowest time so consistency is understood.
DMAIC GB Q TX PG V 4.2.0
D M A I C
Summarized Analysis
% %
Process Step 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 Total Total Steps
Est. Avg. Time (Mins) 1 120 15 120 3 180 7 1 120 5 10 15 90 15 120 2 120 5 8 957 100%
Value-Added 9 9 99 9 99 48 5.0%
Nonvalue-Added
External Failure
Control/Inspection 9 9 8 .8%
Prep/Set-Up
Move 9 9 30 3.1%
Value-Enabling 9 1 .1%
Linking value analysis with cycle time analysis creates a Conclusion: Example
compelling business case for change. 7 steps (-40% of total steps) provide 100% value and
take 48 minutes or 7% of total cycle time.
Displaying this information on process maps emphasizes 11 steps (-61% of steps) are non-value-added, and
focus areas for improvement. consume 94.9% of cycle time.
Tips/Traps
Don’t get stuck on deciding if a step is
value-added or nonvalue-added
– Allow the team to discuss it.
– Consensus with 2/3s majority vote.
Push the team to make the process map a true “as is.”
– The ratio for an average “as is” map is 20% value-
added, 80% nonvalue-added.
– Think of the process map from the point of view of
the product or service.
DMAIC GB Q TX PG V 4.2.0
D M A I C
DMAIC GB Q TX PG V 4.2.0
D M A I C
DMAIC GB Q TX PG V 4.2.0
D M A I C
DMAIC GB Q TX PG V 4.2.0
D M A I C
IMPACT
C In Our
O Control
N
T
R
O
L
Out Of
Our Control
Prioritization Steps
DMAIC GB Q TX PG V 4.2.0
D M A I C
Example
Machines Methods Materials
Why Is There
Difference In The
Variation In Cycle
Time Between Small
And Medium Loans?
Measure- Mother People
ments Nature
IMPACT
High Medium Low
Too many
defects
Complicated Too long for Complexity
C In Our form customer Evaluation
Control Too much number of risk
O review worthiness
N Duplication
of effort
T
R
Too long to Not enough
O Out Of get credit staff
report
Our Control Not well
L trained
DMAIC GB Q TX PG V 4.2.0
D M A I C
DMAIC GB Q TX PG V 4.2.0
D M A I C
IMPACT
C In Our
O Control
N
T
R
O
L
Out Of
Our Control
DMAIC GB Q TX PG V 4.2.0
D M A I C
24
Analyze 6.2
Narrow list of Potential Causes
DMAIC GB Q TX PG V 4.2.0
D M A I C
Sample: A subset of
members that possesses
the same characteristics as
that of the population.
DMAIC GB Q TX PG V 4.2.0
D M A I C
Hypothesis Testing–Introduction 26
DMAIC GB Q TX PG V 4.2.0
D M A I C
Country A
Country B
60.0 62.0 64.0 66.0 68.0 70.0 72.0 74.0 76.0 78.0 80.0
[inch]
DMAIC GB Q TX PG V 4.2.0
D M A I C
To improve processes, we need to identify factors which This way everyone makes the same decisions.
impact the mean or standard deviation.
DMAIC GB Q TX PG V 4.2.0
D M A I C
Kinds Of Differences 29
Continuous data:
Differences in averages
Differences in variation
Differences in distribution
“shape” of values
Discrete data:
Differences in proportions
DMAIC GB Q TX PG V 4.2.0
D M A I C
Hypothesis Testing 30
DMAIC GB Q TX PG V 4.2.0
D M A I C
Nature Of Hypothesis 31
DMAIC GB Q TX PG V 4.2.0
D M A I C
Write the null and alternate hypothesis testing statements for each
Scenario 1: You
scenario have collected delivery time of supplier A and supplier B. You wish to test
below:
whether or not there is a difference in delivery time from supplier A and B.
Scenario 1: statement
Null hypothesis You have : collected delivery time of supplier A and supplier B.
You wish
Alternate to test statement:
hypothesis whether or not there is a difference in delivery time from
supplier A and B.
Scenario 2: You suspect that there is a difference in cycle time to process purchase orders in
site 1 of your company compared to site 2. You are going to perform a hypothesis test to verify
Null hypothesis statement :
your hypothesis.
Scenario 2: You
Scenario 3: You have suspect
implemented that there
process is a difference
improvements in the
to reduce cycle time
cycle time to process
to process
purchase orders
purchase in your
orders incompany.
site 1 ofYou havecompany
your collected cycle time before
compared tothe process
site 2. You are
improvements and after the process improvement was implemented. You are going to perform a
going to perform
hypothesis a that
test to verify hypothesis
the processtest to verifyhave
improvements yourresulted
hypothesis.
in a reduction in cycle
time.
Null hypothesis
Null hypothesis statement
statement : :
Alternate hypothesis statement:
Alternate hypothesis statement:
DMAIC GB Q TX PG V 4.2.0
D M A I C
Hypothesis Testing 33
DMAIC GB Q TX PG V 4.2.0
D M A I C
Innocent, Guilty,
Ho
Set Free Set Free
Set Free Type II
β
Verdict
Verdict
Innocent, Guilty,
Ha Jailed Jailed
Jailed Type I
α
DMAIC GB Q TX PG V 4.2.0
D M A I C
Hypothesis Testing 35
P < α: Reject Ho
P > α : Accept Ho
Reject Ho Accept Ho
Different Not Different
0.00 0.05 1.00
p-value
DMAIC GB Q TX PG V 4.2.0
D M A I C
Some basic statistical tests are shown below with the command for running each
test in Minitab.
Chi-Square Test
Frequency
DMAIC GB Q TX PG V 4.2.0
D M A I C
Y
Continuous Discrete
DMAIC GB Q TX PG V 4.2.0
D M A I C
DMAIC GB Q TX PG V 4.2.0
D M A I C
DMAIC GB Q TX PG V 4.2.0
D M A I C
Regression
Are Y’s NO CHI
Continuous?
SQUARE (χ2)
YES
Are We
Can I Match NO Comparing NO
X’s With X’s? To A 2 Sample t–Test
Standard? HOV
YES YES
Paired t 1 Sample t
Note: In order to use this chart, we are assuming our X’s Paired t-Test
are discrete. Otherwise, use Regression. Ho = Difference = 0
Ha = Difference ≠ 0
Moods Median Test
One Sample t-Test
Ho = Median1 = Median 2 = ...Median n
Ho = µ = Target
Ha = At least one Median is Different
Ha = µ ≠ Target
Homogeniety of Variance (Hov) – Levene’s Test
Two Sample t-test
Ho = σ12 = σ 22 = σ 32 ...σ 2n Ho = µ 1 = µ 2
Ha = At least σ 2 is Different Ha = µ 1 ≠ µ 2
Chi Square
Ho = χ 2 = 0
Ha = χ 2 ≠ 0
DMAIC GB Q TX PG V 4.2.0
D M A I C
1. A six-sigma project is being conducted in the field to improve the cycle time for
warranty repair returns. The warranty return cycle time was measured for a
period of 6 weeks for 4 regions. The Green Belt suspects that there is a
difference in average warranty repair cycle time among each of the regions.
How would you test whether there is a statistically significant difference in mean
cycle time for the different regions?
2. Tungsten steel erosion shields are fitted to the low pressure blading in steam
turbines. The most important feature of a shield is its resistance to wear.
Resistance to wear can be measured by abrasion loss, which is thought to be
associated with the hardness of steel. How would you test whether there is a
statistically significant relationship between resistance to wear and abrasion
hardness of steel?
3. Your business purchases sheet stock from two different suppliers. It has found
an unacceptably large number of defects being caused by thickness beyond
tolerance levels. Data for overall mean thickness data was analyzed and found
to be on target. Data was collected that would identify a potential difference in
the variation of the thickness of the material by supplier.
DMAIC GB Q TX PG V 4.2.0
D M A I C
Data Analysis 42
DMAIC GB Q TX PG V 4.2.0
D M A I C
File Description:
DMAIC GB Q TX PG V 4.2.0
D M A I C
Stability–Minitab
Stability–Minitab Run Chart
Run Chart 44 44
1. Double
Click.
2. Type in a “1.”
It is important to note that the data should be in time Test for Trends: Ho: No fewer runs observed than
order for run charts to be valid. expected. Ha: More runs observed than expected. Trends
result from a gradual increase or decrease in the measure
A run is one or more consecutive points in the same so that there are fewer runs than expected by sampling
direction. A new run begins each time there is a change error alone.
in direction. Do not count points exactly on the median.
Test for Oscillations: Ho: No more runs observed than
Test for Clusters: Ho: No fewer runs observed than expected. Ha: More runs observed than expected.
expected. Ha: Fewer runs observed than expected. Oscillations are similar to mixtures in that they indicate
Clusters result from sampling from one process for a too many runs—more than expected
period of time and then from another process with a by sampling error alone.
different center for a period so there are fewer runs than
expected by sampling error alone.
Since all p-values are less than 0.05, the process is stable
for the south territory. You would also need to check
Test for Mixtures: Ho: No more runs observed than
each of the other territories for stability.
expected. Ha: More runs observed than expected.
Mixtures result from two or more processes with different
centers alternately sampled so there are too many runs
than expected by sampling error alone.
© GE Capital, Inc., 2000
DMAIC GB Q TX PG V 4.2.0
D M A I C
p < α: Reject Ho
p > α: Accept Ho
DMAIC GB Q TX PG V 4.2.0
D M A I C
Data Analysis 46
DMAIC GB Q TX PG V 4.2.0
D M A I C
DMAIC GB Q TX PG V 4.2.0
D M A I C
1. Double
Click
3. Click
OK 2. Click on Graphs,
Select Graphical Summary
Descriptive Statistics
Variable: South
Non-normal
Anderson-Darling Normality Test distribution!
A-Squared: 3.095
P-Value: 0.000
Mean 11.9348
StDev 11.1064
Variance 123.351
Skewness 1.75318
Kurtosis 2.69797
N 46
0 10 20 30 40
Minimum 1.0000
1st Quartile 4.0000
Median 9.0000
3rd Quartile 14.2500
95% Confidence Interval for Mu Maximum 45.0000
95% Confidence Interval for Mu
8.6366 15.2330
7 8 9 10 11 12 13 14 15 16 95% Confidence Interval for Sigma
9.2120 13.9887
95% Confidence Interval for Median
95% Confidence Interval for Median
6.9173 12.0000
DMAIC GB Q TX PG V 4.2.0
D M A I C
Non-Parametric Tests 49
DMAIC GB Q TX PG V 4.2.0
D M A I C
Non-Parametric Tests 50
DMAIC GB Q TX PG V 4.2.0
D M A I C
Spread–Test On Variances 51
Xb
Xa
DMAIC GB Q TX PG V 4.2.0
D M A I C
Spread–Homogeneity Of Variance 52
DMAIC GB Q TX PG V 4.2.0
D M A I C
2 = 2 = 2 = 2 = 2
Ho : σEast σ West σ IFG σ Central σ South
Ha : at least one is different α = .05
DMAIC GB Q TX PG V 4.2.0
D M A I C
2 = 2 = 2 = 2 = 2
Ho : σEast σ West σ IFG σ Central σ South
α = .05
Ha : at least one is different
Homogeneity of Variance Test for total time
95% Confidence Intervals for Sigmas Factor Levels
Central
Bartlett's Test
IFG
Levene's Test
West
5 15 25 35 45 55
We have standardized on
The Levene’s test because
it is more robust to non-
p > α: Accept Ho
DMAIC GB Q TX PG V 4.2.0
D M A I C
DMAIC GB Q TX PG V 4.2.0
D M A I C
~ ~ ~ ~ ~
H :X = X = X = X = X
o 1 2 3 4 5
DMAIC GB Q TX PG V 4.2.0
D M A I C
DMAIC GB Q TX PG V 4.2.0
D M A I C
Recall, when two groups are equal, the confidence intervals overlap.
DMAIC GB Q TX PG V 4.2.0
D M A I C
4. Chose hypothesis test Choose the correct test, given the type of X and Y data.
5. Gather evidence and test/conduct analysis Collect data, run analysis, get p-value
6. Decide to Reject H0, or not reject H0, and If p ≥ 0.05 conclude, no difference between populations
draw conclusion If p < 0.05 conclude, the populations are different
DMAIC GB Q TX PG V 4.2.0
D M A I C
1-Sample Hypothesis 60
1. Ho : µ = constant = T
Ha : µ ≠ constant = T
Ho Ha
2. Ho : σ 2 = constant = T
Ha : σ 2 ≠ constant = T
DMAIC GB Q TX PG V 4.2.0
D M A I C
1-Sample t-Test 61
Ho : µ = constant = T
Ha : µ ≠ constant = T
DMAIC GB Q TX PG V 4.2.0
D M A I C
1-Sample t-Test 62
DMAIC GB Q TX PG V 4.2.0
D M A I C
1-Sample t-Test 63
1. Double click
anywhere on the C5
line to select time as
a variable
4. Click on
Histogram of data
5. Click OK
DMAIC GB Q TX PG V 4.2.0
D M A I C
1-Sample t-Test 64
Histogram of time
(with Ho and 95% t-confidence interval for the mean)
50
40
Frequency
30
20
10
0 _
X Ho
[ ]
-100 0 100
time
DMAIC GB Q TX PG V 4.2.0
D M A I C
2-Sample Hypothesis 65
3. Ho :µ1 = µ2
Ha : µ1≠ µ2
4. Ho : µ1 ≤ µ2
µ1 µ2
Ha : µ1 > µ2
5. Ho : σ1 2 = σ22
Ha : σ1 2 ≠ σ22
σ1
σ2
DMAIC GB Q TX PG V 4.2.0
D M A I C
2-Sample t-Test 66
Ho :µ1 = µ2
Ha : µ1≠ µ2
DMAIC GB Q TX PG V 4.2.0
D M A I C
2-Sample t-Test 67
DMAIC GB Q TX PG V 4.2.0
D M A I C
2-Sample t-Test 68
1. Double click
anywhere on the
C5 line to select
time as a variable
for samples.
2. Click on C3 to chose
LoadType as the
variable for
subscripts.
3. Click Graphs to
show a pictorial.
4. Click on
Boxplots of data.
5. Click OK.
DMAIC GB Q TX PG V 4.2.0
D M A I C
2-Sample t-test 69
100
time
-100
D/H Unload
LoadType
DMAIC GB Q TX PG V 4.2.0
D M A I C
6. Ho : µ1 = µ2 = . . . = µn
DMAIC GB Q TX PG V 4.2.0
D M A I C
Ho : µ1 = µ2 = . . . = µn
DMAIC GB Q TX PG V 4.2.0
D M A I C
DMAIC GB Q TX PG V 4.2.0
D M A I C
1. Double click
anywhere on the C5
line to select time as
a response variable.
2. Double click C2
for Driver as a
factor.
3. Click Graphs
to show a
pictorial.
4. Click on
Boxplots of data.
5. Click OK.
DMAIC GB Q TX PG V 4.2.0
D M A I C
100
time
-100
Driver
Trans, Inc.
Temp Labor
Cap. Log.
JJ Truck
DNGY Inc.
2. These 3 are
alike.
3. These 2 are
definitely
different.
DMAIC GB Q TX PG V 4.2.0
D M A I C
DMAIC GB Q TX PG V 4.2.0
D M A I C
DMAIC GB Q TX PG V 4.2.0
D M A I C
3. Click OK.
DMAIC GB Q TX PG V 4.2.0
D M A I C
1. We have standardized
our Levene’s Test
because it is more
robust for non-normal
data.
2. Our p-value is
less than 0.05.
DMAIC GB Q TX PG V 4.2.0
D M A I C
R-value
y r = –1.0 y r = +1.0
x x
r = +.7 r = –.7
y y
x x
y r=0 y r=0
x x
–1 ≤r ≤1
DMAIC GB Q TX PG V 4.2.0
D M A I C
DMAIC GB Q TX PG V 4.2.0
D M A I C
Y = b 0 + b 1X 1
DMAIC GB Q TX PG V 4.2.0
D M A I C
DMAIC GB Q TX PG V 4.2.0
D M A I C
3. We believe there to
be a linear
relationship.
4. Click OK.
DMAIC GB Q TX PG V 4.2.0
D M A I C
Regression Plot
Y = -29.1923 + 0.174971X
R-Sq = 10.5 %
100
0 - distance.
-100
Entire model is
significant.
Ho: Model = O
Ha: Model ≠ O
DMAIC GB Q TX PG V 4.2.0
D M A I C
Chi-Square Test 85
Y
Continuous Discrete
Continuous
X
Discrete
Chi-Square Test
DMAIC GB Q TX PG V 4.2.0
D M A I C
Chi-Square Tests 86
χ2
The Chi-Square test provides a way to determine if a
given set of data has the hypothesized distribution. For
example, if we roll a die 600 times., the number 1 occurs
96 times, 2–88 times, 3–104 times, 4–107 times, 5–112
times, and 6–93 times. In theory, each number should
occur 100 times. The Chi-Square test enables us to
determine if the observed set of data, the output of 600
rolls of a die, fits an expected model.
The Chi-Square test will also let you know if any factor
has an impact on the output. If there is no difference in
the data (Ho), this means no factor(s) have an impact on
the output. If a difference exists (Ha), one or more
factors have an impact on the output.
DMAIC GB Q TX PG V 4.2.0
D M A I C
g 2
fo - fe
χ =
2
Σ fe
j=1
g = number of categories
Suppose we flip a coin N
fo = observed frequency
= 100 times and observe
fe = expected frequency
63 heads and 37 tails.
Could this ratio of heads
to tails occur by chance or (f0-fe)2
should we conclude that fe
Penny
the coin is somehow
biased ?
Heads 63/50* 3.38
3.38
Tails 37/50*
*=Observed Expected
(f0) (fe) χ 2 = 6.76
DMAIC GB Q TX PG V 4.2.0
D M A I C
DMAIC GB Q TX PG V 4.2.0
D M A I C
Degrees Of Freedom 89
DMAIC GB Q TX PG V 4.2.0
D M A I C
Since the statistic from the data, 6.76, is greater than the table
critical value, the alternative hypothesis is accepted: we are
confident that we saw unusual behavior with that penny.
Use the tables on the next two pages to determine the chi-
square critical value.
DMAIC GB Q TX PG V 4.2.0
D M A I C
DMAIC GB Q TX PG V 4.2.0
D M A I C
DMAIC GB Q TX PG V 4.2.0
D M A I C
= CHIDIST(statistic,dof)
= CHIDIST(6.76,1)
Excel will return the p-value, in this case 0.0093. Since the
probability of being wrong is less than 0.05, we accept the
alternative hypothesis: we are confident that we saw unusual
behavior with that penny.
DMAIC GB Q TX PG V 4.2.0
D M A I C
Test Of Independence 94
Oppose Total = 65
fo = 30 fo = 35
Favor fo = 15 fo = 20 Total = 35
Total = 45 Total = 55
DMAIC GB Q TX PG V 4.2.0
D M A I C
= (65/100) x 45 = 29.25
Note that for each cell, this is the total frequency for the row, times
the total frequency for the column, divided by the total population
(see chart on following page).
DMAIC GB Q TX PG V 4.2.0
D M A I C
fe = frow x fcol
N
Party 1 Party 2
Total = 45 Total = 55
DMAIC GB Q TX PG V 4.2.0
D M A I C
Minitab Example 97
DMAIC GB Q TX PG V 4.2.0
D M A I C
1. Double click on
both columns.
2. Click OK.
Minitab calculates
the expected
frequency (fe)
DMAIC GB Q TX PG V 4.2.0
D M A I C
DMAIC GB Q TX PG V 4.2.0
D M A I C
DMAIC GB Q TX PG V 4.2.0
D M A I C
1. Double click on
both columns
C2 & C7..
3. Click OK.
DMAIC GB Q TX PG V 4.2.0
D M A I C
χ statistic 2
DMAIC GB Q TX PG V 4.2.0
D M A I C
Desired Outcome: Select and perform the appropriate statistical test for a given
project situation
DMAIC GB Q TX PG V 4.2.0
D M A I C
DMAIC GB Q TX PG V 4.2.0
D M A I C
1. State the H0
2. State the Ha
3. State α level
1. H0: ___________________________________________________________________
2. Ha: ___________________________________________________________________
3. α level: ________________________________________________________________
4. Test: _________________________________________________________________
DMAIC GB Q TX PG V 4.2.0
D M A I C
1. State the H0
2. State the Ha
3. State α level
1. H0: ___________________________________________________________________
2. Ha: ___________________________________________________________________
3. α level: ________________________________________________________________
4. Test: __________________________________________________________________
DMAIC GB Q TX PG V 4.2.0
D M A I C
1
D M A I C
IMPROVE
PHASE
OVERVIEW
Improve 8: Improve 9:
Improve 7: Discover Establish
Screen Variable Operating
Potential Relationships Tolerance & Pilot
Causes & Propose Solutions
Solution
DMAIC GB R TX PG V 4.2.0
D M A I C
DMAIC GB R TX PG V 4.2.0
D M A I C
3
The12-Step Process
DMAIC GB R TX PG V 4.2.0
D M A I C
IMPROVE STEP
OVERVIEW
Improve 8: Improve 9:
Improve 7:
Discover Variable Establish
Screen Potential
Relationships & Operating
Causes
Propose Solution Tolerances & Pilot
Solution
DMAIC GB R TX PG V 4.2.0
D M A I C
Propose Solution
IMPROVE STEP
OVERVIEW
Improve 8: Improve 9:
Improve 7:
Discover Variable Establish
Screen Potential
Relationships & Operating
Causes
Propose Solution Tolerances & Pilot
Solution
8.1 Generate a transfer function to optimize the changes in vital X’s (if necessary)
8.2 Propose solution
8.3 Refine solution
DMAIC GB R TX PG V 4.2.0
D M A I C
6
IMPROVE STEP
OVERVIEW
Improve 8: Improve 9:
Improve 7:
Discover Variable Establish
Screen Potential
Relationships & Operating
Causes
Propose Solution Tolerances & Pilot
Solution
DMAIC GB R TX PG V 4.2.0
D M A I C
Statistical Thinking 7
D M A I C
Practical Statistical Statistical Practical
Problem Problem Solution Solution
Problem Characterize the Root cause Verify critical
statement process analysis X’s and ƒ(x)
– Project Y – Stability – Critical X’s Change
– Magnitude – Shape Measure the process
– Impact – Center influence of the Control the
– Variation critical X’s on gains
Data Integrity the mean and – Risk
– MSA Capability variability analysis
– Brainstorm – ZBench ST & LT – Test – Control
potential X’s – Model plans
– Sampling plan – Estimate
Collect data
The Practical-To-Statistical-To-Practical
Transformation Process
DMAIC GB R TX PG V 4.2.0
D M A I C
DMAIC GB R TX PG V 4.2.0
D M A I C
1
Design Of Experiment: Learning Objectives
2
Design Of Experiments
3
DOE Terminology
Independent Variables–X’s
Also called factors Project Y
Factors or variables we select Dependent
in advance
Independent X
The causes (5 M’s and 1 P)
4
DOE Terminology (continued)
X2 = Tire Type
Levels = Brand X & Brand Y
Main Effects For example: Is the On-Time Delivery different for:
a) 62 ft. trucks vs. 66 ft. trucks?
Differences between b) Brand X or Brand Y tires?
each factor level
62 ft.
truck
delivery time.
delivery time.
66 ft.
truck
Y
X
62 66 X Y
Truck Size Tire Type
Deviation from required
delivery time.
Y
X
X Tire Type Y
5
Strategy For Choosing The Appropriate Design
Purpose:
Identify Main effect Some Relationships Optimal factor
critical factors interactions among factor settings
- vital few
Screening DOE–experiments which are used to screen Fractional Factorials–provide a middle ground. They
out the vital few X variables out of many at a very low are used to expand on the study of the vital few X
variables at a low cost.
cost.
Can study the main effects
Can study main effects
Can study selected (and few in number)
Can’t study the interactions
interactions
Optimization DOE–Experiments which are used to Full Factorials–are used to study the vital few X
determine the proper settings for each vital X. variables and all possible interactions. They can be
quite costly when the number of vital few X’s are
large.
Can study main effects
Can study all interactions
6
DOE Benefits
7
The Process Of Experimentation
8
The Process Of Experimentation
9
The Process Of Experimentation (continued)
3. Perform Analysis
Dependent Variable–Gas Mileage (Y)
Identify Factors (independent variables)
– Tire Pressure (X)
– Octane (X)
– Speed (X)
Choose Factor Levels
Independent Variables
(X’s) (Factors) Level - Level +
Tire Pressure (psig) 30 35
Octane 87 92
Speed (mph) 55 65
10
The Process Of Experimentation (continued)
11
The Process Of Experimentation (continued)
12
The Process Of Experimentation (continued)
B
A: Pressure
A
B: Octane
BC
C: Speed
AC
ABC
AB
0 1 2 3 4 5 6 7 8 9
Effect Size
13
The Process Of Experimentation (continued)
- + – + – +
28
26
Mileage
24
22
20
The Main Effects Plot show the average value for each of
the levels of the three independent variables. Each
average is based on 4 data points.
14
The Process Of Experimentation (continued)
30 33
+
-, -, + +, -, +
18 21
Octane +
Speed
– 27
–
26
– +
-, -, - Tire Pressure +, -, -
Each corner of the cube represents an experimental trial from the trial table
For example, when tire pressure is negative (-), octane is negative (-) and
speed is negative (-), the gas mileage is 26 (x,y,z is the convention)
Can you tell which combination of tire pressure, octane and speed is the
best by looking at the cube?
15
The Process Of Experimentation (continued)
5. Record Results
6. Standardize
7. Determine Future Plans
16
17
Full Factorial DOE
18
Full Factorial Layout: 3 Factors, 2 Levels
This is a 3–factor, 2–level full factorial layout arranged in Does this full factorial layout look familiar?
“standard order.”
Where did you see this before?
Each row of the full factorial layout represents a set of
experimental conditions. There are 8 different
experimental conditions in this design layout.
19
Visualizing The Experimental Space
2 Factors = A SQUARE +
Factor
2
–
– +
Factor 1
3 Factors = A CUBE
Factor 2
+
– –
– Factor 3
+
Factor 1
20
The Factorial Pattern Of Experimentation
k Factors Order X1 X2 X3 X4 X5 • • Xk
•
1 – – – – –
k=1 2 + – – – –
3 – + – – –
k=2 4 + + – – –
5 – – + – –
6 + – + – –
7 – + + – –
k=3 8 + + + – –
9 – – – + –
10 + – – + –
11 – + – + –
12 + + – + –
13 – – + + –
14 + – + + –
15 – + + + –
k=4 16 + + + + –
17 – – – – +
18 + – – – +
19 – + – – +
20 + + – – +
21 – – + – +
22 + – + – +
23 – + + – +
24 + + + – +
25 – – – + +
26 + – – + +
27 – + – + +
28 + + – + +
29 – – + + +
30 + – + + +
31 – + + + +
k=5 32 + + + + +
•
•
•
k 2k
21
Replication vs. Repetition
22
Randomization: The Experimenter’s Insurance
Why Randomization?
– Averages the effect of any lurking variables over all of the factors in the
experiment
23
Design Of Experiment
Summary
Provides process performance data under different types of operating
parameters
Establishes cause and effect relationships between a process Y and
possible X’s:
– Verify the Vital Few X’s
– Determine X values for optimal performance
– Predict future process performance
Can be used in both incremental and exponential process improvement
studies
24
Reference Material–
Analyzing DOE Data
25
The Process Of Experimentation
26
Steps in Analysis: Full Factorial, Replicated Designs
This is the “Analyze Data” step, which is one step in the After using Minitab to check the data, the next step is to
process of experimentation on the previous page. plot the residuals.
Steps in analyzing the data: Residuals are the “leftover” variation in the data after you
have accounted for the main cause of variation. Plotting
the residuals should show a normal distribution. If the
In steps 1 & 2, we begin by checking for problems with
residuals are not normally distributed, it is a sign that
the data before doing the DOE analysis on the data.
there is another variable present that has not been
Minitab will plot the data for you. This will help to: accounted for (a “lurking” variable).
Find defects, outlier or patterns in the data
Get a feel for the actual data * We don’t discuss residuals here, but to continue to Step
3, seek assistance from someone who knows how.
Get impressions of what the data is going to tell us, to
guide our analysis
Once the data has been validated and looks acceptable,
you are ready to examine the factor effects.
27
Background For DOE: Capital Logistics GB Case Study
In Step 6 of the Analyze Phase, the Green Belt and the project team at Capital Logistics determined,
based on the results of their Hypothesis Testing, that there were 2 X’s that had an impact on the on-
time delivery of shipments to customers. The 2 X’s were:
– Drivers
– Distance
The team determined that there wasn’t anything that they could do to change the distance that each
truck traveled to make their deliveries–the routes were automated by the system and were
determined to be the shortest routes. They decided to focus on Driver as a factor, as this was under
their immediate control.
The team used the Cause & Effect diagram to brainstorm potential root causes as to why
Driver/Trucking company influenced Time. The resulting list included:
– size of the vehicle – type of tire
– dispatch method – fuel capacity
– type of engine
After reviewing their data further, the team decided to conduct a Design of Experiment. They would
use Size of Vehicle and Tire Type as possible Vital X’s in their DOE. Each of these factors has 2
levels:
– Size of Vehicle is either 66 or 62 feet
– Tire Type is either X or Y
The DOE was set-up and conducted to determine the Vital X that would have the greatest impact on
the timeliness of shipment deliveries to their customers.
28
Examine Factor Effects
29
Factor Names And Interaction Symbols
Total (11)
30
Examine Factor Effects–In Minitab
31
Examine Factor Effects–In Minitab
32
Examine Factor Effects–In Minitab
Name Factors
33
Examine Factor Effects–In Minitab
Note: Ensure that when you enter this data into your
Minitab Worksheet, you put the times in the correct rows.
Your experiment may have run in a different order and
you may have coded your factors differently.
34
Examine Factor Effects–In Minitab
35
Examine Factor Effects–In Minitab
36
Examine Factor Effects–In Minitab
What does the Pareto Chart tell you about the factors &
interactions?
37
Examine Factor Effect–In Minitab
38
Examine Factor Effects–In Minitab
39
Examine Factor Effects–In Minitab
40
Main Effect Plots
-1 1 -1 1
90
60
30
Time
-30
41
Interaction Effect
–
–
Factor A Factor A + Factor A
Response
Response
Response
+
+ –
– + – + – +
Factor B Factor B
Factor B
Adapted from Lawson, John and John Erjavac, Basic Experimental Strategies
and Data Analysis. Provo, UT: Brigham Young University, p. 104.
We use the same calculation for Interaction Effects as we If the lines are not parallel, there is interaction between
did for Main Effects. The only difference is we first need the factors (variables).
to generate the interaction column. We then use this
column for the calculation of the Interaction Effect.
42
Interaction Effect
Truck Size
-1
1
100
Late
50
Mean
On-Time
0
Early -50
-1 1
Tire Type
43
Steps In Analysis: Full Factorial, Replicated Designs
44
Confirm Impressions
45
Confirm Impressions With Minitab
Indicate?
46
Steps In Analysis: Full Factorial, Replicated Designs
47
Summarize Conclusions
48
Summarize Conclusions (continued)
They returned to their Fishbone (Cause & Effect) Diagram with greater
granularity, and asked why these factors would influence on-time performance.
They validated these reasons through field observations.
The team validated that the larger size trucks took longer at some locations. In
addition to the time involved with maneuvering in a cramped environment, on
certain routes these trucks had a higher incidence of tire failure (lateral
movement caused by tight turns increased the stress on the tires, increasing the
number of flat tires).
By using the data & results from the DOE, the team had determined the root
causes of the on-time performance problem and were ready to start to
brainstorm solutions to address/control these “Vital X’s”.
49
50
Reducing The Size Of Factorial Experiments
Full factorial is time-consuming and costly as the number From a Full Factorial (4 factors):
of factors increase.
Number
Also consider the number of 2-way and higher order
interactions as the number of factors increases. Overall Average 1
For 4 factors, let’s look at every possible effect (main & Main Effects: A B C D 4
interaction) and the overall average. Evaluating the
average is what the extra run is used for. 2-way interactions 6
3-way interactions 4
4-way interactions 1
(ABCD)
© GE Capital, Inc., 2000
DMAIC GB S TX PG V 4.2.0
D M A I C
51
Fractional Factorials
Fractional Factorials*:
Provide a middle ground for design of experiment
Are used to expand on the study of the Vital Few X variables at
a low cost
Can study main effects
Can study selected (and few in number) interactions
52
Strategy For Choosing The Appropriate Design
Purpose:
Identify Main effect Some Relationships Optimal factor
critical factors interactions among factor settings
- vital few
Screening DOE–experiments which are used to screen Fractional Factorials–provide a middle ground. They
out the vital few X variables out of many at a very low are used to expand on the study of the vital few X
cost. variables at a low cost.
Can study the main effects Can study main effects
Can’t study the interactions Can study selected (and few in number)
interactions
Optimization DOE–Experiments which are used to
determine the proper settings for each vital X. Full Factorials–are used to study the vital few X
variables and all possible interactions. They can be
quite costly when the number of vital few X’s are
large.
Can study main effects
Can study all interactions
Independent Variables
(X’s) (Factors) Level - Level +
55
Design Of Experiments–Breakout Activity Worksheet (Continued)
Independent Variable (X) Independent Variable (X) Independent Variable (X) Dependent
#1 #2 #3 Variable (Y)
Independent Variable (X) Independent Variable (X) Independent Variable (X) Dependent
#1 #2 #3 Variable (Y)
56
Appendix: Example Of Designed Experiment: GE Card Services
57
DOE Example: Defining The CTQ
Overall
Quality
EX/VG - 57%
0.38 Timeliness
0.59 Speed Of 10% 33%
24% 25% Of Delivery
Approval 0.28
0.23 Accuracy 20% 41%
Amount/
18% 26% 0.24
Type Of Info Insert
12% 47%
0.2 Credit Lines 0.17 Capability
16% 25%
Granted Card Quality 23% 53%
1 0.8 0.6 0.4 0.2 0 1 0.8 0.6 0.4 0.2 0
0% 20% 40% 60% 80% 100% 0% 20% 40% 60% 80% 100%
Excellent Very Good Excellent Very Good
Inseparable CTQ’s
#1 Speed #1 Speed
One Experience To The Customer #2 Accuracy
A project team was working to improve customer When examining the process map, it became clear that
satisfaction with the process of approval for and issuing many faxed and mailed applications needed additional
of new credit cards. information before they could be processed. This often
involved phone or written correspondence with the
One of the key drivers of satisfaction was identified as applicant, which slowed the process considerably.
time to issue cards, i.e., the cycle time between
a consumer’s approval for a credit card, and the time
he/she received it.
58
DOE Example: Causes Of Time Consuming Delays
50%
40%
Frequency %
30%
20%
10%
0%
Msg Wrng Msg/Wrng Msg/ Unknown CBR Msg/ Duplicate
Home Promo Brnch/Dlr/ Wrng Issues Wrng
#/AC Cd Str # SS# Zip
59
DOE Example: The Designed Experiment
Insurance Location on the form for the purchase Not Body* Body*
of the optional credit card insurance
Page Layout Orientation of the form Portrait Landscape
60
DOE Example: Experimental Complications
The applications were paired (eight pairs of applications) The completed applications were sent for imaging, and
and customers were asked to fill out a pair of applications the imaged applications were reviewed by an associate
and provide feedback regarding the two application to record the legibility and completeness of each field
versions. The applications were paired so that customers on each application. A legibility score was assigned
experienced the factors at both levels. using a 5-point scale; missing information was given a
“0” and fields not relevant to the applicant were
Eighty customers filled out a pair of applications in return recorded as not being applicable. Preferred applications
for a $10.00 gift certificate. The order which each paired were given a score of “1,” not preferred “0,” and no
application was presented to the customer was preference was indicated by a “0.5.”
randomized throughout the test to balance the effect of
order on the results. Customers were asked whether they
preferred one application version over another and why.
Applicants signing for credit insurance were asked
whether they understood that they had requested
insurance coverage.
61
DOE Example: Responses
62
DOE Example: Analysis
0.849
Completeness
Legibility and
0.843
0.837
0.831
0.825
Naked Full No Yes No Yes No Yes Body Bottom Landscape
Portrait
App. Type Instruct Flags Boxes Insurance Pg. Layout
* * *
Main Effects Plot – Means For Preference
0.67
0.59
Preference
0.51
0.43
0.35
Naked Full No Yes No Yes No Yes Body Bottom Landscape Portrait
63
DOE Example: Results
Instructions H H No Yes
57% 43% Yes1
Flags No Yes H H
82.9% 84.8% Yes
Boxes H H No Yes
59% 41% No
1 The data does not show that instructions improve legibility and completeness, but it is felt that instructions should be provided for clarification
purposes. Also, consideration should be given to improving the effectiveness of the instructions provided on the application.
+ Effect not significant
1
Step 9
DMAIC GB T TX PG V 4.2.0
D M A I C
2
Principle Of Tolerancing
USL
T Tolerance:
LSL the allowable
Y = f (X) range of X
X
Lower Limit Upper Limit
of Tolerance of Tolerance
DMAIC GB T TX PG V 4.2.0
D M A I C
3
Consideration Of Measurement Variations
USL
X
Lower Limit Upper Limit
of Tolerance of Tolerance
USL
The measurement
variation of X Lower Limit Upper Limit
of Tolerance of Tolerance
Estimate the gage error for Y using the σe from the MSA
study in Step 3.
Establish the gage error for X in Step 10. Use the 5.15σe
from the MSA of X.
DMAIC GB T TX PG V 4.2.0
D M A I C
lbs.
week
USL
3
Target
Weight 2 Transfer
Loss Function
1
LSL
0
1 1.4 2 2.7 3 hrs./week
CTQ requirements: 4
Target to lose 2 lbs/week
At least need to lose 1 lb/week
No more than 3 lbs/week due to health concerns
DMAIC GB T TX PG V 4.2.0
D M A I C
5
Example: Weight Loss Wellness Program (continued)
lbs.
The measurement
variation of Y week
USL
3
Target
Weight 2
Loss
1
LSL
New operating limits
on X due to variability
0 in the Y and X
measurements
1 2 hr/week
Exercise
1.55 2.55
DMAIC GB T TX PG V 4.2.0
D M A I C
Target
Weight 2
Loss
1
LSL
0
1 1.5 2 2.6 3 hrs./week
Exercise
New operating limits on X due to
variability in the Y measurements
DMAIC GB T TX PG V 4.2.0
D M A I C
Target
Weight 2
Loss
1
LSL
0
1 1.5 2 2.6 3 hrs./week
Exercise
New operating limits on X due to
variability in the Y measurements
DMAIC GB T TX PG V 4.2.0
D M A I C
8
Summary
DMAIC GB T TX PG V 4.2.0
D M A I C
D M A I C
CONTROL
PHASE
OVERVIEW
DMAIC GB U TX PG V 4.2.0
D M A I C
DMAIC GB U TX PG V 4.2.0
D M A I C
3
The12-Step Process
Step Description Focus Tools Deliverables
Define
A Identify Project CTQ’s Project CTQ’s
B Develop Team Charter Approved Charter
C Define Process Map High Level Process Map
Measure
1 Select CTQ Characteristics Y Customer, QFD, FMEA Project Y
2 Define Performance Y Customer, Blueprints Performance Standard for
Standards Project Y
3 Measurement System Y Continuous Gage R&R, Data Collection Plan & MSA
Analysis test/Retest, Attribute Data for Project Y
R&R
Analyze
4 Establish Process Y Capability Indices Process Capability for
Capabilities Project Y
5 Define Performance Y Team, Benchmarking Improvement Goal for
Objectives Project Y
6 Identify Variation Sources X Process Analysis, Prioritized List of all X’s
Graphical Analysis,
Hypothesis Tests
Improve
7 Screen Potential Causes X DOE-Screening List of Vital Few X’s
8 Discover Variable X Factorial Designs Proposed Solution
Relationships
9 Establish Operating Y, X Simulation Piloted Solution
Tolerances
Control
10 Define & Validate Y, X Continuous Gage R&R, MSA
Measurement System on Test/Retest, Attribute
X’s in Actual Application R&R
11 Determine Process Y, X Capability Indices Process Capability Y, X
Capability
12 Implement Process Control X Control Charts, Mistake Sustained Solution,
Proofing, FMEA Documentation
DMAIC GB U TX PG V 4.2.0
D M A I C
CONTROL STEP
OVERVIEW
DMAIC GB U TX PG V 4.2.0
D M A I C
4
CONTROL STEP
OVERVIEW
DMAIC GB U TX PG V 4.2.0
D M A I C
CONTROL STEP
OVERVIEW
DMAIC GB U TX PG V 4.2.0
D M A I C
Statistical Thinking 7
D M A I C
Practical Statistical Statistical Practical
Problem Problem Solution Solution
Problem Characterize the Root cause Verify critical
statement process analysis X’s and ƒ(x)
– Project Y – Stability – Critical X’s Change
– Magnitude – Shape Measure the process
– Impact – Center influence of the Control the
– Variation critical X’s on gains
Data Integrity the mean and – Risk
– MSA Capability variability analysis
– Brainstorm – ZBench ST & LT – Test – Control
potential X’s – Model plans
– Sampling plan – Estimate
Collect data
The Practical-To-Statistical-To-Practical
Transformation Process
DMAIC GB U TX PG V 4.2.0
8
Process Control Plan Worksheet
Plan
PlanFor
ForDoing
DoingThe
TheWork
Work Checking
CheckingThe
TheWork
Work Response
ResponseTo
ToSignals
Signalsof
ofChange
Change
Process Step Procedure Key Process Method For Corrective Action Preventive Action
Monitoring
(From Process (From & Output Recording For Containment Procedure For Process
Standards
Map Attached) SOPs) Measures Data Improvement
Refer To Diagram Refers To Root Cause X’s For Each For Each Identifies Who Describes What Must Be Done When
That Illustrates Document That Measure, Measure, Should Do What We Fail Process Sigma Standard.
Project Y’s
Process Steps Describes How Describes Any Describes How With The
Describes What Must Be Done To
Separated By The Task Should Process Sigma Target, Numeric The Monitored Defective Output.
Prevent Special Causes That Worsen
Function. Shows Be Done, Or For Each Limits, Or Data Should Be
Describes What Performance Or To Incorporate
Transfers Between Refers To A Project Y Tolerances To Recorded, Who
Should Be Done These Special Causes That Improve
Functions, And Document That Which A Process Should Record
For Those Ill- Performance.
Which Function Is Describes The Should Conform The Data And
Served By These
Responsible At Step. How.
Defects.
Each Step.
D M A I C
Patent
Best Protection against independent development by others
Gives right to exclude others from making, using selling, the invention or 20
years from filing
Can wait 2-3 years for patent to issue
Right granted by the government for new, useful, nonobvious inventions
Trade Secret
Rights exist only so long as actually kept secret
Protects against theft or misappropriation, not independent development
Does not require inventiveness in the patent sense
Copyright
Protects original works of authorship fixed in a tangible medium
Does not protect against independent development
Trademark
Defines source of goods or services
Arbitrary, fanciful marks get stronger protection
DMAIC GB U TX PG V 4.2.0
D M A I C
10
Intellectual Property Infringement Avoidance Clearance And Assessment
It is critical to assure that improvement or new designs developed in the course of the
project do not infringe on the Intellectual Property of others, external to GE. GE has a
clear policy on infringement avoidance. “Intellectual Property Infringement Avoidance
Process”. The team must conduct the necessary reviews to assure no infringement
(consult the IP Business Champion for guidance). In general, if the answer is “yes” to
any of the questions below, consult with the IP Business Champion and IP Designee
for your business to obtain clearance before proceeding.
1) Is the team aware of any non-GE businesses who use substantially the same process or
sell substantially the same process?
2) Will the project result in a product or process that is significantly different than a current GE
product or process, and will it be commercially significant?
3) Have or will any services or products been obtained from an outside party I connection with
this project, and is the project result commercially significant?
4) Will the product or process be exposed to non-GE parties, and be commercially
significant?
5) Will the product or process be exposed to the public, and be commercially significant?
If the project obtained clearance from the Infringement Avoidance screen, then
consider whether there is Intellectual Property to Capture and Protect, via Patent,
Trade Secret, Copyright or Trademark. If the project results in a process or product
that is new, non-obvious and useful and will be of significant commercial value. It may
be patentable. Consult the IP Business Champion and IP Designee for your business
and submit the new process or product for evaluation through the IDEAS website.
DMAIC GB U TX PG V 4.2.0
D M A I C
DMAIC GB V TX PG V 4.2.0
D M A I C
DMAIC GB V TX PG V 4.2.0
D M A I C
DMAIC GB V TX PG V 4.2.0
D M A I C
DMAIC GB V TX PG V 4.2.0
D M A I C
5
Process States
1. Chaos
Process out-of-control, producing non-conforming product
Even the level of nonconformance is unstable
Assignable causes dominate the output
Fixes don’t work for very long
DMAIC GB V TX PG V 4.2.0
D M A I C
6
Process States (continued)
DMAIC GB V TX PG V 4.2.0
D M A I C
7
Process States (continued)
DMAIC GB V TX PG V 4.2.0
D M A I C
DMAIC GB V TX PG V 4.2.0
D M A I C
40
Upper Control Limit*
(UCL)
30
Measurement
20 Average
10
0 10 20
DMAIC GB V TX PG V 4.2.0
D M A I C
Zone B
Zone C
Average
Zone C
Zone B
Zone A
Lower Control Limit
(LCL)
Source: Memory Jogger Plus, ©1994 GOAL/QPC
DMAIC GB V TX PG V 4.2.0
D M A I C
Minitab Rules 11
+3σ
+2σ A
B
+1σ
C
C
-1σ
B
-2σ
A
-3σ
DMAIC GB V TX PG V 4.2.0
D M A I C
Control limits are not the same as tolerances. While the data collected for the control charts
Control limits are what you have demonstrates:
Tolerances are what you want
Mean = 40 days
Example:
UCL= 46 days
DMAIC GB V TX PG V 4.2.0
D M A I C
PROCESS A PROCESS B
Lower Upper Lower Upper
Spec. Limit Spec. Limit Spec. Limit Spec. Limit
DMAIC GB V TX PG V 4.2.0
D M A I C
Input Output
Variables (Y’s)
(X’s)
DMAIC GB V TX PG V 4.2.0
D M A I C
Establish Develop
Ensure Data Collect Data
Data Operational
Consistency & Monitor
Collection Definitions
& Stability Consistency
Goals & Procedures
DMAIC GB V TX PG V 4.2.0
D M A I C
DMAIC GB V TX PG V 4.2.0
D M A I C
VARIABLE CHART
ATTRIBUTE CHART
DMAIC GB V TX PG V 4.2.0
D M A I C
Variable or
Attribute Data?*
Variable
Variable Attribute
Attribute
Defects or Defects or
Defective Defective
Individuals
Individuals &
& Xbar
Xbar &
&
Moving
Moving Range
Range uu
Range
Range pp cc np
np
There are many different types of control charts. The If the data being analyzed is discrete (attribute), you
chart pictured here is a helpful guide to decide which type would need to determine:
of chart to use and when. The primary determinant of If you have a constant lot (sample) size
which type of control chart to use is the type of data or if the lot (sample) size is not constant
being analyzed. If you are counting defects, you may have
multiple defects per unit. If you are counting
If the data being analyzed is continuous (variable), you defectives, the unit either is or is not defective.
would use either: In this module we will study the different types of control
The “Individuals & Moving Range” (I-MR): use if charts and when to use each one.
Rational Subgrouping is not possible (you want to see
each transaction on the control chart) What you want the Control Chart to tell you is:
The Xbar & Range (X - R) use if Rational Subgrouping
is possible. Typically subgroup size will be between 3 - 5. Is my level of non-conformance consistent?
Use (X - S) if subgroup size is 7 or greater Am I controlling my significant X’s?
TM
* Refer to Memory Jogger for further details.
DMAIC GB V TX PG V 4.2.0
D M A I C
DMAIC GB V TX PG V 4.2.0
D M A I C
DMAIC GB V TX PG V 4.2.0
D M A I C
21
DMAIC GB V TX PG V 4.2.0
D M A I C
Variable or
Attribute Data?*
Variable
Variable Attribute
Attribute
Defects or Defects or
Defective Defective
Individuals
Individuals &
& Xbar
Xbar &
&
Moving
Moving Range
Range uu
Range
Range pp cc np
np
DMAIC GB V TX PG V 4.2.0
D M A I C
DMAIC GB V TX PG V 4.2.0
D M A I C
24
DMAIC GB V TX PG V 4.2.0
D M A I C
Variable or
Attribute Data?*
Variable
Variable Attribute
Attribute
Defects or Defects or
Defective Defective
Individuals
Individuals &
& Xbar
Xbar &
&
Moving
Moving Range
Range uu
Range
Range pp cc np
np
DMAIC GB V TX PG V 4.2.0
D M A I C
DMAIC GB V TX PG V 4.2.0
D M A I C
Individuals
4 Moving Range
3
2
1
0
DMAIC GB V TX PG V 4.2.0
D M A I C
DMAIC GB V TX PG V 4.2.0
D M A I C
DMAIC GB V TX PG V 4.2.0
D M A I C
Instructions The quality team has 25 weeks of historical data, contained in the
file listed above. You have been asked to examine their data and
believe an Xbar and R Chart, or an I/MR Chart would be the
appropriate tool to use.
Time 10 minutes
DMAIC GB V TX PG V 4.2.0
D M A I C
31
Minitab Output
Response Time
I and MR Chart for Response
600 1
Individual Value
500
400
1 1 1 11 1
300 1 1
1 11 1 1 1 11 1 11 1
1 1 1
200 5 5 5
55 3.0SL=205.2
2 44 7 4
100 7 44
X=84.00
4 2 4 77 7
0 6 2
-3.0SL=-37.19
-100
Subgroup 0 500 1000
500 1
1
Moving Range
400
300
1 1 11
1 1 1 1
1
200 1 11 1 11 1 1 11 1
2 1 1 1 1 111 1
3.0SL=148.9
100
R=45.57
0 2 2 2 2
22 -3.0SL=0.00E+00
22
22 2 222 2
DMAIC GB V TX PG V 4.2.0
D M A I C
DMAIC GB V TX PG V 4.2.0
D M A I C
33
DMAIC GB V TX PG V 4.2.0
D M A I C
Attribute Data
Variable or
Attribute Data?*
Variable
Variable Attribute
Attribute
Defects or Defects or
Defective Defective
Individuals
Individuals &
& Xbar
Xbar &
&
Moving
Moving Range
Range uu
Range
Range pp cc np
np
DMAIC GB V TX PG V 4.2.0
D M A I C
Important Definitions 35
A Defect
A single characteristic that does not meet requirements
A Defective
A unit that contains one or more defects
DMAIC GB V TX PG V 4.2.0
D M A I C
Constant Variable
Sample Size Sample Size
For example:
Is it pass/fail for the part or process?
Or is it several sub-components or sub-processes
passed/failed for the part or process)?
DMAIC GB V TX PG V 4.2.0
D M A I C
37
DMAIC GB V TX PG V 4.2.0
D M A I C
P-Chart
Variable or
Attribute Data?*
Variable
Variable Attribute
Attribute
Defects or Defects or
Defective Defective
Individuals
Individuals &
& Xbar
Xbar &
&
Moving
Moving Range
Range uu
Range
Range pp cc np
np
DMAIC GB V TX PG V 4.2.0
D M A I C
P-Chart 39
DMAIC GB V TX PG V 4.2.0
D M A I C
DMAIC GB V TX PG V 4.2.0
D M A I C
The loan department has made a change to the loan application process. They
think the change will help weed out the bad loan risks before the applications
are actually submitted. If they are right, the approval rate for submitted
applications should increase. They want to check to see if this change in the
process has had the desired effect.
1. Is there evidence that the new process made a difference?
2. What is the number of opportunities for defect in this process measure?
3. Is this an input, process, or output variable? Why?
3.0SL = 97.71
New Process
•
P-Chart Of Loan Approval Rate
• 93.79
% Applications Approved
Implemented
May 1 To November 5 •
• •
90
• • • •• •
89.86
•• •• • •
• • • P = 85.94
•
• 82.02
80 •
• •
• • 78.09
-3.0SL = 74.18
70
0 10 20 30
Week (Time Order)
Note: “SL” is the standard Minitab notation for “Sigma Limits” which are the same as Control Limits.
DMAIC GB V TX PG V 4.2.0
D M A I C
Reference Materials–
Attribute Control Charts
C-Chart
Proceed to page 61
DMAIC GB V TX PG V 4.2.0
D M A I C
Attribute Data
Variable or
Attribute Data?*
Variable
Variable Attribute
Attribute
Defects or Defects or
Defective Defective
Individuals
Individuals &
& Xbar
Xbar &
&
Moving
Moving Range
Range uu
Range
Range pp cc np
np
DMAIC GB V TX PG V 4.2.0
D M A I C
C-Chart 44
DMAIC GB V TX PG V 4.2.0
D M A I C
DMAIC GB V TX PG V 4.2.0
D M A I C
An applications group of a large insurer recently made some changes to the design of
their application. The old applications had 10 fields for the customer to fill out; the new
application has 5 fields for the customer to fill out. They are now monitoring the incoming
applications to ensure the new design has had an impact on completeness. A random
sample of 100 applications are reviewed each day and the total number of incomplete
fields are plotted.
The chart for this data is on the following page.
1. Why is the control chart a C-Chart?
2. Is the number of errors in a sample stable over the 20–day sample?
3. What questions does the chart raise?
Extra Credit!! Calculate baseline sigma for the first 20 days of service for the new
application.
1 5 6 9 11 6 16 7
2 5 7 9 12 6 17 7
3 11 8 8 13 7 18 8
4 5 9 8 14 5 19 12
5 8 10 7 15 6 20 13
DMAIC GB V TX PG V 4.2.0
D M A I C
3.0SL=15.87
15
Number of Incomplete Fields
10
C=7.600
0 -3.0SL=0.00E+00
0 10 20
Day of Sample
DMAIC GB V TX PG V 4.2.0
D M A I C
48
DMAIC GB V TX PG V 4.2.0
D M A I C
Attribute Data
Variable or
Attribute Data?*
Variable
Variable Attribute
Attribute
Defects or Defects or
Defective Defective
Individuals
Individuals &
& Xbar
Xbar &
&
Moving
Moving Range
Range uu
Range
Range pp cc np
np
DMAIC GB V TX PG V 4.2.0
D M A I C
U-Chart 50
Chart for defects per unit, with variable lot (subgroup) size
Same logic as C-Chart, except variable lot (subgroup) size (n)
DMAIC GB V TX PG V 4.2.0
D M A I C
Note: Do not turn the page until you have completed the activity.
DMAIC GB V TX PG V 4.2.0
D M A I C
DMAIC GB V TX PG V 4.2.0
D M A I C
3
Sample Count
3.0SL=2.114
2
U=1.764
-3.0SL=1.415
0 10 20 30
Sample Number
DMAIC GB V TX PG V 4.2.0
D M A I C
54
DMAIC GB V TX PG V 4.2.0
D M A I C
NP-Chart
Variable or
Attribute Data?*
Variable
Variable Attribute
Attribute
Defects or Defects or
Defective Defective
Individuals
Individuals &
& Xbar
Xbar &
&
Moving
Moving Range
Range uu
Range
Range pp cc np
np
DMAIC GB V TX PG V 4.2.0
D M A I C
NP-Chart 56
DMAIC GB V TX PG V 4.2.0
D M A I C
DMAIC GB V TX PG V 4.2.0
D M A I C
An applications group of a large insurer recently made some changes to the design of
their application. They are now monitoring the incoming applications to ensure the new
design has had an impact on completeness. 150 applications are chosen randomly and
reviewed each day. The number that are incomplete are plotted each day
1 12 6 14 11 2 16 11
2 22 7 16 12 15 17 17
3 14 8 19 13 8 18 12
4 15 9 11 14 12 19 11
5 12 10 1 15 12 20 21
DMAIC GB V TX PG V 4.2.0
D M A I C
25
3.0SL=23.13
20
# of Incomplete Applications
15
NP=12.85
10
Day of Sample
5
-3.0SL=2.567
0 10 20
DMAIC GB V TX PG V 4.2.0
D M A I C
Rule of Thumb:
– Select a subgroup size that will provide an average defect/defective of
approximately
C, U, NP > 5.0
– For NP-charts, to select the appropriate sample size such that 95% of the
subgroups will have at least one defective, use the relationship
n= 3
p
DMAIC GB V TX PG V 4.2.0
D M A I C
Variable or
Attribute Data?*
Variable
Variable Attribute
Attribute
Rational Yes
No
Subgrouping
No Constant Yes
Possible? Lot Size?
Defects or Defects or
Defective Defective
Individuals
Individuals&& Xbar
Xbar&&
Moving
Moving Range
Range
Range
Range uu pp cc np
np
© GE Capital, Inc., 2000
DMAIC GB V TX PG V 4.2.0
D M A I C
DMAIC GB V TX PG V 4.2.0
D M A I C
Summary–Control Charts 63
DMAIC GB V TX PG V 4.2.0
D M A I C
64
DMAIC GB V TX PG V 4.2.0
D M A I C
DMAIC GB V TX PG V 4.2.0
D M A I C
Continuous Data
Control limits are calculated using control chart factors and the
Range-Bar (an estimate of short-term sigma)
Control chart factors were invented by Shewart in the 1920’s to avoid
long-hand calculation
Control chart factors are shown in the table according to the sample
size for each subgroup
Individual control charts are considered to have a sample of size 2,
the number of data points make up the moving range
Discrete Data
Control limits are calculated using a formula that estimates sigma
without
the necessity to transform the data
The normal estimate for sigma (under radical) is then multiplied by
three
DMAIC GB V TX PG V 4.2.0
D M A I C
Median ˜ 1 + X˜ 2 + … X˜ k)
(X ˜ + A2R
UCLx = X
and Range X
˜= k LCLx = X
˜ - A2R
< 10, but
usually
3 to 5 (R1 + R2 +…Rk) UCLR = D4R
X
˜ and R R= k LCLR = D3R
˜
k = # of subgroups, X = median value within each subgroup
ΣXi
*X = n
DMAIC GB V TX PG V 4.2.0
D M A I C
DMAIC GB V TX PG V 4.2.0
D M A I C
Fraction
p(1 - p)
Defective For each subgroup: *UCLp = p + 3
Variable, n
p = np/n
usually ≥ 50 For all subgroups:
p-Chart p = ∑np/ ∑n *LCLp = p - 3 p(1 - p)
n
UCLnp = np + 3
Number For each subgroup: np(1 - p)
Defective Constant, usually np = # defectives
≥ 50 For all subgroups:
np = ∑np/k LCLnp = np - 3
np-Chart np(1 - p)
DMAIC GB V TX PG V 4.2.0
D M A I C
DMAIC GB V TX PG V 4.2.0