treqna base manual ed
TRANSCRIPT
Treqna Base Manual
First Edition 1 © All Rights Reserved TreQna 2005
treqna.com
Treqna Base Manual
First Edition 2 © All Rights Reserved TreQna 2005
DEFINEStep D1 – Map Project Page 4 Step D2 – Approve Project Page 36
MEASUREStep M1 – CTQ Characteristics & Standards Page 45Step M2 – Measurement System Analysis Page 82Step M3 – Data Collection Plan Page 116
ANALYZEStep A1 – Baseline Process Page 145Step A2 – Performance Objective Page 191Step A3 – Identify drivers of Variation Page 197
IMPROVEImprove Step I1 – Screen for Vital Xs Page 293Improve Step I2 – Study Interaction between Xs Page 298Improve Step I3 – Define Improved Process Page 303
CONTROLControl Step C1 – MSA on Xs Page 317Control Step C2 – Improved Process Capability Page 319Control Step C3 – Establish Control Plan Page 321
C ontents
Treqna Base Manual
First Edition 3 © All Rights Reserved TreQna 2005
DEFINE
MEASURE
ANALYZE
IMPROVE
CONTROL
Map project
CTQ Characteristics and standards
Measurement System Analysis
Baseline process
Screen for vital X’s
Defined Improve Process
MSA on X’s
Approve project
Data collection
Performance objective
Identify drivers of variation
Study interaction between X’s
Improved Process capability
Establish control plan
Treqna Base Manual
First Edition 4 © All Rights Reserved TreQna 2005
Define Phase
DEFINE MEASURE ANALYZE IMPROVE CONTROL
Tools
Deliverables
Step
Step D.2Step D.1
CBAFMEA
Survey,Focus Group,Interview,Charter,COPIS
Project feasibilityProject CharterCOPIS
Approve projectMap project
Treqna Base Manual
First Edition 5 © All Rights Reserved TreQna 2005
Topics
About Projects
Customer, CTQs and VOC
COPIS
Project Charter
Treqna Base Manual
First Edition 6 © All Rights Reserved TreQna 2005
Characteristics of a good project
Should have clearly defined scope
Should have significant impact on customer
Should be in line with business objectives
Should have synergy with other projects (must not conflict)
Should be focused on key CTQ
Should generate bottom-line hard dollar benefits, Cost Avoidance or Productivity
increase
Treqna Base Manual
First Edition 7 © All Rights Reserved TreQna 2005
Sources for project ideas ?
Project topics usually surface as issues in a product or process. The project ideas
may also come from:
Customer dashboards
Customer Surveys
Other related projects
Internal issues
The tools which can be used to identify project ideas:
Brainstorming
QFD
Root cause analysis
Treqna Base Manual
First Edition 8 © All Rights Reserved TreQna 2005
Topics
About Projects
Customer, CTQs and VoC
COPIS
Project Charter
Treqna Base Manual
First Edition 9 © All Rights Reserved TreQna 2005
Who is the customer? Internal Customer or External Customer ?
Identifying the Customer
CustomerOutput Input
Supplier
Process
Customer receives the output of a process
Internal customer: E.g. Marketing person for manufacturing company is an
internal customer.
External customer: E.g. Buyer of the product or services
Treqna Base Manual
First Edition 10 © All Rights Reserved TreQna 2005
Critical to Quality (CTQ)
Customer to CTQ
External Customer Internal Customer
Customer driven CTQ’s Process driven CTQ’s
Identify Customer Capture customer requirement Derive and Define CTQ’s
CTQ’s from any process are translated from customer requirements
What is a CTQ?A CTQ (Critical to Quality) also known as a KPI (Key Process Indicator) is a metric that measure
some aspect of a product or process which is critical to the customer. The customer defines acceptable levels for CTQs using specification limits.
Treqna Base Manual
First Edition 11 © All Rights Reserved TreQna 2005
Voice of the Customer (VoC)
The "voice of the customer" is a process which is used to capture the
requirements/feedback from the customer (internal or external). It helps us
understand the customers expectations.
The voice of the customer can be captured in a variety of ways:
Direct discussion or interviews
Surveys
Focus groups
Complaints
Others ways like, Customer specifications, Observation, Field reports etc.
Treqna Base Manual
First Edition 12 © All Rights Reserved TreQna 2005
VOC Tools Evaluation
SurveyAdvantages:
Low administration cost
Relatively faster results in case of telephone
survey
Excellent approach to reach larger end customer
base
Least trained resources required to execute
Disadvantages:
Low response rate with Mail surveys
Unclear understanding by customer
Incomplete responses to mail surveys
Potential risk of biased response due to interviewer influence
Focus Groups
Advantages:
Help gather qualitative and in depth information
Excellent for understanding and getting CTQ
definitions
Not prone to understanding and interpretation
gaps
Disadvantages:
Bias due to limited participants
Limited data due to qualitative focus
Can have a lot of anecdotal information
Treqna Base Manual
First Edition 13 © All Rights Reserved TreQna 2005
Interviews
Advantages:
Complex questions can be asked and a wide
range of information gathered
Both qualitative and quantitative data can be
collected
Better communication and interpretation
Disadvantages:
Higher cost alternate
Time consuming
Requires trained, experienced resources to execute
Customer Complaints
Advantages:
Specific feedback
Provides opportunity to respond appropriately to
dissatisfied customer
Disadvantages:
Sample size issue
Situation specific, might be anecdotal
Prone to small sample bias
VOC Tools Evaluation
Treqna Base Manual
First Edition 14 © All Rights Reserved TreQna 2005
Survey Development
Objectives
Why is the survey needed?
Who will use the survey results?
What specific information is required with the help of this survey?
Who needs to be surveyed?
Logistics
How much time is available to administer the survey?
Who will conduct the survey?
What is the medium to conduct the survey?
What are the financial resources available to conduct the survey?
Design
Draft questions and validate against objectives.
Devise measurement scale.
Design report out structure
Administration
Determine sample size and sampling strategy
Pilot survey on selected group (ensure the methodology)
Designing survey question: Survey results will be meaningless unless you ask the right questions. The question must contain enough specifics so the respondent can give a meaningful answer.
Types of questions:•open questions: Open questions allow the customer to respond to a question in his or her own words.•closed questions. Closed questions offer the customer a choice of specific responses from which to select. Multiple choice, rating scale and yes/no questions are examples of closed questions.
Bias from Design of Questions: The questions you ask your customers must be properly worded in order to achieve good end results. Avoid the following:•Leading questions-They inject interviewer bias.•Compound questions-They may generate a partial or no response.•Judging questions-They can lead to guarded or partial responses.Ambiguous or vague questions-They produce meaningless responses•Acronyms and jargon-These may be unknown to the respondent.Double negatives-They may create misunderstanding.•Long surveys-They discourage respondent participation.
Other sources of bias in survey:Question sequenceSample BiasNo-response biasUnclear questions
Interviewer / facilitator induced bias
Treqna Base Manual
First Edition 15 © All Rights Reserved TreQna 2005
Understanding Customer Needs
Real vs. Stated Needs: Use questions like “Why do you want…?” until you understand the need
Perceived Needs: The same product or service may be perceived differently by customers
Intended vs. Actual Usage: Be careful of how customers say they will use a product/service versus how they actually use it.
Internal Customers: Internal customers have needs associated with job security, prestige, etc.
Effectiveness vs. Efficiency Needs: The external customer will generally express effectiveness needs–those relating to the value they receive from the product or service. The internal customer, however, will generally express efficiency needs, those relating to the amount of resources allocated or consumed in meeting customer needs.
Treqna Base Manual
First Edition 16 © All Rights Reserved TreQna 2005
Topics
About Projects
Customer, CTQs and VoC
COPIS
Project Charter
Treqna Base Manual
First Edition 17 © All Rights Reserved TreQna 2005
COPIS
Puts process and customer in perspective
Depicts process flow
Highlights process boundaries and interdependencies
COPIS maps customer and process interaction by defining customer requirements and steps taken to deliver the desired output
CCustomer
OOutput
PProcess
IInput
SSupplier
Thought Process
Process Flow
COPIS and SIPOC are used interchangeably.
Treqna Base Manual
First Edition 18 © All Rights Reserved TreQna 2005
COPIS - Components
Customer: Recipient of the process output.
Output: Anything produced for the customer (internal or external). Outcome of the process.
Process: Group of activities required to transform inputs in to customer desired output
Input: Material or knowledge required to produce the desired output
Supplier: Source that supplies the input
Metric: A measure of compliance with customer expectation or established standard
CCustomer
OOutput
PProcess
IInput
SSupplier
Treqna Base Manual
First Edition 19 © All Rights Reserved TreQna 2005
Step3: Identify process steps
Step2: Identify customer requirements
Step1: Identify Customers, Outputs, Inputs, Suppliers and Process Name
COPIS - Components
CCustomer
OOutput
PProcess
IInput
SSupplier
Treqna Base Manual
First Edition 20 © All Rights Reserved TreQna 2005
Topics
About Projects
Customer, CTQs and VoC
COPIS
Project Charter
Treqna Base Manual
First Edition 21 © All Rights Reserved TreQna 2005
Project Charter
Project charter is a written roadmap that defines key questions or issues to be addressed by the project. It defines the project's purpose and its intended outcomes.
The critical elements of a project charter are:
Business Case
Problem and Goal Statement
Project Scope and Boundaries
Project Team
Project Timelines
Communication Plan
Treqna Base Manual
First Edition 22 © All Rights Reserved TreQna 2005
Critical Elements Of A Charter
Business Case: The need to do the project
Problem and Goal Statement: Project problem and improvement goal in distinct and measurable terms
Project Team: Who will be involved in the project and what role will he or she play?
Project Scope: Area of influence and resources for the project
Project Timelines: Project plan and key milestones
Communication Plan: Communication strategy through the project
Treqna Base Manual
First Edition 23 © All Rights Reserved TreQna 2005
The Business Case
Project Business Case has three critical elements:
It tells us why is the project worth doing
It tells us why is it important to do it now
It tells us the consequences of both, NOT doing the project and delaying it
Business case also talks about:
Brief introduction to the process/business where project is undertaken
How does the project help in achieving business goals.
Treqna Base Manual
First Edition 24 © All Rights Reserved TreQna 2005
The Problem Statement
The problem statement:
Represents the problem in measurable (quantifiable) terms. For instance, yield, defect rate, ppm, sigma etc.
Measures the problem against a target (if one exists)
Questions that a problem statement must answer
What is the problem?
Where is the problem?
What is the magnitude of the problem?
Over what period has the problem been recorded?
What is the impact of the problem?
Treqna Base Manual
First Edition 25 © All Rights Reserved TreQna 2005
The Problem Statement Examples
Problem Statement: “Our customers are not happy with the delay in monthly credit card account statement mailing”
Poorly defined problem statement – generic in nature. It does not highlight the magnitude or the impact of the issue at hand.
Problem Statement: “In the last 3 months 26% of the monthly credit card account statement were delayed by 10 or more days, due to which business had to waive off $280,000 in late fee charges.”
Well defined problem statement – specific and direct. Clearly brings out the magnitude of the problem and its consequences to the business.
Treqna Base Manual
First Edition 26 © All Rights Reserved TreQna 2005
Things To Watch Out For
Potential Pitfalls:
Anecdotal: The problem statement is based on assumption (guess) or based on insufficient data?
Leading: Problem statement giving hint to a prejudged root cause? (implied solution, this might happen in business case also)
Immeasurable: Challenges in collecting data or data can not be collected (Insight into a problem is limited by measurability of a problem)
Scope: Problem statement too narrowly or broadly defined
Treqna Base Manual
First Edition 27 © All Rights Reserved TreQna 2005
The Goal Statement
A Goal Statement defines the improvement objective that a project team sets out to
achieve. It must
Be time bound
NOT suggest or assume a solution
Have a specific, measurable target
The goal statement would usually start with the verb, e.g. reduce, increase, eliminate etc.
Treqna Base Manual
First Edition 28 © All Rights Reserved TreQna 2005
Goal Statement: “We intend to reduce the no. of delayed credit card statements.”
Poorly defined goal statement – generic in nature. The target and time frame to achieve
it have not been defined.
Goal Statement: “Reduce the percentage of delayed monthly credit card account
statement from 26% to 2% by 26th Nov 2004.”
Well defined goal statement – is specific, measurable and time bound.
The Goal Statement Example
Treqna Base Manual
First Edition 29 © All Rights Reserved TreQna 2005
Project Scope and Boundaries
Project Scope and Boundaries provide clear demarcation between what the project team
can and cannot influence.
While defining the scope of a project, you must identify
Processes under study
Resources available to the project team
Constraints, if any, on the project team
You must also define boundaries (start and stop points) for each process under study to
avoid ambiguity
Treqna Base Manual
First Edition 30 © All Rights Reserved TreQna 2005
Steps in Scoping a Project
Identify customer expectations – Who is the customer? What is the output? What does the customer expect from the output?
Translate customer expectations to process CTQs – which process CTQs best represent customer want
Map process – detail the process(es) under study to identify areas of specific focus
Identify resources – what are the resources available to the project team
Understand constraints – what are the other CTQs or parameter that can be influenced by the process but need to remain as is
Treqna Base Manual
First Edition 31 © All Rights Reserved TreQna 2005
Timelines & Project Plans
Establishing critical milestones helps ensure focus through the life of the project, timelines should be aggressive, yet realistic
Project Plan: An activity by activity plan to help meet project timelines.
Identifies order of execution for critical tasks
Identifies interdependencies, if any, between tasks
Gives an estimate of time requirement
Helps identify potential roadblocks and plan contingency measures
Treqna Base Manual
First Edition 32 © All Rights Reserved TreQna 2005
Communication Plan
Communication plan:
To prevent sudden surprises: Keep all stakeholders informed about project status
To ensure that timely communication of vital information to those who need it
To help efficient management of information logistics
Good communication plan would address the following:
Who would communicate? (Project Owner)
What to communicate? (Project Status, schedule, plan, challenges, successes, failures)
When and how frequently? (Tollgates, weekly, monthly…)
Reason for communication? (importance)
Recipients of communication? (Project Manager, Process Owner, Champion/Sponsor, Process Team, Cross-functional Teams, Process Teams etc..)
Other things to be covered in the plan include the choice of media, information residence,
information source and comments regarding details of information
Treqna Base Manual
First Edition 33 © All Rights Reserved TreQna 2005
Team Roles
Roles which should be defined in the Project Charter:
Project Manager – GB/BB or MBB
Mentor - BB or MBB
Project Champion
Project Sponsor
Project team members
While defining roles:
The project champion and the sponsor might be the same person
Team members are usually from the process or related to the product for which the project is undertaken. Involvement of support staff is also common.
It is not unusual to have team members specific to a particular activity or phase
Green belt - An employee of an organization who has been trained on Six Sigma and executes projects as a part of his/her full time job. Green belts are usually supervised or mentored by a Black belt.Black belt – An employee of an organization who has been trained on Six Sigma and executes/mentors projects. Black belts are Six Sigma specialists and mentor Green Belts, Black belts are usually mentored by Master Black BeltsMaster Black Belt – Master Black Belts are Six Sigma Quality experts that are responsible for the strategic deployment of Six Sigma resources within an organization. Master Black Belt main responsibilities include training and mentoring BBs and GBs; selection, execution and support of Six Sigma projects.Black belts usually report into a Master Black Belt.Champion – A member from the leadership team who reviews the progress of projects that he/she champions. The champion is also responsible for supporting the project and help it overcome any challengesSponsor – Usually the process owner of the process on which a project is being done, he/she provides the necessary resources for execution of the project
Treqna Base Manual
First Edition 34 © All Rights Reserved TreQna 2005
Potential Pitfalls
Common mistakes made while selecting a project team
Time commitment from team members is not clearly defined
Incorrect mix – lack of influence (hierarchy), knowledge (subject matter expertise) or competence (statistical or technical skills)
Champion’s role is not clearly defined
Treqna Base Manual
First Edition 35 © All Rights Reserved TreQna 2005
Topics covered
About Projects– Characteristics of a good project– Sources for project ideas
Customer, CTQs and VoC– Internal or External Customer– Critical to Quality Metrics– Voice of the Customer – VoC tools
COPIS
Project Charter– Business Case– Problem and Goal Statement– Project Scope and Boundaries– Project Team– Project Timelines– Communication Plan
Treqna Base Manual
First Edition 36 © All Rights Reserved TreQna 2005
Define Phase
DEFINE MEASURE ANALYZE IMPROVE CONTROL
Tools
Deliverables
Step
Step D.2Step D.1
CBAFMEA
Survey, Focus Group, Interview, Charter, COPIS
Project feasibilityProject CharterCOPIS
Approve projectMap project
Treqna Base Manual
First Edition 37 © All Rights Reserved TreQna 2005
Topics
Project Risks
Cost Benefit Analysis
Project Go/No Go decision
Treqna Base Manual
First Edition 38 © All Rights Reserved TreQna 2005
Project Risks
Brainstorm to identify potential risks associated with each phase of the project – risks that restrict its execution or cause it to fail
Assess the likelihood of the risk to occur
Evaluate risks associated with external factors influencing the process or product under study
Identify contingency measures to mitigate identified risks.
Evaluate project feasibility under light of associated risks and contingency plans
Treqna Base Manual
First Edition 39 © All Rights Reserved TreQna 2005
Topics
Project Risks
Cost Benefit Analysis
Project Go/No Go decision
Treqna Base Manual
First Edition 40 © All Rights Reserved TreQna 2005
Cost Benefit Analysis
Identify costs associated with the project
Would new technology or changes to existing technology be required?
Is there some new infrastructure that would be required?
Translate improvement target, if possible, in to quantifiable dollar benefit
Would the raw material costs come down?
Would the no. of people required to do the job reduce?
Understand business budget constraints
Are there any budget constraints?
Are there other projects with equal or high priority?
A Cost Benefits Analysis (CBA) is a calculation which is done in order to evaluate the costs that a project/change may incur and if it will have generate enough benefits to justify the costs.
Treqna Base Manual
First Edition 41 © All Rights Reserved TreQna 2005
Topics
Project Risks
Cost Benefit Analysis
Project Go/No Go decision
Treqna Base Manual
First Edition 42 © All Rights Reserved TreQna 2005
Project Go/ No-go
Evaluate the following before undertaking a project
Potential risks associated with the project, the contingency plans to mitigate those risks and the residual exposure to such risks
The champion’s commitment to mitigate risks and drive contingency measures, if any require his influence
Financial commitment for the completion of the project (allocate budget)
Treqna Base Manual
First Edition 43 © All Rights Reserved TreQna 2005
Topics Covered
Project Risks
– Identify and plan for risks that may delay the project
Cost Benefit Analysis
– Evaluate costs associated with the project, its benefits and its importance from the business stand point
Project Go/No Go decision
– Contract the champion’s commitment to ensure the success of the project
Treqna Base Manual
First Edition 44 © All Rights Reserved TreQna 2005
DEFINE
MEASURE
ANALYZE
IMPROVE
CONTROL
Map project
CTQ Characteristics and standards
Measurement System Analysis
Baseline process
Screen for vital X’s
Defined Improve Process
MSA on X’s
Approve project
Data collection
Performance objective
Identify drivers of variation
Study interaction between X’s
Improved Process capability
Establish control plan
Treqna Base Manual
First Edition 45 © All Rights Reserved TreQna 2005
Measure Phase
Continuous Gage R&R, Short Form Gage,
Test/ Retest Study,Attribute Gage R&R
Identify gage error
Measurement system analysis
Step M.2
Tools
Deliverables
Step
Step M.3Step M.1
Process Map, QFD,FMEA,Pareto
Sampling StrategySegmentation factors
Identify project Y metricEstablish performance standards
Data collectionCTQ characteristics and standards
DEFINE MEASURE ANALYZE IMPROVE CONTROL
Treqna Base Manual
First Edition 46 © All Rights Reserved TreQna 2005
Topics
Selecting CTQ Characteristics
Process mapping
CTQ Drill down tree
QFD
FMEA
Performance Standards
Treqna Base Manual
First Edition 47 © All Rights Reserved TreQna 2005
Process mapping
Process Mapping:
Is a graphical representation of activities, steps, information, resources and their interactions.
Process Map is also called a flowchart.
Process Mapping is a first step in understanding how and why a process behaves the way it does.
Treqna Base Manual
First Edition 48 © All Rights Reserved TreQna 2005
A process is a series of logical steps that transform inputs (raw materials) in to customer defined outputs.
A process may be largely affected by one or more of the following factors:
personnel who operate the processes;
materials which are used as inputs (including information);
machines or equipment being used in the process (in process execution or monitoring / measurement;
methods (including criteria and various documentations used along the process);
work environment
Understanding a process
CustomerOutput Input
Supplier
Process
Treqna Base Manual
First Edition 49 © All Rights Reserved TreQna 2005
------------------- Process Boundary --------------------
Process Controls
Metrics and Measurements
Process Boundary defines the process limit. It helps understand the scope of the process and its constraints.
Process controls help ensure the process is consistent in behavior
Metrics and Measurements are the means to measure conformance with requirements placed by customers on outputs and processes on inputs.
Components of a process
Treqna Base Manual
First Edition 50 © All Rights Reserved TreQna 2005
Why map a process?
Mapping a process has many uses:
Compare actual vs. assumed (designed) flow
Identify redundant, duplicate or non value add steps
Identify data collection points in the process
Study interaction and interdependencies between different departments (highlight hand off points)
Treqna Base Manual
First Edition 51 © All Rights Reserved TreQna 2005
How to map a process?
To map a process:
Understand the process boundaries and scope
Observe the process and identify its steps
Arrange the process steps in the order that they are followed
Identify the outputs, the customers, and their key requirements
Identify the inputs, the suppliers, and the process’s key requirements
Treqna Base Manual
First Edition 52 © All Rights Reserved TreQna 2005
Frequently used symbols
Process step Decision box
Process Start / End
Off page connector
Document Multi-document
Process steps connector
Input to a process
Treqna Base Manual
First Edition 53 © All Rights Reserved TreQna 2005
Types of process maps
Alternate path process map: used for complex and/or large process. Alternate paths replace decision boxes in the map. The percentage depiction of alternate paths makes this map more informative. 80%
20%
Human Resource
Operations
Finance
Three frequently used process maps are:Basic Process map: used when the process is small and simple
Cross functional process map: used when the process has many handoffs between different departments. It is also known as deployment map.
Treqna Base Manual
First Edition 54 © All Rights Reserved TreQna 2005
Topics
Selecting CTQ Characteristics
Process mapping
CTQ Drill down tree
QFD
FMEA
Performance Standards
Treqna Base Manual
First Edition 55 © All Rights Reserved TreQna 2005
CTQ Drill down Tree
E.g. Ticket Counter at Movie theatre. From customer VOC to project metric
Better Billing Process
Good Customer Service
Timely Process
Curt and Polite Accurate BillingKnowledgeable Less queue time
Big Y
A CTQ drill down tree tool assists in choosing a project metric. It shows linkage between the project metric and the company goals. Once established it can be used in projects to help finalize relevant project metrics.
A ‘Y’ is a dependent output variable. The higher the ‘Y’ in a hierarchy of ‘Y’ the bigger it is considered to be. A ‘X’ is an independent variable that effects the ‘Y’.
Treqna Base Manual
First Edition 56 © All Rights Reserved TreQna 2005
Topics
Selecting CTQ Characteristics
Process mapping
CTQ Drill down tree
QFD
FMEA
Performance Standards
Treqna Base Manual
First Edition 57 © All Rights Reserved TreQna 2005
Quality Function Deployment
Quality Function Deployment (QFD) is a structured approach to defining customer needs or requirements and translating them into specific plans to produce products to meet those needs.
It helps to understand:
What customer wants when the customer requirement is not explicit or is indirect.
How important it is to the customer
What the customer does not want
Treqna Base Manual
First Edition 58 © All Rights Reserved TreQna 2005
QFD -Objectives
Understand Quality Function Deployment (QFD) as a tool.
Describe the steps of QFD.
Analyze output from QFD.
Treqna Base Manual
First Edition 59 © All Rights Reserved TreQna 2005
Definition of QFD
QFD is a methodology and tool to identify and translate customer needs and wants into measurable features and requirements – converting the “what’s” in to “how’s”
QFD links the needs of the customer (end user) with design, development, manufacturing, and service functions.
Used to identify Critical to Quality Characteristics (CTQ’s)
QFD is also known as the “House of Quality”
Treqna Base Manual
First Edition 60 © All Rights Reserved TreQna 2005
18
9
Ingredient cost
48
3
9
Temperature (cold)
423936
93Good service
32Should be reasonably priced
94Should taste good
934Quenches thirst
Quantity
Time to prepare
Ingredient mix
Customer Needs
Customer’s importance rating
Customer’s Expectations
Measures that capture customers requirement
Impact a CTQ has on the customer’s expectations
Overall CTQ impact on customer needs
QFD Matrix
××
×Correlation between measures
Strong Positive Strong Negative Positive Negative×× ×
Elements of QFD matrix:Rows: The rows of the QFD matrix represent the customer needs (VOC) Columns: The columns of the QFD matrix represent the CTQs that measure the customer’s need.The cross section matrix cells are filled with values 1,3 and 9 depending on the correlation between the customer needs and the CTQs. 1 represents low correlation, 3 medium correlation, and 9 high correlation. The cell is left blank where no correlation exists.The column next to needs contains the relative importance given to needs by customer. The roof of the house is used to depict correlation between various measuresWhen the matrix is complete, to prioritize the CTQs or requirement:
Multiply the strength of each relationship (1 for weak, 3 for moderate and 9 for strong) by the priority number (1 to 5) for each corresponding customer expectation. Add the results and enter the sum for each requirement at the bottom of the matrix. This is the output of the QFD exercise.
It is not necessary to complete every cell of the QFD matrix. Typically, a QFD matrix would have 30-50% of the cells filled.Caution: An empty row indicates that the need is not captured by any measure. An empty column indicates that the measure does not fulfill any need and is redundant.
Treqna Base Manual
First Edition 61 © All Rights Reserved TreQna 2005
Analyzing the QFD matrix
Look for blank rows and columns-
blank row means that we don’t have any measure to satisfy customer requirement and
a blank column means that we have a measure that does cater to any customer requirement.
Identify processes/measures which have high impact on customer CTQ’s
Resolve negative correlations between measures/ processes
QFD can be used in iteration to translate customer needs to process CTQs.
For instance - customer needs can be translated to functional requirements using a QFD. A second level of QFD drill down can translate these requirements to product characteristics. A third level QFD can then translate the characteristics to process output measures.
Treqna Base Manual
First Edition 62 © All Rights Reserved TreQna 2005
Topics
Selecting CTQ Characteristics
Process mapping
CTQ Drill down tree
QFD
FMEA
Performance Standards
Treqna Base Manual
First Edition 63 © All Rights Reserved TreQna 2005
FMEA - Objectives
Understanding what an FMEA does
Types of FMEAs
When to use
Learn steps involved in making of FMEA
Treqna Base Manual
First Edition 64 © All Rights Reserved TreQna 2005
Failure Modes and Effects Analysis
FMEA is a structured tool to:
Identify failure modes for a process
Estimate its effect and the risk associated with it
Identify and prioritize probable causes of failure
Evaluate and develop control plan to contain the risk
Treqna Base Manual
First Edition 65 © All Rights Reserved TreQna 2005
Types of FMEAs
System FMEA: Used to analyze potential failure modes in systems during the design or concept stage.
Design FMEA: Used to analyze potential failure modes in products before they are released to production.
Process FMEA: Used to analyze assembly line, manufacturing, transactional or other such processes.
Treqna Base Manual
First Edition 66 © All Rights Reserved TreQna 2005
When is FMEA useful?
Evaluate and control potential risk associated with product, systems in there design phase
Evaluate potential risk when existing systems, processes or products are changed
Proactive tool to evaluate and contain risk associated with any process or change in the process.
FMEA can be used at different phases for different purpose.Measure – CTQ’s identificationImprove – Risk analysis on solutionControl – To develop process control plan
Treqna Base Manual
First Edition 67 © All Rights Reserved TreQna 2005
Preparing a FMEA
Identify process and map the process.
Identify input and output of the process
Understand the effect of inputs (process variables) on output of the process
Rank inputs according to importance
List ways in which inputs can vary and the associated failure modes and effects on outputs
Assign severity, occurrence and detection to each of the items and calculate risk priority number
Prioritize causes with high RPN and develop control plan to minimize risk
FMEA:•Is a team exercise•Living document and should be updated regularly to evaluate risk and make control plans
Treqna Base Manual
First Edition 68 © All Rights Reserved TreQna 2005
FMEA: Definition of terms
Failure Mode
Way in which a process can fail to meet expectation
Any fluctuation which can cause break down in process
Any fluctuation in process variable which can result in defect or non conformance
Potential Cause
A deficiency leading to a Failure Mode
Any source of variation in inputs or process variables
Potential Effect
The impact on customer if the failure mode in not contained or avoided
Customer could be internal or external. The impact could be on the business or employees also (which is internal customer)
Treqna Base Manual
First Edition 69 © All Rights Reserved TreQna 2005
Identifying causes
Causes can be identified using:
Brainstorming
Cause and effect diagram
Ask the question - “What can cause process to fail or what can cause a failure to impact the customer?”
Treqna Base Manual
First Edition 70 © All Rights Reserved TreQna 2005
Rating and RPN
Severity
How severe is the impact of failure mode on customer?
Occurrence
How frequent the cause leading to failure mode can occur?
Detection
What is the possibility of detecting the failure mode if it occurs?
Risk Priority Number
RPN signifies the amount of risk associated with failure mode
RPN = Severance * Occurrence * Detection
Treqna Base Manual
First Edition 71 © All Rights Reserved TreQna 2005
Rating Scale
Many scales are available for rating severity, occurrence and detection. For simplicity (recommended for GBs) one can use the following scale:
Low- e.g. not detectableHigh- e.g. almost assured failure
High- e.g. effecting customer
9
MediumMedium-Medium3
High- e.g.. always detectable
Low – e.g. remote likelihood
Low- e.g. Low or no effect customer
1
DetectionOccurrenceSeverityRating
Treqna Base Manual
First Edition 72 © All Rights Reserved TreQna 2005
Failure Modes and Effects analysis (FMEA)
Date Prepared By:
Process Step Potential failure mode
Potential failure effect
SEV.
Potential Causes
OCC
Current Controls
DET
RPN
RecommendationsSEV.
OCC
DET
RPN
How does it look?
List process steps or product parts
List Failure Modes for each process step
Rate the Severity of the Effect to the customer
Rate how often a particular cause or Failure Mode occurs
List the causes for each Failure ModeList the Effects of each
Failure Mode
Rate the detection based on current controls
Calculate RPN and give recommendations for reducing RPN. Calculate new RPN post implementation of actions
Treqna Base Manual
First Edition 73 © All Rights Reserved TreQna 2005
Topics
Selecting CTQ Characteristics
Process mapping
CTQ Drill down tree
QFD
FMEA
Performance Standards
Treqna Base Manual
First Edition 74 © All Rights Reserved TreQna 2005
Performance Standard
A Performance Standard defines
The customer want
Clearly whether a process is performing well or not
E.g. loan approval within 24 hours, first call resolution
A Performance Standard translates the “Voice of Customer” a measurable metric.
Treqna Base Manual
First Edition 75 © All Rights Reserved TreQna 2005
Elements - CTQ performance standard
Performance Standard
CTQProject Y
CTQUnit of Measure
CTQ Target
Specification Limits
Defect Definition
An Example
Turn Around Time
Days
3 Days
Not greater than5 days
Each new accountApplication whichTakes more than
5 days
CTQ Definition
Time from application to new account opening
Treqna Base Manual
First Edition 76 © All Rights Reserved TreQna 2005
Operational Definition
A precise description that clarifies:
What the CTQ is
And how to measure it.
Why do we need to have operational definitions?
It removes ambiguity in understanding between team members
It clearly defines a “standard” way to measure a CTQ
Ensure that CTQ representation is correct and independent of different times and operators
Treqna Base Manual
First Edition 77 © All Rights Reserved TreQna 2005
Target, Specification limit and Defect
One can understand Target as “a reference point to shoot at”, however Specification Limits (SL) define the acceptable tolerance in deviation from target. Target and SL are defined using VOC.
E.g. VOC - “average time to process a loan should be 3 days but it should not exceed 7 days.”
Would translate into :
Target: 3 days
Upper SL: 7 days (situation where LSL is not required!)
Treqna Base Manual
First Edition 78 © All Rights Reserved TreQna 2005
Target
Specification Limit
Defect
Defect
Anything that results in customer dissatisfaction. Nonconformance to the CTQ.
Target, Specification limit and Defect
Treqna Base Manual
First Edition 79 © All Rights Reserved TreQna 2005
Definitions
Unit (U)The output of a process.
Opportunity (OP)A characteristic of the unit, always at least 1 or more.
Defect (D)Anything that results in a non-conformance with set quality standards
Note: All Opportunities should be independent of each other and must be of significance to the customer
Treqna Base Manual
First Edition 80 © All Rights Reserved TreQna 2005
Defects per UnitTotal Defects per Unit. If there is more than 1 opportunity per unit then the Defects
produced may be more than the number of units. A strict measure of quality.
Total OpportunitiesTotal number of units multiplied by opportunity per unit.
Defects per OpportunityTotal Defects divided by total opportunities.
Defects per Million OpportunitiesDefects per opportunity multiplied by a million.
Formulas
Treqna Base Manual
First Edition 81 © All Rights Reserved TreQna 2005
Topics Covered
Selecting CTQ Characteristics
Process mapping
CTQ Drill down tree
QFD
FMEA
Performance Standards
Treqna Base Manual
First Edition 82 © All Rights Reserved TreQna 2005
Measure Phase
Continuous Gage R&R, Short Form Gage,
Test/ Retest Study,Attribute Gage R&R
Identify gage error
Measurement system analysis
Step M.2
Tools
Deliverables
Step
Step M.3Step M.1
Process Map, QFD,FMEA,Pareto
Sampling StrategySegmentation factors
Identify project Y metricEstablish performance standards
Data collectionCTQ characteristics and standards
DEFINE MEASURE ANALYZE IMPROVE CONTROL
Treqna Base Manual
First Edition 83 © All Rights Reserved TreQna 2005
Topics
Measurement System Analysis - Concepts
Types of MSA
Treqna Base Manual
First Edition 84 © All Rights Reserved TreQna 2005
Understand the terms: Precision , Accuracy, Resolution , Stability, Linearity, Repeatability, Reproducibility,
Evaluated the measurement system and established variation induced by it in process data.
Project team’s consensus on measurement system
And define calibration standards
Measurement System Analysis - Objectives
Treqna Base Manual
First Edition 85 © All Rights Reserved TreQna 2005
Measurement System & Process Variation
Process variation consists of the true variation of process and the variation due to measurement system.
Total Variation = True Process Variation + Measurement System Variation
In this step we try to reduce the variation induced by measurement system. If the measurement system is not studied and calibrated:
Analysis of data may give misleading results
The variation problem may get fixed by fixing measurement system and the project may not be required
One might disturb the process operations without realizing that the cause of variation is measurement system
A Gage study tells us:•Amount of measurement error•Source of measurement error
Once the measurement error is quantified using gage study the next step is to minimize to the level when it becomes acceptable. The gage study is repeated until the level of acceptance is achieved. Until the problem of measurement variation is fixed the data can not be used for analysis and project purpose.
Treqna Base Manual
First Edition 86 © All Rights Reserved TreQna 2005
Process Variation
Process True Variation producing parts different lengths for parts
Input OutputProcess
Variation Induced by measurement system
True Process Variation
Perceived total Variation
Because our knowledge about a problem is limited by the data available, it becomes critical to study variation induced by measurement system - the data is only as good as the measurement system that produces it.
Treqna Base Manual
First Edition 87 © All Rights Reserved TreQna 2005
Sources of Variation
Observed Process Variation
Actual Process Variation
Measurement Variation
Long Term Process Variation
Short Term Process Variation
Sample Variation Gage Variation
Operator Variation
(Repeatability)
Accuracy (Bias) Precision(Reproducibility) Stability Linearity`
Treqna Base Manual
First Edition 88 © All Rights Reserved TreQna 2005
Measurement Variation – Key concepts
Accuracy / Bias – Difference between the standard and observation.
Example - A weighing scale measuring greater than “0” in free state
Repeatability – Variation in measurements of the same unit measured by the same operator using the same equipment.
Example – Observed difference in the length of a pencil measured by the same operator using the same measuring scale
Reproducibility – Variation in measurements of the same unit measured by different operators using the same equipment.
Example – Observed difference in the length of a pencil measured by different operators using the same measuring scale.
Stability – Variation in measurement of the same unit measured by the same operator using the same equipment over time
Example – Observed difference in the length of a pencil measured by different operators using the same measuring scale on different days.
Linearity – Consistency of accuracy across the entire range of measurement system.
Example – Observed difference in the length of a pencil when measured using different ends of the same measuring scale.
Treqna Base Manual
First Edition 89 © All Rights Reserved TreQna 2005
Understanding Precision and Accuracy
Precision: The variation caused due to gage (noise) should be less than the process variation otherwise it will be difficult to detect process variation.
Accuracy: Minimal or no difference between an observed reading average and a standard.
Resolution: Resolution as defined is the smallest unit in which the gage can measure. Ideally the gage should be able to resolve the tolerance into approximately ten levels.
Treqna Base Manual
First Edition 90 © All Rights Reserved TreQna 2005
Understanding Precision and Accuracy
Precise and accurate
==
At target and precise
Accurate but not precise
At target but not precise
Precise but not accurate
Target Precise but away from target
Treqna Base Manual
First Edition 91 © All Rights Reserved TreQna 2005
Measurement System Variability
Method
Material
Measurement
Machine
Environment
People
Fishbone may be used as a tool to identify Sources of variation in the measurement system
Some other terms associated with gage:Repeatability– Variation when one operator repeatedly measures the same unit with the same measuring equipment.
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 a large gap of time.
Linearity–The consistency of the accuracy of gage across the entire range of the measurement system.
Treqna Base Manual
First Edition 92 © All Rights Reserved TreQna 2005
Topics
Measurement System Analysis - Concepts
Types of MSA
Treqna Base Manual
First Edition 93 © All Rights Reserved TreQna 2005
Types of MSA
Attribute Gage: Done when discrete data is involved
Gage ANOVA: with continuous data and the most exhaustive gage study. Requires resource commitment and time. Continuous data can also be run through short form gage and test-retest.
Destructive Gage: Done when the event is destructive in nature and can not be repeated.
Treqna Base Manual
First Edition 94 © All Rights Reserved TreQna 2005
Improving the Measurement System
NOT Precise and NOT Accurate
At target but not precise
Precise but NOT Accurate
At target but not precise
Precise and Accurate
At target but not precise
Treqna Base Manual
First Edition 95 © All Rights Reserved TreQna 2005
Attribute GAGE
Checks gage for:
Repeatability
Reproducibility
Accuracy
Treqna Base Manual
First Edition 96 © All Rights Reserved TreQna 2005
Attribute Gage standards
Ideally an attribute gage is treated as adequate if:
Repeatability: 90% of repeated measures by an operator match
Reproducibility: 90% of repeated measures across all operators match
Accuracy: 90% of all individual measures by across all operators match with the standard
Treqna Base Manual
First Edition 97 © All Rights Reserved TreQna 2005
Attribute Gage standards
Repeatability test is more sensitive if there are greater no. of repeated measurements of the same unit. For instance, it is better to measure 10 units 9 times, than to measure 30 units 3 times.
Reproducibility test is more sensitive if the same unit is measured by more operators. For instance, it is better to have 10 units measured by 9 operators than to have 30 units measured by 3 operators.
Accuracy is calculated by taking each measurement as a data point. Therefore, whether we have 30 units measured 3 times by 3 operators or 10 units measured 3 times by 9 operators or 10 units measured 9 times by 3 operators, there are 270 data points to calculate accuracy.
Treqna Base Manual
First Edition 98 © All Rights Reserved TreQna 2005
Sample Size for the Attribute Gage
Consider the desired resolution to determine the sample size
Greater the desired resolution, higher is the sample size requirement.
If the desired resolution is 0.5%, will a sample size of 100 be enough?
A desired resolution of 0.5% means that the gage needs to distinguish a difference of half percentage point. If the sample size is hundred, the gage is good to distinguish a difference of one percentage point only (1/100), which may not solve the purpose.
Treqna Base Manual
First Edition 99 © All Rights Reserved TreQna 2005
Attribute gage sheet
Repeatability: % times each operator matches on all three measures for the same unit. Total 90 opportunity exist in the example.
Accuracy % times each individual measure matches with the standard. Since there are total 270 measures, there are 270 opportunities
Reproducibility: % times all operators match on all repeated measures for the same unit. Total 30 opportunities exist in the example
TRUE First Operator Second Operator Third Operator
Sample Answer Trial1 Trial2 Trial3 Trial1 Trial2 Trial3 Trial1 Trial2 Trial3
1 N N N N N N N Y N N2 Y Y Y Y Y Y Y Y Y Y3 Y Y Y N Y Y Y Y Y Y4 N N N N Y Y Y N N N5 N N N N N N N N N N6 Y Y Y Y Y Y Y Y Y N7 Y N Y N Y N Y Y Y Y8 N N N N Y N Y N N N9 N N N N N N N N N N
10 N N N N N N N Y N Y11 Y Y Y Y Y Y Y Y Y Y12 Y N N N Y Y Y Y Y Y13 Y Y Y Y Y Y Y Y Y Y14 N N N N N N N N N N15 N N N N N N N Y Y Y16 N N N N N N N N N N17 Y Y Y Y Y Y Y Y Y Y18 N N N N N N N N N N19 Y Y Y Y Y Y Y Y Y Y20 Y N N N Y Y Y Y Y Y21 Y Y Y Y Y Y Y Y Y Y22 N N N N Y Y Y N N N23 N N N N Y Y Y N N N24 N N N N N N N Y Y Y25 N N N N N N N N N N26 Y Y Y Y Y Y Y Y Y Y27 N N N N N N N N N N28 N N N N N N N N N N29 N N N N N N N N N N30 Y N Y N Y Y Y Y Y Y
Treqna Base Manual
First Edition 100 © All Rights Reserved TreQna 2005
Test Re-test study
When to use it?
To check for precision of gage or measurement system. Test re-test is used when same operator does the measurements. It is good practice to use it before opting for complex gage ANOVA.
Remember that if a gage is good on precision, the problem of accuracy can be solved by calibration (correction).
Treqna Base Manual
First Edition 101 © All Rights Reserved TreQna 2005
Test Re-test standard
Precision: Ideally precision should be less than 10% of tolerance. Tolerance is defined by the customer. For example, +10 seconds of specifications would translate into tolerance of 20 seconds.
Accuracy: Can be only measured when one knows the standard value.
Treqna Base Manual
First Edition 102 © All Rights Reserved TreQna 2005
Test re-test gage
Let us take an example where we have readings for diameter of a ball. The tolerance given is 16 mm (+8mm) and the standard diameter of ball is 98mm .
104, 95, 101, 99, 95, 97, 96, 105, 101, 106, 94, 100, 98, 103, 98, 96, 100, 98, 95, 103
Treqna Base Manual
First Edition 103 © All Rights Reserved TreQna 2005
Test re-test gage
Make Run chart to find out if the data is stable
Observation
C1
2018161412108642
107.5
105.0
102.5
100.0
97.5
95.0
Number of runs about median:
0.17906
13Expected number of runs: 11.00000Longest run about median: 3Approx P-Value for Clustering: 0.82094Approx P-Value for Mixtures:
Number of runs up or down:
0.04762
16Expected number of runs: 13.00000Longest run up or down: 2Approx P-Value for Trends: 0.95238Approx P-Value for Oscillation:
Run Chart of C1
Treqna Base Manual
First Edition 104 © All Rights Reserved TreQna 2005
Test re-test gage
Run Descriptive test to calculate mean and standard deviation
Mean=99.2, S=3.6
S=3.6 > 10% of 16 = 1.6
The inaccuracy or bias can be calculated as 99.2-98=1.2
106104102100989694
Median
Mean
10110099989796
A nderson-Darling N ormality Test
V ariance 13.116Skew ness 0.367278Kurtosis -0.959324N 20
M inimum 94.000
A -Squared
1st Q uartile 96.000M edian 98.5003rd Q uartile 102.500M aximum 106.000
95% C onfidence Interv al for M ean
97.505
0.36
100.895
95% C onfidence Interv al for M edian
96.235 101.000
95% C onfidence Interv al for S tDev
2.754 5.290
P -V alue 0.425
M ean 99.200S tDev 3.622
9 5 % Confidence Inter vals
Summary for C1
Therefore, the measurement device has inadequate precision.
Treqna Base Manual
First Edition 105 © All Rights Reserved TreQna 2005
EV and AV
Total variation (R&R) in a measurement system is further classified as:
Equipment Variation: The variation within operator, within equipment or gage, within the method. This variation comes from the parts of measurement system or process. This is also termed as Repeatability variation.
Appraiser Variation: The variation introduced between different operators, different parts, and different methods. This is also termed as Reproducibility variation.
Total variation (R&R) = EV + AV
The mathematical equation used for representing the relationship between total variation, EV and AV is
(Std. Dev (R&R))2 = (Std. Dev (Repeatability))2 + (Std. Dev. (Reproducibility))2
Treqna Base Manual
First Edition 106 © All Rights Reserved TreQna 2005
How a gage plays with Spec Limits
What is the challenge with having high R&R error ?
Following scenario explains why..
Centre circle is the true data point and the limits on left and right show the measurement variation
The gage variation might cause this data to pass on lower specification limit
The gage variation might cause this data to fail on the upper specification limit
The gage variation might cause this data to fail on the lower specification limit
The gage variation might cause this data to pass on upper specification limit
LSL USL
Treqna Base Manual
First Edition 107 © All Rights Reserved TreQna 2005
Gage ANOVA - Objectives
Gage ANOVA study:
Quantify amount of measurement variability
Identify amount of measurement variation from different sources
Data Type: Continuous
Treqna Base Manual
First Edition 108 © All Rights Reserved TreQna 2005
Gage ANOVA - Data Format
The data should be arranged in columns as in the following example. The recommended numbers of operators, parts and trials are 3, 10 and 3 respectively. Which would give one 90 data points. (3*10*3=90)
Operator Part Trial Response1 1 1 20.02 1 1 19.13 1 1 19.54 1 1 18.21 1 2 21.22 1 2 20.03 1 2 19.04 1 2 22.01 2 1 17.02 2 1 19.53 2 1 19.64 2 1 18.01 2 2 19.52 2 2 20.13 2 2 21.04 2 2 18.9
Treqna Base Manual
First Edition 109 © All Rights Reserved TreQna 2005
Gage ANOVA Minitab
Choose Stat> Quality Tools> Gage Study> Gage R&R Study (Crossed).
Treqna Base Manual
First Edition 110 © All Rights Reserved TreQna 2005
Gage ANOVA – Minitab
Operator Part Trial Response1 1 1 20.02 1 1 19.13 1 1 19.54 1 1 18.21 1 2 21.22 1 2 20.03 1 2 19.04 1 2 22.01 2 1 17.02 2 1 19.53 2 1 19.64 2 1 18.01 2 2 19.52 2 2 20.13 2 2 21.04 2 2 18.9
Select method of analysis as ANOVA
Treqna Base Manual
First Edition 111 © All Rights Reserved TreQna 2005
Minitab Session Output
Gage R&R Study - ANOVA Method
Two-Way ANOVA Table With Interaction
Source DF SS MS F PPart 9 18343.8 2038.20 31.5583 0.000Operator 2 530.8 265.40 4.1093 0.034Part * Operator 18 1162.5 64.59 3.9663 0.000Repeatability 30 488.5 16.28Total 59 20525.7
Gage R&R %Contribution
Source VarComp (of VarComp)Total Gage R&R 50.475 13. 30Repeatability 16.283 4.29Reproducibility 34.192 9.01Operator 10.041 2.65Operator*Part 24.151 6.37Part-To-Part 328.936 86.70Total Variation 379.411 100.00
Study Var %Study VarSource StdDev (SD) (6 * SD) (%SV)Total Gage R&R 7.1046 42.627 36.47Repeatability 4.0353 24.212 20.72Reproducibility 5.8474 35.084 30.02Operator 3.1687 19.012 16.27Operator*Part 4.9144 29.486 25.23Part-To-Part 18.1366 108.820 93.11Total Variation 19.4785 116.871 100.00
Number of Distinct Categories = 3
P-value < 0.05 says significant variation. In the example Parts, operator and the interaction all are significant factors
Look for percentage contribution for Total Gage R&R, which is addition of Repeatability and Reproducibility percentage contribution. In the example it is 13.3% which is more than acceptable 10%
Percentage contribution:It represents the percentage of variation contributed by gage in the process.Recommended percentage contribution from Total Gage R&R should ideally be less than 10%. However one should consult with BB/MBB if the value is between 10-15%, where one might accept the gage error and gets go ahead with the project. Project should be evaluated by the mentor and acceptance would depend on the process, business and the project champion.If tolerance is beyond 15% it is recommended that the gage be corrected before repeating the gage study.
No. of Distinct Categories:This number represents the number of distinct groups in the data. For example if you have 20 part being evaluated and the number of distinct categories are 2. This means that the most of the parts are not different enough, and the data can be divided into two groups. In such a case the precision of the gage is not enough for the process. Minimum number of distinct categories should be at least 5.
Treqna Base Manual
First Edition 112 © All Rights Reserved TreQna 2005
Gage ANOVA – Graphical Output
Per
cent
Part-to-PartReprodRepeatGage R&R
100
50
0
% Contribution
% Study Var
Sam
ple
Ran
ge
16
8
0
_R=4.7
UCL=15.36
LCL=0
1 2 3
Sam
ple
Mea
n
100
75
50
__X=80.85UCL=89.69
LCL=72.01
1 2 3
Part10987654321
100
75
50
Operator321
100
75
50
Part
Ave
rage
10 9 8 7 6 5 4 3 2 1
100
75
50
Operator
1
23
Gage name:Date of study :
Reported by :Tolerance:Misc:
Components of Variation
R Chart by Operator
Xbar Chart by Operator
Weight by Part
Weight by Operator
Operator * Part Interaction
Gage R&R (ANOVA) for Weight
In By part graph, there are large differences between different parts.
Components of variation graph shows percent contribution by category
In by operator graph, there isn't much difference between operators
X bar graph by operator shows most of the points out of control, which says most of the variation is due to parts
Treqna Base Manual
First Edition 113 © All Rights Reserved TreQna 2005
Destructive Testing
Destructive testing is unique in the sense that the measured characteristic is different after the measurement process. Few examples of destructive events are:
Crash testing
Tensile strength of material
Call quality evaluation for call centre (unless it is done on recorded call)
Destructive Testing is a unique case since an operator can measure a part multiple times in the earlier case where we used Gage ANOVA
Treqna Base Manual
First Edition 114 © All Rights Reserved TreQna 2005
Destructive Testing and Gage
Assumptions and restrictions:
In such a setup the event happens once and can not be repeated. However if the event is recorded than the measurement can be done multiple times. E.g. call quality check done on recorded calls in call centers.
Each operator measures unique batch.
One must be able to assume that all parts in one batch are identical. The reason it is necessary is otherwise the part to part variation will mask the measurement variation. This means that none of the part is measured by multiple operators
Treqna Base Manual
First Edition 115 © All Rights Reserved TreQna 2005
Topics Covered
MSA Concepts
Understand Accuracy and Precision and their components.
Understand sources of variation
•Perceived vs. True process variation
•Equipment vs. appraiser variation
Why gage error is a challenge
Gage standards
Types of MSAs
Attribute gage
Test retest
Gage R&R
Treqna Base Manual
First Edition 116 © All Rights Reserved TreQna 2005
Measure Phase
Continuous Gage R&R, Short Form Gage,
Test/ Retest Study,Attribute Gage R&R
Identify gage error
Measurement system analysis
Step M.2
Tools
Deliverables
Objective
Step M.3Step M.1
Process Map, QFD,FMEA,Pareto
Sampling StrategySegmentation factors
Identify project Y metricEstablish performance standards
Data collectionCTQ characteristics and standards
DEFINE MEASURE ANALYZE IMPROVE CONTROL
Treqna Base Manual
First Edition 117 © All Rights Reserved TreQna 2005
Topics
Data Collection Plan
Data Segmentation
Sampling
Treqna Base Manual
First Edition 118 © All Rights Reserved TreQna 2005
Data Collection Plan
Following are the steps involved in data collection plan
1. Data Collection Goals
2. Data definition and
procedures
3. Validate measurement
system
4. Collect data and
monitor
•Why collect data?
•What data to collect?
•Establish operational definitions
•Establish data collection procedures
•Pilot operational definitions and procedures and refine
•Establish sampling strategy
•Conduct MSA and validate measurement system
•Collect data
•Monitor for data consistency and make final adjustments
Treqna Base Manual
First Edition 119 © All Rights Reserved TreQna 2005
Data Collection Plan
Till now we have established:
Metric for the project - What is it that we are trying to measure? E.g. If the metric is‘No of New Accounts each day’. One might have to collect data items like – Touch time per application, day of week, No. of applications reworked etc. as additional data that will help us segment and analyze later.
Operational Definitions – We know how to define operational definition. Also we know that operational definition needs to be piloted so that we can fix loose ends in the definition.
Measurement system validation – Depending on the type of data and measurement system setup we know what type of gage to use.
Treqna Base Manual
First Edition 120 © All Rights Reserved TreQna 2005
Data Collection Sheet
What should one capture while collecting data?
Data – Project metric Y or segmentation factor X
Data Type – Continuous vs. Discrete
Data Source – Source/s of data. E.g. Application, process step
Data Range – Possible set of values data can take
Segmentation Factors – All the factors which come out of brainstorming
Time Period – For what time period data is collected
One should also capture if there are any special changes in process during the time period. Such reasons might produce outliers in the data
Monitoring Data Collection
Question collectors for the understanding of operations definitions.
Verify collected data with source, use sampling or ad hoc QC
Ensure that the procedures are followed.
Treqna Base Manual
First Edition 121 © All Rights Reserved TreQna 2005
Topics
Data Collection Plan
Data Segmentation
Sampling
Treqna Base Manual
First Edition 122 © All Rights Reserved TreQna 2005
Data Segmentation
What is Data Segmentation?
Segmentation involves dividing data into logical categories for analyzing data. For instance, while recording the errors made by a data entry process, the project manager may choose to capture the step at which the error occurred, the operator who made the error and so on.
Tools Used for Data Segmentation:
Brainstorming- Members of the process and project team
SME- Subject matter expert for the process can give valuable inputs
Treqna Base Manual
First Edition 123 © All Rights Reserved TreQna 2005
Data Segmentation
Example:
A project is being done on Accuracy of Application processing, the possible segmentation factors are
AccuracyVintage of processor
Work type
Touch time
Day of WeekNo of Processors
These segmentation factors are logical groups that may point to be significant x’s during the analyze phase.
Treqna Base Manual
First Edition 124 © All Rights Reserved TreQna 2005
Topics
Data Collection Plan
Data Segmentation
Sampling
Treqna Base Manual
First Edition 125 © All Rights Reserved TreQna 2005
Sampling - Objectives
Understand sampling and why is it required
Advantages of sampling
Different sampling techniques
Treqna Base Manual
First Edition 126 © All Rights Reserved TreQna 2005
Population vs. Sample
Population:
The entire set of data, the universal set.
Symbol used is N
The population mean represented by µ
Standard deviation is represented by s
Sample:
The part or subset of a population, the universal subset
signifies the sample mean
s signifies standard deviation of a sampleX
Sampling tries to answer few critical questions asked while doing a project:How much data is required?How and when to collect data?
Sampling also reduces the cost involved in collecting data by reducing the amount of data required.
Treqna Base Manual
First Edition 127 © All Rights Reserved TreQna 2005
Sampling
Sampling is the process of collecting subset or portion of data from population and predicting population characteristics
Population N=1000
Sample N=50
Treqna Base Manual
First Edition 128 © All Rights Reserved TreQna 2005
Why Sample ?
When to sample
The cost associated with data is very high
We are measuring a high volume process
When not to sample
When the subset of data may not be able to predict population characteristics. E.g.. If each unit of data is unique
Apart from other reasons to do sampling, one of the critical reasons is the time required to collect population data. Sampling is an efficient way to collect data for any kind of project or analysis.
Treqna Base Manual
First Edition 129 © All Rights Reserved TreQna 2005
Representative Sample
Sample is Representative:
When the sample has same statistics as of the population
How to guarantee a representative sample:
Suitable sampling strategy based on the process
Understand the nature of process
Understand the characteristics of population
Sample must be “representative” of the population, as the data collected otherwise would not deliver any results and might give misleading inferences about the population.
Treqna Base Manual
First Edition 130 © All Rights Reserved TreQna 2005
Sampling Bias
Bias occurs when systematic differences are introduced into the sample as a result of the selection process.
A sample that is biased will not be representative of the population
A sample that is biased will lead to incorrect conclusions about the population
A biased sample would have incorrect information about the population and will lead to biased conclusions. A bias in the sample can not be eliminated with increasing the sample size.
Treqna Base Manual
First Edition 131 © All Rights Reserved TreQna 2005
Sampling Bias Types
Convenience sampling:
This is also called as accidental bias. It occurs we use the data what is available instead of making as effort to collect data. E.g. interviewing the first person you met, pick friends and neighbors for survey.
Systematic sampling:
Collecting data at a particular time, or following a pattern in collecting data which has interference with underlying process. E.g. sampling process on Sunday when only one person works in the process.
Judgmental sampling:
When the data collector or surveyor collects data using his judgment or opinion.
Treqna Base Manual
First Edition 132 © All Rights Reserved TreQna 2005
Sampling Bias Types
Bias by Environment:
This type of bias is introduced when the environmental conditions surrounding process have changed since the sample was collected. E.g. When the call handle time data was collected it was tax season.
Measurement Bias:
Measurement bias could arise if the operational definitions are not correct or the data collectors have interpreted the operational definition differently.
Non-Response Bias:
Non-response bias happens when the respondents to a survey tend to differ systematically from respondents to the survey.
Treqna Base Manual
First Edition 133 © All Rights Reserved TreQna 2005
Sampling Strategy – Random Sampling
Random Sampling:
Samples picked up from assembly line without any order or system. Each item has equal probability of getting selected in the sample.
Production Line
Sample
This is the most commonly used method of sampling. Use random sampling when information about stratification is unknown. Random sampling is a population approach and so care must be taken when the process is cyclical or if the data can be stratified.In random sampling each unit has equal probability of being selected in the sample. Using random sampling method avoids bias being introduced in the sampling process.For practical purpose one can number the data coming out of process and generate random list in excel or on minitab and accordingly sample data.
Treqna Base Manual
First Edition 134 © All Rights Reserved TreQna 2005
Stratified Random Sampling
Stratified Random Sampling
Production Line
Sample 1.1
Sample 1.2
Assembly data stratified in to two categories and random samples collected for both
Stratified random sampling is used when the population has different groups (strata). In such situation it is necessary that both groups are represented in the sample and samples are collected from each group (it is like doing random sampling for each group). The size of the sample depends on population size of each the group. E.g. To sample from a assembly line preparing nuts of two sizes A and B respectively. The two sample sizes for nuts sized A and B would be calculated based on production of size A and size B nuts respectively.
Treqna Base Manual
First Edition 135 © All Rights Reserved TreQna 2005
Systematic Sampling
Systematic Sampling:
Production Line
Sample
Every fourth item selected for sample
Random sampling and stratified sampling are done with historical data. When we have real time data coming in systematic sampling is done.
Unlike random sampling the frequency of collecting data is fixed in systematic sampling.
E.g. selecting every fourth call in the call centre for call barging, selecting 2 application every alternate hour for quality check.
Care must be taken and process must be studied for any underlying structures.
Treqna Base Manual
First Edition 136 © All Rights Reserved TreQna 2005
Rational Sub grouping
Random Sampling:
Production Line
Sample
Three items selected with a frequency of one hour. In this case sample has subgroup size of three.
1st hour2nd hour3rd hour
Rational sub-grouping is a process based sampling strategy. The rational subgroups depend on the nature and type of process from where the data is being selected.
Rational sub-grouping assists us to understand the shift, which is the difference between the long term variability and short term variability of the process. We will study this in capability analysis under Analyze phase.
Treqna Base Manual
First Edition 137 © All Rights Reserved TreQna 2005
Sampling Strategy - Tips
Use following pointers for sampling strategy for a given process:
It is always better to collect small sample spread over longer time period than one large sample over a shorter time period.
Sample more frequently for unstable process and less frequently for stable process
Sample more frequently for process with short cycle time and less frequently for process with long cycle time
To understand the sampling frequency one should understand the objective of data collectionThe most important issue to remember when considering sample frequency is the data collection objective. The sampling frequency is driven by the objective of data collection, e.g. if the data collected is for monitoring process one might want to sample data daily. However if the objective is to collect data for the same process to study capability one might want to collect data for few months by sampling few data points each week or month.
Treqna Base Manual
First Edition 138 © All Rights Reserved TreQna 2005
Sample Size Calc - Continuous Data
Sample represents population characteristics. We can use confidence interval formula and solve it for n to get sample size:
As standard error is also represented by delta, hence:
Solving for n we get:
The equation for 95% confidence becomes (using Z table):
Standard errorXCI nX *2α/Z σ±
n/2Z α=Δ σ
22
Zn
⎟⎟⎟⎟⎟⎟
⎠
⎞
⎜⎜⎜⎜⎜⎜
⎝
⎛
Δ
∗=
σ/α
σ x 1.96 n2
⎟⎠⎞
⎜⎝⎛=
Δ
Where, Δ error or precision required from sample in representing the populationis standard deviation
N is sample sizeThe value of Zα/2 depends on confidence level we choose, the corresponding value can taken out from Z table.
Finite Population CorrectionThe central limit theorem and the standard errors of the mean and of the proportion are based on the premise that the samples selected are chosen with replacement. However, virtually in all scenarios finite population sampling is conducted without replacement from populations that are of a finite size N.When sample is more than 5% of population size we use Finite population correction factor which is:The sample size is than:N (finite) = n(1 + n/N)
1−−
=N
nNFPC
Treqna Base Manual
First Edition 139 © All Rights Reserved TreQna 2005
Estimating Standard Deviation
Use control limits from an existing X chart
= (UCL – X) / 3 or (X – LCL) / 3
Collect some data from population and calculate std deviation (recommended – 30 data points)
Get realistic estimate from business (based on past experience)
Treqna Base Manual
First Edition 140 © All Rights Reserved TreQna 2005
Estimate Δ
Take estimate from business. It may be based on the resolution of measurement. For instance, if the business measures cycle time in days, do not set Δ in minutes or hours
Collect some data from population and calculate Δ using the sample
Treqna Base Manual
First Edition 141 © All Rights Reserved TreQna 2005
Sample Size Calc - Discrete Data
The formula to calculate sample for proportions data is:
nP1P
ZP /2⎟⎠⎞⎜
⎝⎛ −
± α
⎟⎠⎞⎜
⎝⎛
⎟⎟⎟
⎠
⎞
⎜⎜⎜
⎝
⎛= − P1PZn
2/2
Δα
Again if the sample is more than 5% of population finite population correction should be used
Z is the Z score (normal data) and P represents proportionate defective and a represents confidence required from sample. Za/2 is 1.96 for 95% confidence.
This uses the confidence interval formula for attribute (discrete) data:
Treqna Base Manual
First Edition 142 © All Rights Reserved TreQna 2005
Estimating p and Δ
Estimate p
Calculate proportionate defective using small sample
Use historical control charts to estimate
Use subject matter expertise to estimate
Estimate Δ
Use subject matter expertise to estimate
Estimate using formula:
( )n
P1PZ BB/2
−= αΔ
Treqna Base Manual
First Edition 143 © All Rights Reserved TreQna 2005
Topics Covered
Data Collection Plan
Data Segmentation
What is segmentation?
Why segment?
Sampling
What is sampling?
Why sample?
Sampling challenges – Types of Bias
Types of Sampling
Treqna Base Manual
First Edition 144 © All Rights Reserved TreQna 2005
DEFINE
MEASURE
ANALYZE
IMPROVE
CONTROL
Map project
CTQ Characteristics and standards
Measurement System Analysis
Baseline process
Screen for vital Xs
Define Improved Process
MSA on Xs
Approve project
Data collection
Performance objective
Identify drivers of variation
Study interaction between Xs
Improved Process capability
Establish control plan
Treqna Base Manual
First Edition 145 © All Rights Reserved TreQna 2005
Benchmarking
-Concept of Benchmarking
-Benchmarking Types-Benchmarking as
process
Improvement goal for project Y
Performance Objective
Step A.2
Tools
Activities
Deliverables
Objective
Step A.3Step A.1
-Root Cause Analysis (RCA)-Process map analysis-Graphical analysis-Statistical Analysis
-Study data for shape, stability and normality
-Capability analysis-Short Term vs. Long Term
sigma-Understanding shift,
Common cause variation, special cause variation
1-Sample T-test, 2-Sample T-test, One-way ANOVA, Mood’s Median, Homogeneity of Variance, Simple Linear Regression, Correlation/Scatter Diagrams, Chi Square–Test of Independence, Chi Square–Test for Goodness of Fit, Pareto, Cause and Effect diagram, VA/NVA analysis
Descriptive test, run chart, capability pack
List all statistically significant X’sProcess capability for project Y
Identify drivers of variationBaseline Process
Analyze Phase
DEFINE MEASURE ANALYZE IMPROVE CONTROL
Treqna Base Manual
First Edition 146 © All Rights Reserved TreQna 2005
Baseline Process - Objectives
• Study data for stability, shape
• Statistically indicate the nature of the problem
• Calculated the baseline capability of the process.
Describe the process by its
• Descriptive statistics
• Nature of distribution
• Understand Specification Limits and Centering or Target Values
• Calculate: probability of a defect and process capability
• Use Minitab for these purposes
Treqna Base Manual
First Edition 147 © All Rights Reserved TreQna 2005
Stability, Shape and Normality
Stability
Tool Used: Run Chart
Stability is lack of excessive variation.There should be no trends, clusters, mixtures or oscillations.
Shape
Tool Used: Descriptive Stats,Histogram
Study distribution characteristics
Normality
Tool Used: Descriptive test, NormalityTest
Check if the data follows a normaldistributions
Treqna Base Manual
First Edition 148 © All Rights Reserved TreQna 2005
Important thing to remember is that data should be in a time order before a run chart is plotted. If the data is sorted or it does not follow the time order in any other way, run chart loses its significance. Run chart studies the stability of process over a period of time by studying any abnormality is data like trends, clusters and so on.
Run chart can be plotted for:Individual valuesMeans or Medians of subgroups (if subgroups are present)
Apart from looking for process stability over a period of time, run chart can also give graphical representation of pre and post process changes comparison as it is plotted on time scale.
The only difference between run chart and control chart comes from the control limits.
Stability - Run Chart
Tool Utility: To track process over a period of time for stability.Tests for if there are any trends, clusters, oscillations in the process. Indicates any special cause variation in the process.
How: Minitab. STATS > QUALITY TOOLS > RUN CHART
Treqna Base Manual
First Edition 149 © All Rights Reserved TreQna 2005
Run Chart - Minitab
MINITAB FILE: RunChart.mtw
Treqna Base Manual
First Edition 150 © All Rights Reserved TreQna 2005
The graph you see in the output window of Minitab is the run chart. Run chart is used for stability and helps in identifying if there is any special cause in the process and when does it appear on time scale. Below the graph you would see hypothesis tests for different measures.Run Chart provides two tests for randomness:Based on the number of runs about the median (Mixtures and Clusters)Based on the number of runs up or down. (Trends and Oscillations)
P-values for the following special cases should be read from run chart out put:Trend:Ho: No fewer runs observed than expectedHa: Fewer runs observed than expectedA trend is a sustained drift in the data, either up or down. Might indicate if the process is about to go out of control or is out of control.Oscillations:Ho: No more runs observed than expected Ha: More runs observed than expectedOscillation occurs when the data fluctuates up and down rapidly, indicating that the process is not steadyClusters:Ho: No fewer runs observed than expected. Ha: Fewer runs observed than expectedClusters are group to group variability which may indicate variation due to special causes, such as measurement problems or sampling from a bad group of parts.Mixtures:Ho: No more runs observed than expected. Ha: More runs observed than expectedMixture can be identified by absence of points near the center line. Mixtures often indicate combined data from two populations, or two processes operating at different levels.`
Run Chart - Subgroup Size 1
• Choose ‘C2’ for Data• Set Subgroup Size to ‘1’• Click Ok
Treqna Base Manual
First Edition 151 © All Rights Reserved TreQna 2005
Studying Stability
Ho: Data is random, special causes not presentHa: Data is not random, special causes present
Treqna Base Manual
First Edition 152 © All Rights Reserved TreQna 2005
Studying Stability
Treqna Base Manual
First Edition 153 © All Rights Reserved TreQna 2005
If you have done rational subgrouping in your data. It can be used for run chart as well. In the input window you can mention subgroup size or column representing subgrouping in the data.
The run chart would than plot mean or median depending on your choice. It also plots the variation within the subgroup apart from the stability over a period of time.
Run Chart – Subgroup Size more than 1
Treqna Base Manual
First Edition 154 © All Rights Reserved TreQna 2005
Shape - Descriptive Stats
Click on Stat > Basic Statistics > Display descriptive Statistics
MINITAB FILE: DescriptiveStats.mtw
Descriptive stats helps studying the shape of data distribution, measures of central tendency and dispersion.It also helps to find out what kind of issue we have with the process: whether it is issue with centering or issue with variation ?
Treqna Base Manual
First Edition 155 © All Rights Reserved TreQna 2005
Tabular Output
1.Double click on “C1 Agent A”2.Click Ok
Results for: DescriptiveStats.MTW
Descriptive Statistics: Agent A
Variable N N* Mean SE Mean StDev Minimum Q1 Median Q3
Agent A 250 0 50.424 0.535 8.460 19.000 45.750 51.000 56.000
Variable Maximum
Agent A 71.000
Treqna Base Manual
First Edition 156 © All Rights Reserved TreQna 2005
Graphical Output
Click on Stat > Basic Statistics > Graphical Summary
MINITAB FILE: DescriptiveStats.mtw
1. Double click on “C1 Agent A”2. Click Ok
Graphical summary from minitab lists all the basic statistics of the data which includes measures of central tendency, variation(spread) and there graphical representation.Following are the key statistic output from graphical summary.Normality P valueMeanMedianStandard DeviationVarianceKurtosis (measure of skew ness of data)Quartiles
In graphical representation:Histogram (with o without normal curve)Box plot and Dot plot
In the following pages we would cover few basis statistic tools.
Treqna Base Manual
First Edition 157 © All Rights Reserved TreQna 2005
Different Shapes
Normal
Skewed
Central Tendency: Mean
Dispersion: Standard Deviation
Central Tendency: Q1 or Q3
Dispersion: Stability Factor, defined as Q1/Q3
Treqna Base Manual
First Edition 158 © All Rights Reserved TreQna 2005
Different Shapes
Bi-modal
Central Tendency: Median
Dispersion: Range or Span
Such situation suggests data coming from two processes and not one. One should go back and have a look at the processes, separate the data by process and than analyze.
Long Tailed
Treqna Base Manual
First Edition 159 © All Rights Reserved TreQna 2005
Normality
One can test normality of data using:
• Descriptive test
• Normality test
Hypothesis used for the test is:
H0: data follow a normal distribution
H1: data do not follow a normal distribution
Treqna Base Manual
First Edition 160 © All Rights Reserved TreQna 2005
99.73%
95.46%
Standard Deviations
3210- 1- 2- 3
68.26%
Area under a Normal Curve
The probability of a randomly selected value of being within 2 Standard Deviations from the mean is 95.46%
+-
Treqna Base Manual
First Edition 161 © All Rights Reserved TreQna 2005
Z - Value
For any value of X in a distribution there is an equivalent Z – value. The Z – value is a measure of the number of Standard Deviations that will fit between X and the Mean.
σμX-=Z
z
x 3 σ 2 σ 1 σ μ
Statistically,
The Z-value is used to transform any normal distribution in terms of the standard normal distribution with a Mean of 0 and Standard Deviation of 1.The Z – value is important as it allows us to compare to dissimilar distributions by using standard deviation as the unit of measure.
Note : Z Value and Z capability are not the same.
Treqna Base Manual
First Edition 162 © All Rights Reserved TreQna 2005
Z - Value
43210- 1- 2- 3- 4
18161412108642
Standard Deviations
Units of Measure
USLUSL
ZUSL
LSLLSL
ZLSL
Χ
2 24
210-14
sX- USL ZUSL ====
3 26
24-10
s LSL-X ZLSL ====
10X =
2s =
A Simple Z Capability CalculationFor this calculation assumeUSL = 14 LSL = 4Nominal = 10Note :
Nominal is also called Target ValueHow the calculation changes for LSL and USL, this is done to ensure that the resulting
number is always positive.The total area under any distribution is always 1.The Shaded Region represents the probability of a value being outside specification limits.
Treqna Base Manual
First Edition 163 © All Rights Reserved TreQna 2005
The Total Probability of a defect is calculated by adding the probability of defects that will occur at values lesser than the LSL and greater than USL. These values can be obtained by using the Z capability values you calculated earlier for both ZLSL and ZUSL.. Using the Z table, then find the Z-value that corresponds to this total probability of defect, this is ZBench and it represent long term process capability.To calculate process capability for the short term, just add Z Shift to the long term capability.
Note :Z Shift is empirically set at 1.5If it is not mentioned whether the data used is for sort or long term, always assume it to be long term. If the data is for short term, your Z – Bench will be short term capability, then subtract Z shift to obtain Z long term.1 – Z bench is the probability of a conformance also called yield, this yield can directly be converted to Sigma
by using a sigma conversion table.
Calculating Z – Bench value
USLLSLTotal P(d)P(d)P(d) +=
ZLT = from Z table
ZST = ZLT + Z shift
LSLP(d) USLP(d)
USL T X LSL
Treqna Base Manual
First Edition 164 © All Rights Reserved TreQna 2005
Single-Tail Z Table A (Values of Z from 0.00 to 4.99)
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-0010.10 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-0010.20 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-0010.30 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-0010.40 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-0010.50 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-0010.60 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-0010.70 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-0010.80 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-0010.90 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-0011.10 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-0011.20 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-0021.30 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-0021.40 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-0021.50 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-0021.60 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-0021.70 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-0021.80 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-0021.90 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-0022.10 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-0022.20 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-0022.30 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-0032.40 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-0032.50 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-0032.60 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-0032.70 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-0032.80 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-0032.90 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-0033.10 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-0043.20 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-0043.30 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-0043.40 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-0043.50 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-0043.60 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-0043.70 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-0053.80 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-0053.90 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-0054.10 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-0054.20 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-0064.30 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-0064.40 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-0064.50 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-0064.60 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-0064.70 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-0074.80 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-0074.90 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
Treqna Base Manual
First Edition 165 © All Rights Reserved TreQna 2005
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-0075.10 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-0075.20 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-0085.30 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-0085.40 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-0085.50 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-0085.60 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-0095.70 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-0095.80 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-0095.90 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-0106.10 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-0106.20 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-0106.30 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-0116.40 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-0116.50 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-0116.60 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-0116.70 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-0126.80 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-0126.90 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-0137.10 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-0137.20 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-0137.30 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-0147.40 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-0147.50 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-0147.60 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-0157.70 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-0157.80 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-0157.90 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-0168.10 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-0168.20 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-0178.30 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-0178.40 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-0178.50 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-0188.60 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-0188.70 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-0198.80 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-0198.90 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-0209.10 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-0209.20 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-0219.30 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-0219.40 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-0219.50 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-0229.60 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-0229.70 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-0239.80 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-0239.90 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
Single-Tail Z Table A (Values of Z from 5.00 to 9.99)
Treqna Base Manual
First Edition 166 © All Rights Reserved TreQna 2005
Standard Norm
al D
istribution Table BZ
This Is K
nown A
s The Standard N
ormal D
istribution W
hich H
as A M
ean Of Zero (=
0) An
d A Stan
dard D
eviation of On
e (s = 1)
Treqna Base Manual
First Edition 167 © All Rights Reserved TreQna 2005
Standard Norm
al D
istribution Table B
This Is K
now
n A
s The Stan
dard Norm
al Distribu
tion Wh
ich Has A
Mean O
f Zero (= 0) A
nd A
Standard
Deviation of O
ne (s =
1)
Z
Treqna Base Manual
First Edition 168 © All Rights Reserved TreQna 2005
The Total Probability of a defect is calculated by adding the probability of defects that will occur at values lesser than the LSL and greater than USL. These values can be obtained by using the Z capability values you calculated earlier for both ZLSL and ZUSL.. Using the Z table, then find the Z-value that corresponds to this total probability of defect, this is ZBench and it represent long term process capability.To calculate process capability for the short term, just add Z Shift to the long term capability.
Note :Z Shift is empirically set at 1.5If it is not mentioned whether the data used is for sort or long term, always assume it to be long
term. If the data is for short term, your Z – Bench will be short term capability, then subtract Z shift to obtain Z long term.1 – Z bench is the probability of a conformance also called yield, this yield can directly be
converted to Sigma by using a sigma conversion table.
Process Capability - Objectives
• Calculate Process Capability for Discrete Data
• Calculate Process Capability for Continuous Data
• Understand rational subgroups and use it for sampling and data analysis
• Perform the following calculations:
• within, between and total sum of squares
• the long and short-term standard deviation and Z-values
• understand the difference between short-term process capability and long-term process capability
• establish whether the process has control issue or requires technology improvement
Treqna Base Manual
First Edition 169 © All Rights Reserved TreQna 2005
Choosing a Capability Tool
Start
Is data discrete ?
Run Discrete Data Process Capability
report (excel spreadsheet)
ZST = ZLT + 1.5(ZShift = 1.5)
Is data normal ?
Treat data as discrete for this step and run Discrete Data Process Capability
report (see notes)
Run Capability Analysis (normal data)
Is target specified?
Read ZLT from Capability Analysis
report and 1.5 to arrive at ZST
Run Capability Sixpack (see next page for
interpretation)Is sub group size
> 1?Read ZLT from Capability Analysis report and 1.5 to
arrive at ZST
No
No
No
No
Yes
Yes
Yes
Yes
To convert continuous data to discrete follow the following steps1. Minitab>Stat>Tables>Tally Individual variables2. Under Variable select column that has data on project Y3. Select “Count” and “Cumulative percent”4. Look for yield @ USL or LSL in session window5. Calculate Z value using Abridge table
Treqna Base Manual
First Edition 170 © All Rights Reserved TreQna 2005
Process Capability for Discrete Data
ZLT : Long term capability can be calculated using excel formula normsinv(yield)
ZST : Short term capability is computed by adding 1.5 (standard shift) to the long term
sigma
Treqna Base Manual
First Edition 171 © All Rights Reserved TreQna 2005
Process Capability - Continuous Data
• Objective: Practice generating a Minitab Capability Sixpack
• Directions:
• Open file Penlength.mtw
• The capability of a Pen manufacturing plant is being calculated. The ‘length of the pen’ is the measure.
• USL = 10.2 cms
• LSL = 9.8 cms
• The data was collected in batches of 10(Subgroup size = 10)
• Stat > Quality Tools > Capability Sixpack > Normal
• In ‘Options’ – Enter Target = 10
If subgroup size = 1, the short term sigma value reported in the Capability Analysis report is invalid as there are no sub groups and hence no sub group variation.
If upper and lower specification limits are provided without a target, Minitab assumes that the target is the mid point of the specification range. If only one specification limit is entered and the target is left blank, then Minitab approximates the target with the mean of the data.
Treqna Base Manual
First Edition 172 © All Rights Reserved TreQna 2005
Process Capability - Continuous Data
Treqna Base Manual
First Edition 173 © All Rights Reserved TreQna 2005
Process Capability - Continuous Data
Treqna Base Manual
First Edition 174 © All Rights Reserved TreQna 2005
Process Capability - Continuous Data
Treqna Base Manual
First Edition 175 © All Rights Reserved TreQna 2005
Interpreting the Capability Sixpack
Always check for S Chart stability to judge whether rational sub grouping is acceptable- S chart in control indicates sub groups are acceptable- S chart out of control indicates sampling is not acceptable
-If the graph shows an R Chart instead of S then,If R chart is out of control, Xbar chart interpretation may not be reliable as the control limits on the Xbar chart are arrived at using the Rbar value from the R chart.If the R chart is not in control, the rational sub grouping is not correct. Within subgroup variation is higher than between sub group variation.
Treqna Base Manual
First Edition 176 © All Rights Reserved TreQna 2005Note: Subtract 1.5 to get long-term Sigma level
Abridged Yield to Sigma Conversion TableLong-Term
YieldST Process
SigmaDefects Per 1,000,000
Defects Per 100,000
Defects Per 10,000
Defects Per 1,000
Defects Per 100
99.99966%99.9995%99.9992%99.9990%99.9980%99.9970%99.9960%99.9930%99.9900%99.9850%
99.9770%99.9670%99.9520%99.9320%99.9040%99.8650%99.8140%99.7450%99.6540%99.5340%
99.3790%99.1810%98.930%98.610%98.220%97.730%97.130%96.410%95.540%94.520%
93.320%91.920%90.320%88.50%86.50%84.20%81.60%78.80%75.80%72.60%
69.20%65.60%61.80%58.00%54.00%
50%46%43%39%35%
31%28%25%22%19%16%14%12%10%8%
3.4581020304070100150230330480680960
1,3501,8602,5503,4604,6606,2108,19010,70013,90017,80022,70028,70035,90044,60054,80066,80080,80096,800115,000135,000158,000184,000212,000242,000274,000
308,000344,000382,000420,000460,000500,000540,000570,000610,000650,000
690,000720,000750,000780,000810,000840,000860,000880,000900,000920,000
0.340.50.81234710152333486896135186255346466621819
1,0701,3901,7802,2702,8703,5904,4605,4806,6808,0809,68011,50013,50015,80018,40021,20024,20027,40030,80034,40038,20042,00046,00050,00054,00057,00061,00065,00069,00072,00075,00078,00081,00084,00086,00088,00090,00092,000
0.0340.050.080.10.20.30.40.71.01.52.33.34.86.89.613.518.625.534.646.662.181.9107139178227287359446548668808968
1,1501,3501,5801,8402,1202,4202,7403,0803,4403,8204,2004,6005,0005,4005,7006,1006,5006,9007,2007,5007,8008,1008,4008,6008,8009,0009,200
0.00340.0050.0080.010.020.030.040.070.10.150.230.330.480.680.961.351.862.553.464.666.218.1910.713.917.822.728.735.944.654.866.880.896.8115135158184212242274308344382420460500540570610650690720750780810840860880900920
0.000340.00050.00080.0010.0020.0030.0040.0070.010.0150.0230.0330.0480.0680.0960.1350.1860.2550.3460.4660.6210.8191.071.391.782.272.873.594.465.486.688.089.6811.513.515.818.421.224.227.430.834.438.24246505457616569727578818486889092
6.05.95.85.75.65.55.45.35.25.15.04.94.84.74.64.54.44.34.24.14.03.93.83.73.63.53.43.33.23.13.02.92.82.72.62.52.42.32.22.12.01.91.81.71.61.51.41.31.21.11.00.90.80.70.60.50.40.30.20.1
Treqna Base Manual
First Edition 177 © All Rights Reserved TreQna 2005
Long Term & Short Term Capability
ZLT = Long term process sigma
Long term sigma is also the sustained process capability
ZST = Short term process sigma
The process at entitlement is performing at its best. This can happen when process short term sigma (Zst) is as high as possible.
ZST = ZLT + Shift
Where shift is the difference between short term process capability and long term process capability.
Shift is taken as 1.5 if the data lacks rational sub grouping*
* Discussed Later
ZST assumes:Theoretically, process performance is more consistent in the short term than the long term, hence short term sigma is always better than the long term sigma.
Treqna Base Manual
First Edition 178 © All Rights Reserved TreQna 2005
Therefore, six sigma is a measure of the short term capability of a process. If a normal process is centered at 6 standard deviations from each specification limit (or is on target) in the short term, assuming standard shift, the process would tend to move 1.5 standard deviations towards either specification limit in the long term. This means, the process will be 4.5 standard deviations from one specification limit and 7.5 standard deviations from the other.
The assumed shift of 1.5 sigma comes from manufacturing industry. Shift compensates for the variation which is non-random in nature and does not get captured while calculating short term sigma. It accounts for the changes in process capability over many cycles of that process.
LSL ± 6s
Centered Distribution
TUSL
4.5sT
mmDistribution Shifted 1.5σ
LSL USL
The Shift
Treqna Base Manual
First Edition 179 © All Rights Reserved TreQna 2005
As discussed in the previous slide, the process mean typically shifts by 1.5 standard deviations around the target. Short term sigma is a measure of the process at a particular point in time whereas long term sigma is a measure of the process over a period of time.
LSL USLT
Time = t
Short-Term capability
Long-Term capability
Visual Representation of ST & LT
Time = t + x
Treqna Base Manual
First Edition 180 © All Rights Reserved TreQna 2005
Non-Random Variation:ShiftsTrendsCycles
Types of Variation
Common Cause variation• Random process variation• Also called as inherent variation of any given process• Variation comes from
– People– Materials– Methods– Machines– Measurements– Mother Nature
Special Cause variation • Non-random process variation• Assignable cause• Can generate outliers
Treqna Base Manual
First Edition 181 © All Rights Reserved TreQna 2005
To create rational subgroups find:A time period that is short enough such that you do not have special causes of variationThis time period will differ from process to process
A subgroup should only contain common cause variation. There is variation between different subgroups. This variation is accounted primarily by special causes (though this might include some common cause variation also)In other words, short term capability should have common cause variation only. Long term capability has common cause and special cause variation
Special Cause Variation – between groups
RationalSubgroups
Common Cause Variation- Within Group
The objective behind doing Rational Subgrouping is to have subgroups with only common cause variation and minimal special cause variation within them. The attempt is made to collect samples of data in such a way that special cause accounts for variation between different subgroups.
Rational Subgrouping
TimeTime
Treqna Base Manual
First Edition 182 © All Rights Reserved TreQna 2005
Rational Subgroups
Characteristics of good rational subgroup:
– Samples to be collected over a large period of time so that each sample represents a rational subgroup defined by some logic(which usually is a one process cycle)
– Attempt to minimize variation within the subgroup and maximize the variation between subgroups
Examples of sampling to create rational subgroups:
– Sample each hour, 3 applications
– Sample each day, 3 applications
– Sample each shift, 3 applications
In high-volume situations, use the Xbar and R-chart to evaluate stability of short/long-term stability.
In low-volume situations, process measurements typically come “one-at-a-time.”Use the X & RM chart.
Treqna Base Manual
First Edition 183 © All Rights Reserved TreQna 2005
SSTotal : deviation of each individual point from the overall meanSSBetween: deviation of each subgroup mean from the overall mean (variation of the means)SSWithin :deviation of each individual point from its corresponding group mean (variation of the individuals within group)Xij : Individual observation
Xj : Subgroup mean
X : Overall Mean
Components of Variation
( ) ( )2
11
2
1
2
11jij
ijj
jij
ijXXNXX
NKKNK
XX −+⎟⎠⎞
⎜⎝⎛ −=− ∑∑∑∑∑
=====
[Samples size n=3 at different times
SSTTotalSST
TotalSSW
WithinSSW
WithinSSB
BetweenSSB
Between
Where K=number of subgroups and N= subgroup size
Total variation representing Long term process capability
The Shift defined by special cause variation
Common cause variation representing Short term process capability
[}}SSW
SST
SSB
Treqna Base Manual
First Edition 184 © All Rights Reserved TreQna 2005
Standard Deviation & Sum of Squares
One Sample
Short-Term
Long-Term
( )1gn
XXσ̂
g
1j
n
1i
2
ij
LT−
−=
∑ ∑= =
( )( )1ng
XXσ
g
1j
n
1i
2 jij
ST −
−=
∑ ∑= =ˆ
( )( )1n
XXσ̂
n
i
2 i
−
−=
∑Where “n” is the sample size
Where “n” is the subgroup size and “g” is the number of subgroups. Numerator denotes SSW
Where “n” is the subgroup size and “g” is the number of subgroups. Numerator denotes SST. X represents overall mean.=
Treqna Base Manual
First Edition 185 © All Rights Reserved TreQna 2005
Z Long-term: Which is the actual capability of the process and what is current. It is always calculated w.r.t. the population mean. SL - m is the distance between specification limits given by customer and the true population mean.
Z Short-term:Z short term is the best process can deliver if the current distribution is assumed around the Target (and not mean). This means assuming the same variation and if the distribution were to centered around Target what will be process capability. For the same reason while calculating Z short term we use Target value and not population mean.SL - T is the distance the specification is from the target.
ZST is bigger than ZLT, so the shift is
Higher the shift bigger the control issue in process. Typical assumed shift when rational subgrouping not present is 1.5 sigma.
The Equation
Z Equation
Z = Specification Limit – Central TendencyStandard Deviation
Variations
Z – Short Term(ZST ) or Long Term(ZLT)
Specification Limit – USL or LSL
Central Tendency – Mean(m) or Target(T)
Standard Deviation – σST or σLT
LTLT σ
μSLZ
−=
STST σ
TSLZ
−=
ZShift=ZST-ZLT
Treqna Base Manual
First Edition 186 © All Rights Reserved TreQna 2005
Relating ST Capability & Process Shift
Short Term Sigma
• Short Term Sigma signifies the best process can deliver as of current.
• Entitlement of process with current technology
• Any attempt to increase the capability would require improvement in current technology
Shift
• Shift in the distribution over a period of time, the difference between long term and short term capability
• Caused by assignable special causes
• Signifies control issues in the process
Seeing Short term Sigma and Shift together provides insight into what needs to be changed to improve Long Term process capability.
Poor Control;Poor Control;Poor TechnologyPoor Technology
Poor Control;Poor Control;
Poor TechnologyPoor TechnologyGood Control;Good Control;
Good TechnologyGood Technology
ZZshiftshift
ZZstst
2.5
2.0
1.5
1.0
0.5
1 2 3 4 5 6
The graph to your right helps identify the issue in your process. Once you baseline your process and understand it’s short term capability and shift, you will be in a position to determine whether your process needs better controls, improved technology or both.
A typical DMAIC project aims at improving controls by eliminating special cause variation. DFSS projects on the other hand are run to improve technology so that the process can reach higher capability.
Treqna Base Manual
First Edition 187 © All Rights Reserved TreQna 2005
ZST is based on the analysis of subgroups. If special cause variation is also present along with common cause variation, ZST will not reflect the correct (best) picture of present technology.
A small ZShift does not necessarily indicate good control, but consistency among the subgroups or need to revisit sub groups. However, if ZShift is greater than 1.5, this is a definite indication of a controlproblem.
Summary LT vs. ST
LONG-TERM CAPABILITY (ZLT)
• Actual current process capability
• Way to improve involves improving control (reducing shift) and technology
• 6s means 4.5s with assumed shift of 1.5s, otherwise 6s - (shift)
SHORT-TERM CAPABILITY (ZST)
• Best process performance –Entitlement
• Way to improve involves improving technology as it is the inherent capability of process at current
• 6σ means ZST = 6σ
Treqna Base Manual
First Edition 188 © All Rights Reserved TreQna 2005
Step Summary – Baseline Process
Discrete Data Characteristics• Defect, Defectives• DPMO• For capability Use ZLT from Z table corresponding to DPMO
Continuous data characteristics• Central Tendency – mean, median• Variation – variance, standard deviation• For capability use Z values from minitab capability reports
Normality• Process with only random variation will have such distribution• Can always be related to normal distribution which has mean of zero and
standard deviation of oneSample:
• Subset of population which has same characteristics as population.
Treqna Base Manual
First Edition 189 © All Rights Reserved TreQna 2005
Basic Statistics tools:Use graphical summary, box plots, run charts(for stability), normality graph
from minitab to understand process
Rational Subgrouping• Attempt is made to collect samples with only random variation(common
cause variation)• Between subgroup variation is special cause variation• Derive SSB and SSW to understand ZST, ZLT and Shift
Z-value• Signifies number of standard deviation that can be fit between customer
specification limit and mean• Corresponds to probability of defect, which is area under the curve outside
the customer specification limit• Represents process capability
Step Summary – Baseline Process
Treqna Base Manual
First Edition 190 © All Rights Reserved TreQna 2005
Shift
• Difference between ZLT and ZST
• Signifies process control issues
• Assume shift to be 1.5s to estimate ZST if rational subgrouping not present
• Attributed by special and assignable causes
What the customer feels is the process Long term capability. Short term capability is the inherent capability of process – “Entitlement”
Step Summary – Baseline Process
Treqna Base Manual
First Edition 191 © All Rights Reserved TreQna 2005
Benchmarking
-Concept of Benchmarking
-Benchmarking Types-Benchmarking as
process
Set Improvement goal for project Y
Performance Objective
Step A.2
Tools
Activities
Deliverables
Objective
Step A.3Step A.1
-Root Cause Analysis (RCA)-Process map analysis-Graphical analysis-Statistical Analysis
-Study data for shape, stability and normality
-Capability analysis-Short Term vs. Long Term
sigma-Understanding shift,
Common cause variation, special cause variation
1-Sample T-test, 2-Sample T-test, One-way ANOVA, Mood’s Median, Homogeneity of Variance, Simple Linear Regression, Correlation/Scatter Diagrams, Chi Square–Test of Independence, Chi Square–Test for Goodness of Fit, Pareto, Cause and Effect diagram, VA/NVA analysis
Descriptive test, run chart, capability pack
List all statistically significant X’sProcess capability for project Y
Identify drivers of variationBaseline Process
Analyze Phase
DEFINE MEASURE ANALYZE IMPROVE CONTROL
Treqna Base Manual
First Edition 192 © All Rights Reserved TreQna 2005
Setting a goal for the Y metric
Following are approaches which one could adopt to define the Goal for Y metric:
• Benchmarking: One can target to achieve best in industry
• Learning Curve-Based: One could also use multi level goal in terms of sigma
for the process metric.
• Arbitrary Defect Reduction: This is frequently used with discrete metric.
E.g.reduce DPMO by 50%
Six sigma should always have aggressive but achievable target
Treqna Base Manual
First Edition 193 © All Rights Reserved TreQna 2005
Benchmark and Baseline
Z Long Term
Benchmark is the best in industry, process baseline tells where the process stands against Industry
Benchmark : World Class Performance
Z Short Term : The best performance by the process with current technology
Baseline Process Sigma : The sustained long term sigma for the process
Z Short TermBenchmark
Treqna Base Manual
First Edition 194 © All Rights Reserved TreQna 2005
Benchmarking
Definition:
An improvement process in which a company measures its performance against that
of best in class companies, determines how those companies achieved their
performance levels and uses the information to improve its own performance. The
subjects that can be benchmarked include strategies, operations, processes and
procedures.
Where to use Benchmarking:
• Compare process performance against industry’s best
• Understand market competition
• Establish improvement targets for your process
Benchmarking is a continuous process which gives insight into where does your competitor stand and how do they achieve those levels of service excellence. It also provides a platform for best practice sharing across industry.
The knowledge derived from the process of benchmarking can be used to improve service, products, support functions and systems. It also provides target for process capability improvement.
In the current market scenario where information flows seamless and the access is universal, benchmarking is definitely a tool to facilitate service excellence achievement .
Treqna Base Manual
First Edition 195 © All Rights Reserved TreQna 2005
Types Of Benchmarking
Strategic BenchmarkingIt is used where organizations seek to improve their overall performance by examining the long-term strategies and approach. It involves considering high level aspects such as core competencies, developing new products and services.
Performance/ Competitive BenchmarkingIt is used where organizations consider their positions in relation to performance characteristics of key products and services. Usually focuses is the relevant segment of market.
Process BenchmarkingIt is used when the focus is on improving specific critical processes and operations. Benchmarking partners are sought from best practice organizations that perform similar work or deliver similar services. It helps in driving short term goals of company.
Functional BenchmarkingIt is used when organizations look to benchmark with market players drawn from different business sectors to find ways of improving similar functions or work processes. This sort of benchmarking can lead to innovation and dramatic improvements.
Treqna Base Manual
First Edition 196 © All Rights Reserved TreQna 2005
Types Of Benchmarking
Internal BenchmarkingIt involves seeking information from within the same organization, for example, from business units located in different areas. The main advantages of internal benchmarking are that access to sensitive data and information are easier; standardized data is often readily available; and, usually less time and resources are needed.
External BenchmarkingExternal Benchmarking involves seeking outside organizations that are known to be best in class. External benchmarking provides opportunities of learning from those who are at the leading edge.
International BenchmarkingIt is used where partners are sought from other countries because best practitioners are located elsewhere in the world and/or there are too few benchmarking partners within the same country to produce valid results
Treqna Base Manual
First Edition 197 © All Rights Reserved TreQna 2005
Benchmarking Plan
Primarily following are the things to be considered in your benchmarking plan:
• Focus of Study
• Current assessment of area requiring improvement
• Selection of process for benchmarking
• Description of current process
• Purpose of study
• Outline of areas/issues for questioning
• Benchmarking partners
• rationale for selection
• initial contact
• development of detailed questions
Treqna Base Manual
First Edition 198 © All Rights Reserved TreQna 2005
• Site Visit
•confirm and arrange site visit
•develop briefing package
•finalize and communicate plans
•conduct visit
• Post Visit
•points of clarification / follow up
•review all data
•initial report sent to partners for confirmation
Benchmarking Plan
Treqna Base Manual
First Edition 199 © All Rights Reserved TreQna 2005
Other sources for setting a goal
Following are few more options which drive the goal for process metric:
• Corporate mandate
• Compliance/legal requirement
• Voice Of Customer
• Industry Standards (e.g. CMM, ISO)
Treqna Base Manual
First Edition 200 © All Rights Reserved TreQna 2005
What is Benchmarking?– Benchmarking is a continuous process– Valuable tool to gather market knowledge and improve services, products and
systems– Process of knowing how competitor is achieving higher levels of service
excellence– Benchmarking is a time consuming and labor intensive process
Common Benchmarking Mistakes:– Not establishing the baseline– Confusing benchmarking with participating in a survey. – Thinking there are pre-existing "benchmarks" to be found.– The process is too large and complex to be manageable– Confusing benchmarking with research– Picking a topic that is too intangible and difficult to measure – Not researching benchmarking partners thoroughly– Not having a code of ethics and contract agreed with partners
Summary – Performance Objective
Treqna Base Manual
First Edition 201 © All Rights Reserved TreQna 2005
Benchmarking
-Concept of Benchmarking
-Benchmarking Types-Benchmarking as
process
Set Improvement goal for project Y
Performance Objective
Step A.2
Tools
Activities
Deliverables
Objective
Step A.3Step A.1
-Root Cause Analysis (RCA)-Process map analysis-Graphical analysis-Statistical Analysis
-Study data for shape, stability and normality
-Capability analysis-Short Term vs. Long Term
sigma-Understanding shift,
Common cause variation, special cause variation
1-Sample T-test, 2-Sample T-test, One-way ANOVA, Mood’s Median, Homogeneity of Variance, Simple Linear Regression, Correlation/Scatter Diagrams, Chi Square–Test of Independence, Chi Square–Test for Goodness of Fit, Pareto, Cause and Effect diagram, VA/NVA analysis
Descriptive test, run chart, capability pack
List all statistically significant X’sProcess capability for project Y
Identify drivers of variationBaseline Process
Analyze Phase
DEFINE MEASURE ANALYZE IMPROVE CONTROL
Treqna Base Manual
First Edition 202 © All Rights Reserved TreQna 2005
Bridging the Gap -Practical to Statistical
Practical Problem
Statistical Problem
Statistical Solution
D
M
A
I
C• MSA on Vital X’s
• Control Plan
• Project Y• Project Charter• MSA• Data Collection
• Process Characteristics
- Centering and Variation
• Establish Process Capability
• Determine Statistical Goal (Benchmarking)
• Critical X’s
• Transfer function Y=f(X)
• Pilot/Implement Solution
Practical Solution
This section is the beginning of statistical problem solving process. By the end of this section we would have established Critical X’s
Treqna Base Manual
First Edition 203 © All Rights Reserved TreQna 2005
Tools to be used in this section
Graphical Tools1. Fishbone analysis
2. Matrices
3. Pareto
4. Histograms
5. Dot Plot
6. Box and whisker plot
7. Process map analysis
• VA/NVA analysis
• Value mapping
8. Hypothesis Tests
Cause and Effect diagram is used in measure for identifying segmentation factors. It is also used to prioritize X’s which affect project Y
Treqna Base Manual
First Edition 204 © All Rights Reserved TreQna 2005
Cause and Effect Diagram (Fishbone)
Use the tool:
• To get a visual map of all causes
• While brainstorming for causes and trying to capture thoughts
• Correctly identify root causes and categorize them
Treqna Base Manual
First Edition 205 © All Rights Reserved TreQna 2005
How to work with a Fishbone
Define your problem statement. It should start with “Why”, for example “Why is the
cycle time high?”
Use the basic categories as framework:
– People
– Materials
– Methods
– Machines
– Measurements
– Mother Nature
Make your core team understand the problem statement and objective of exercise
Cause and Effect diagram is a team tool and so it is important to have members in your team who understand the process well.
Treqna Base Manual
First Edition 206 © All Rights Reserved TreQna 2005
Building the Diagram– Brainstorm possible causes and attach them to appropriate categories– Relate each cause with the problem statement– Try to bucket small related causes and prioritize
The Structure of Diagram:
P u t in y o u rQ u e s t io nh e r e
E n v ir o n m e n t
M e a su r e m e n ts
M e th o d s
M a te r ia l
M a c h in e s
P e r s o n n e l
S a m p l e F i s h b o n e
P u t in y o u rQ u e s t io nh e r e
E n v ir o n m e n t
M e a su r e m e n ts
M e th o d s
M a te r ia l
M a c h in e s
P e r s o n n e l
S a m p l e F i s h b o n e
How to work with a Fishbone
The attached diagram is an example to the structure, which once filled gives complete picture of problem and causes.
Treqna Base Manual
First Edition 207 © All Rights Reserved TreQna 2005
Prioritization Matrix
Few prioritization matrices which can be used are:• Criterion v/s criterion• QFD• Control/Impact
Why use it ?• To narrow down systematically to critical X’s by weighing each option w.r.t. its
importance and effect on Y
How does it help?• Helps six sigma team to focus on X’s on priority• Systemizes the approach of prioritizing as it based on on X’s weighed against each
other w.r.t. to there effect on Y • Quickly surfaces critical X’s from disagreements within the team
The matrix only prioritizes…the output still needs to be ratified with data and an objective outlook
Treqna Base Manual
First Edition 208 © All Rights Reserved TreQna 2005
Fishbone to Prioritization Matrix
Translate the Cause and Effect diagram to a Prioritization Matrix
• Items in Quadrant 1 are ‘Must do’• Items in Quadrant 2 should be dropped as they complicate processes without significant impact• Items in Quadrant 3 should only be implemented if Project Y does not show necessary improvements after
implementing items from Q1• Items in Quadrant 4 are NOT to be done
Treqna Base Manual
First Edition 209 © All Rights Reserved TreQna 2005
Pareto Chart
Use the tool:
To filter vital few from trivial many X’s affecting process Y.
A Pareto is very handy when:
– You want to prioritize which causes to eliminate first
– You want to display information objectively to others
Treqna Base Manual
First Edition 210 © All Rights Reserved TreQna 2005
The Pareto Principle
The 80:20 rule
“ 20% of the causes account for 80% of the effect”
The 80:20 rule originated from Vilfredo Pareto, an Italian economist who studied the distribution of wealth in a variety of countries around 1900. He discovered a common phenomenon: about 80% of the wealth in most countries was controlled by a consistent minority -- about 20% of the people. His observation eventually became known as either the "80:20 rule" or "Pareto's Principle".
While making Pareto the defects grouped by different segments depending on what makes sense to the business. Following are few segments: •By Cost•By Financial defects•By risk•By compliance/ legal errors
Treqna Base Manual
First Edition 211 © All Rights Reserved TreQna 2005
Pareto Chart - Minitab
1) Collect data
S o f t S k i l l s I s s u e P o o r K n o w l e d g e N o P r o b l e m R e s o l u t i o nP r o c e s s A 5 8 5 3 9 4 9 3 9 0P r o c e s s B 8 2 1 5 4 0 8 5 4 8P r o c e s s C 1 4 9 9 9 4 9 9P r o c e s s D 9 4 6 6 3 0 7 6 3 1P r o c e s s E 4 1 5 3 1 1 2 2 7 7
C a l l s M o n i t o r e dS o f t S k i l l s I s s u e P o o r K n o w l e d g e N o P r o b l e m R e s o l u t i o n
P r o c e s s A 5 8 5 3 9 4 9 3 9 0P r o c e s s B 8 2 1 5 4 0 8 5 4 8P r o c e s s C 1 4 9 9 9 4 9 9P r o c e s s D 9 4 6 6 3 0 7 6 3 1P r o c e s s E 4 1 5 3 1 1 2 2 7 7
C a l l s M o n i t o r e d
2). Total results and arrange data by Segmentation Factors
T o t a l d e f e c t sS o f t S k i l l s I s s u e 2 9 1 7P o o r K n o w l e d g e 1 9 2 7 6
N o P r o b l e m R e s o l u t i o n 2 0 4 5
T o t a l d e f e c t sS o f t S k i l l s I s s u e 2 9 1 7P o o r K n o w l e d g e 1 9 2 7 6
N o P r o b l e m R e s o l u t i o n 2 0 4 5
Following are the steps involved in making of pareto chart:
1. Data collection by different defect categories. While doing so we capture:Types or categories of defectsFrequency of defectsFor further analysis we might do a drill down on financial loss, risk, compliance defect and so on. This helps in building an approach towards problem
2. Put sum of frequencies of defects by category
Treqna Base Manual
First Edition 212 © All Rights Reserved TreQna 2005
Pareto
Tool Utility: Useful tool identify factors/segments which contribute to the majority of the effect
How: Minitab > Quality Tools > Pareto Chart
Treqna Base Manual
First Edition 213 © All Rights Reserved TreQna 2005
Pareto Minitab Inputs
Choose ‘Chart defects table’ – Labels in ‘Defects’ and Frequencies in ‘Counts’
Treqna Base Manual
First Edition 214 © All Rights Reserved TreQna 2005
Pareto Minitab Output
Treqna Base Manual
First Edition 215 © All Rights Reserved TreQna 2005
Process Map Analysis
Tool Utility:
Useful tools to approach process from customer perspective. Understand what does customer feel and what customer values. It solves two important purpose:
– Identification of problems and improvement opportunity in the process
– Sense of urgency to improve as it is seen from customer perspective
Types Of Analysis:
What does the customer feel?. “The moment of Truth”
Which parts of the services provided, customer is ready to pay? ”Value to customer”
Wait time in process
Further on the three key components to the process map analysis:•Moments of truth helps us on focusing on customer and this paradigm change in approach changes the comfort level at which the process is operating. With this approach we try to study the customer – process interaction. We also study the touch points of customer with the process and improve the experience.
•Value analysis is process of studying the value add activities and non-value add activities in the process. Every activity within the process is seen as whether it adds any value to customer or not.
•Wait time is related to the process workflow.
Treqna Base Manual
First Edition 216 © All Rights Reserved TreQna 2005
Value Analysis
Value-Added Work
Nonvalue-AddedWork
Value-EnablingWork
Process steps which are non essential but enable or support the Value-Adding Tasks to be done better/faster.
For every activity in the process ask the question
“Is Customer
Willing To Pay For Them?”
Answers to this question amount to Value added work which changes the product/ services physically and so is essential to the process.
This is part of process steps which is Non-Essential for production and delivery product/ service.
Customer Is not willing to pay for process step.
Notes on Value enable work:Value enable work is the category which confuses the most. The only distinguishing factor between non value add and value enable work comes from customer’s perspective where customer is definitely not ready to pay for non-value added work.
Given the current condition of any given process we might not be able to take out the value enable steps from the process. They aid in delivery of product and services.
Indicators of non value add in the process:Approval process involves multiple departmentsToo many supervisorsMultiple reporting
Typical non-value add is rework. Which is a process step repeated, a step which usually takes you back in the process. E.g. Defect caught in downstream step being sent to upstream step in process.
Treqna Base Manual
First Edition 217 © All Rights Reserved TreQna 2005
Types Of Non-Value-Added Work
SetupInitialization
InventoryMovement
Inspection(Quality Check)
Failure(Internal/External)
Wait / Delay
Process failures:• Failures happen two ways, Internal and External. Internal refers to defect found during processing due to upstream process and being corrected. E.g. Rework• External failure are the defects reported by customer.
Inspection(Quality Check):• Internal process step dedicated to check products/ services for defects.
Wait/ Delay:• Products/ services waiting or queuing to get processed in front of process step. It is inventory building in front of a process step. It is usually due to bottlenecks. E.g. Backlogs
Setup / Initialization:• Steps involved in setup or preparation of subsequent step. E.g. Changing settings of machine in manufacturing plant
Inventory Movement:Physical movement of inventory from one point of processing to other point of processing.
Treqna Base Manual
First Edition 218 © All Rights Reserved TreQna 2005
Workflow and Time allocation
Processing Time (Product Touch Time)Wait / Delay Time
Cycle Time
T1 T2
T2+T2
While analyzing workflow for a process one needs to go at product/service unit level and track the flow of that identity. This is very detailed process map with time allocation with each small step. In the process we categorize process step into value add, non value add and value enabling.
Concept of time allocation in process:Cycle TimeTotal time spent from start point of process to end point of process. It is also important to look at this metric from customer point of view. E.g in call centre the average wait time for call might be 60 seconds but customer might be going through longer wait due to IVR, which we don’t capture.
Process TimeIt is the actual spent on any unit of service or product which goes through the whole process. (Note: this also includes steps which are non value add).
Wait / Delay Time:Total idle time spent by any product or service while going through process steps.
Treqna Base Manual
First Edition 219 © All Rights Reserved TreQna 2005
Workflow: Disconnects in Process Steps
Unclear work responsibility
Unclear processing requirements
Redundancies
Tricky handoffs between process steps
Conflict in job goals
Process disconnects more often than not lead to delay and wait time in the process. It also leads to rework or resource under utilization.Most of the disconnects arise due to following reasons:
Unclear work responsibility: Happens when a process step is not owned by a particular person. The processing such environment happens in a very ad hoc manner leading to delay.
Unclear processing Requirements: Operational definitions for process step do not exist. This might lead to rework due to defects caught in later process steps.
Redundancies: When one process step is duplicated amount two or multiple processors. This happens when one processor is unaware of other processor working.
Tricky Handoffs: Work transfer from one person or step to another step or person without any operational definitions in place. Leading to delays or rework.
Conflict in Job goals: When the job goals are overlapping or do not cover complete processing involved leading to conflict or incomplete work.
Treqna Base Manual
First Edition 220 © All Rights Reserved TreQna 2005
Analyzing Workflow: “Track the unit flow”
Agent fills app with customer
Agent fills app with customer
Agent mails app using USPS
Agent mails app using USPS
App received in mail room
App received in mail room
Mail room sorts by department
Mail room sorts by department
Data entry apps placed in data
entry tub
Data entry apps placed in data
entry tub
Apps transferred to data entry
every two hours
Apps transferred to data entry
every two hours
Data entry operator screens app for completeness
Data entry operator screens app for completeness
Is app complete ?
Route to Resolution team
Route to Resolution team
Procure missing information from
agent
Procure missing information from
agent
Complete data entry and forward to audit
team
Complete data entry and forward to audit
team
Is information
entered correct ?
Correct informationCorrect
information
Route to underwriterRoute to
underwriter
Can we issue policy ?
Issue policyIssue policy
Decline policyDecline policy
Order printOrder print
Sort issues and declines
Sort issues and declines
Create policy package for Issued policy and letter for
declined policies
Create policy package for Issued policy and letter for
declined policies
Apply postage and mail to agent using SPS
Apply postage and mail to agent using SPS
Agent receives letter / package and notifies clientAgent receives letter /
package and notifies client
Yes
Yes
Yes
No
No
No
Cycle time is the total time taken from the point at which the customer requests a good or service until the good or service is delivered to the customer.The components of cycle time are processing time, inventory movement time, inspection time,wait time. Cycle time metric should always be seen with caution. Lot of times we might just look at one part of process which might be working at its best. But when the customer might think otherwise as you might have another process step having lot of delays. The customer always looks at total cycle time.
Treqna Base Manual
First Edition 221 © All Rights Reserved TreQna 2005
Value Analysis
Process Step 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 Total Time % TotalTime Taken (min) 2 2 10 21 120 3 4 5 3 7 9 12 1 2 3 21 240 10 475 100%Value Add 24 5%Value enabling 9 2%Rework 6 1%Transportation 261 55%Inventory / queue 147 31%Audit / Inspection 24 5%Prep / Set up 4 1%
Process Step 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 Total Time % TotalTime Taken (min) 2 2 10 21 120 3 4 5 3 7 9 12 1 2 3 21 240 10 475 100%Value Add 24 5%Value enabling 9 2%Rework 6 1%Transportation 261 55%Inventory / queue 147 31%Audit / Inspection 24 5%Prep / Set up 4 1%
Usually, a majority of the time is spent in Non Value add or Value Enabling work
Value analysis breaks the process into small steps categorized into value add and non value add. When we put our cycle time analysis against value analysis it gives tremendous insight into improvement opportunities. It tells exactly where we spent maximum time in process and whether or not it is value added work.
It is an eye opener exercise to understand the process, we might come the data where non-value added work adds to maximum in cycle timeNote: We might come across a situation where for a process step we may not have any data available. In such a scenario it is best to estimate time than skip the step.
Conclusion: Example4 steps (<25% of total steps) provide 100% value and take 24 minutes or 5% of total cycle time.14 steps (>75% of steps) are non-value-added, and consume 95% of cycle time.
How to Interpret the Matrix:
Look for the longest cycle time component. Is it a value-added step? If not, can you still reduce time?
Look for frequent category of non-value add steps in your process?. They might be driven due to things outside
process.
Look for overall breakup of cycle time into value add, value enable, and non-value add. This might give overview of
improvement opportunities.
Treqna Base Manual
First Edition 222 © All Rights Reserved TreQna 2005
Histogram
Tool Utility:Study variation in process. Gives visual display of data distribution as bar graph (frequency graph). Useful tool to determine if distribution is centered around target, and how is the variation against the customer specification limits.
How:Minitab.Histogram can also be made using excel sheet. In order to do so:- create a frequency table with or without defined classes(e.g. age 10-20, or without classes 10, 11 etc.)- Plot the frequency against individual X values or Classes on X-axis.
1 0 09 59 08 58 07 57 06 56 05 5
1 0
5
0
T e s t G r a d e s
# of
Stu
dent
s
1 0 09 59 08 58 07 57 06 56 05 5
1 0
5
0
T e s t G r a d e s
# of
Stu
dent
s
Treqna Base Manual
First Edition 223 © All Rights Reserved TreQna 2005
Histogram – Minitab
MINITAB FILE: DescriptiveStats.mtwClick on Graph > Histograms
1.Click on ‘Simple’2.Click Ok
Treqna Base Manual
First Edition 224 © All Rights Reserved TreQna 2005
Histogram is the graphical representation of the distribution of population plotted. It helps understanding following for any given population:
• Shape: Look for the distribution of bars. Are there multiple modes in the distribution?. This might give us idea if the given data is mix of two or more different population sets.
• Center: Where is the distribution centered?. Is it off the process target?.• Spread: How is the spread of bars in histogram?. Is the spread large compare to
requirement?(look against the specification limits). Note that spread has got nothing to do with Normality of data, as you might have normal data with very large spread when compared against specification limits from customer.
The options at the bottom of histogram input window gives option to put specification limits.
Histogram Output
Treqna Base Manual
First Edition 225 © All Rights Reserved TreQna 2005
Dot Plot
Tool Utility: Another tool to study variation in process. Also used for quick comparison of two or more groups for variation. As a part of descriptive statistical analysis used in initial stages.
How: Use Minitab GRAPH > DOTPLOTClick on ‘Simple’, Click OkDouble Click on C1, Click ok
Treqna Base Manual
First Edition 226 © All Rights Reserved TreQna 2005
Dot plot can be plotted from GRAPHS> DOTPLOTIn Minitab Release 14 the Dotplot from Graphs menu gives multiple option for Single Y, Multiple Y, Single group, Multiple group and there combination.
Dot Plot - Minitab
MINITAB FILE: DescriptiveStats.mtw
Treqna Base Manual
First Edition 227 © All Rights Reserved TreQna 2005
When Dotplot plotted for multiple operators (groups) gives us comparative visual representation of their:•Variation•Center• and spread
Dot Plot Output
Treqna Base Manual
First Edition 228 © All Rights Reserved TreQna 2005
Box And Whisker Plot
Tool Utility: Useful tool to compare two or more groups on variation and median. The tool plots:
- Quartiles- Median- Outliers, if any
The box tells about the spread of data and is a visual used for comparing two or more data groups.
How: Minitab > Graph > Boxplot
Treqna Base Manual
First Edition 229 © All Rights Reserved TreQna 2005
Box plot - How to Read ?
Maximum Observation that falls within the upper limit = Q3 + 1.5 (Q3 - Q1)
75th Percentile (Q3)
Median (50th Percentile)
25th Percentile (Q1)
Minimum Observation that falls within the lower limit = Q1 - 1.5 (Q3 - Q1)
Outlierany point outside the lower or upper limit*
Treqna Base Manual
First Edition 230 © All Rights Reserved TreQna 2005
Box plot - Minitab
MINITAB FILE: BoxPlots.mtwMinitab > Graph > Boxplot
• Choose ‘One Y’ ‘With Groups’, Click Ok.
• Use C6 for Data,• C7 for Subscripts• Click ok
Treqna Base Manual
First Edition 231 © All Rights Reserved TreQna 2005
Statistical Hypothesis Testing
Treqna Base Manual
First Edition 232 © All Rights Reserved TreQna 2005
The Nature Of Statistical Problems
DesiredProcess Current
Process
LSL USL
Poor Precision, Focus on reducing Variation
T
Poor Accuracy, Focus on Centering the process
DesiredProcess
CurrentProcess
LSL
T
USL
While working on Six Sigma Projects we come across two types of problems with data. •Variation (Spread)•CenteringThough working on one problem has some impact on other also, more often than not we would target only one problem at a time.
In Analyze phase we run statistical tests to determine which X’s have an effect on the Y. These tests may be testing for variance or for mean depending on what is our Y metric.
Treqna Base Manual
First Edition 233 © All Rights Reserved TreQna 2005
A Statistical Hypothesis
Definition of Hypothesis: Statement about the parameters of the Population
In hypothesis testing there are two hypotheses of interest.
– The null hypothesis (H0)
– The alternative hypothesis (HA)
Treqna Base Manual
First Edition 234 © All Rights Reserved TreQna 2005
Concept Of Hypothesis Testing
1. All processes have variation.
2. Different Samples from the same process may vary
Are these Samples from the same distribution or are they truly from different processes.
3. Hypothesis testing differentiates between sampling error (Precision) and true process differences.
Treqna Base Manual
First Edition 235 © All Rights Reserved TreQna 2005
Why Do Hypothesis Testing?
Identification and validation of factors which impact process performance.
Validate improvement in process
Statistically validate our statement about the characteristics of population
Helps in decision making process. Enabler for data driven approach.
Treqna Base Manual
First Edition 236 © All Rights Reserved TreQna 2005
Hypothesis Testing
In Hypothesis Testing one has to make a decision:– to either to accept null hypothesis or– to reject null hypothesis
The hypotheses are always statements about the population parameters.
To define a statistical Test we– Choose a statistic (called the test statistic)– State your null and alternate hypothesis– Divide the range of possible values for the test statistic into two parts
• The Acceptance Region• The Critical Region
•The Acceptance Region (The range of values of the test statistic that indicate the Null Hypothesis is true.)• The Critical Region (The range of values of the test statistic that indicate the Null Hypothesis is false.)
Treqna Base Manual
First Edition 237 © All Rights Reserved TreQna 2005
Nature of Ho and Ha
Null Hypothesis (Ho ):
Null hypothesis always represents Status Quo, No difference
True unless proven otherwise
Represented by = or > or <
Alternative Hypothesis (Ha ):
It is the conclusion we are trying to make from data
Signs used in Minitab:≠ or < or >
Treqna Base Manual
First Edition 238 © All Rights Reserved TreQna 2005
Criminal Trial Analogy
• In the court defendant is always treated as innocent until proven guilty.
• The hypothesis than is:
• Ho: Person is not guilty.
• Ha: Person is guilty.
• There after once all the evidence(data) is collected we make a decision based on it:
• If there is sufficient evidence(statistically significant) we reject null hypothesis and
accept alternate(defendant is guilty).
• If there is not enough evidence we accept null hypothesis (defendant is innocent)
Hypotheses are statements about population parameters. Relating to the previous example about the height of people from two different countries, we could state: Ho: μB ≤ μA Ha: μB > μA
The statement is about the population means, not the sample means.
Treqna Base Manual
First Edition 239 © All Rights Reserved TreQna 2005
TruthTruth
Ho Ha
Ver
dict
Ver
dict
Ho
HaInnocent,
Jailed
Guilty,Set Free
Innocent,Set Free
Guilty,Jailed
Innocent Guilty
Set
Free
Jaile
d
Based on the evidence there are four possible outcomes of decision we make:
Decision Error
Treqna Base Manual
First Edition 240 © All Rights Reserved TreQna 2005
• 1 - α = Probability of correctly accepting the null hypothesis (detecting no change when there is none).
• 1 - β = Probability of correctly accepting the alternative hypothesis (detecting a change when there is one).
• It is not possible to commit a Type I and Type II decision error simultaneously
Decision Error - Types
TruthTruth
Ho Ha
Ver
dict
Ver
dict Ho
HaType I Error
α
Type II Errorβ
Correct Decision1 - α = Confidence
Correct Decision1 - β = Power
Innocent Guilty
Set
Free
Jaile
d
For any statistical testing procedure define1. α = P[Rejecting the null hypothesis when it is true] = P[ type I error]2. β = P[accepting the null hypothesis when it is false] = P[ type II error]The symbol α is called the level of significance. Usually we have α set at .05, which means we have 95%(1- α) confidence in accepting null hypothesis.The Acceptance Region and The Critical Region are chosen form underneath the sampling distribution of the test statistic when H0 is true.The Critical Region lies in the tails of sampling distribution of the test statistic when H0 is true.
Treqna Base Manual
First Edition 241 © All Rights Reserved TreQna 2005
Types of HypothesisOne Sample Hypothesis
1. Ho : μ = constant = T
Ha : μ ≠ constant = T
2. Ho : σ2 = constant
Ha : σ2 ≠ constant
Ho Ha
T
In this and following slides read:μ as mean and σ as standard deviation
Treqna Base Manual
First Edition 242 © All Rights Reserved TreQna 2005
Types of HypothesisTwo Sample Hypotheses
3. Ho : μ1 = μ2
Ha : μ1≠ μ2
4. Ho : μ1 ≤ μ2
Ha : μ1 > μ2
5. Ho : σ12 = σ2
2
Ha : σ12 ≠ σ2
2
μ1 μ2
σ2
σ1
Treqna Base Manual
First Edition 243 © All Rights Reserved TreQna 2005
Types of HypothesisMulti Sample Hypotheses
6. Ho : μ1 = μ2 = . . . = μn
Ha : at least one not equal
7. Ho : σ12 = σ2
2 = . . . = σn2
Ha : at least one not equal
Treqna Base Manual
First Edition 244 © All Rights Reserved TreQna 2005
4 5 6 7 8 9 10 11 12 13 14 15 16
4 5 6 7 8 9 10 11 12 13 14 15 16
α/2 α/2
α/2
μ = 10
μ = 10
Acceptance Region1 - α
Acceptance Region1 - α
One-Sided Test
Ex. Ho : μ1 < 10Ha : μ1 > 10
Two-Sided Test
Ex. Ho : μ1 = 10Ha : μ1 ≠ 10
One-sided & Two-sided Hypothesis Tests
The symbols used for two-sided tests are = and ≠. The symbols used for one-sided tests are ≥, ≤, < and >
Treqna Base Manual
First Edition 245 © All Rights Reserved TreQna 2005
Alpha denotes the probability of making a Type I error
P-value is the probability value based on which we make judgment on hypothesis.
Any p-value less than 0.05 means we reject the null hypothesis and accept the
alternate hypothesis
The P-Value
p < a: Reject Ho
p > a: Accept Ho
p < a: Reject Ho
p > a: Accept Ho
Treqna Base Manual
First Edition 246 © All Rights Reserved TreQna 2005
Hypothesis Tests Summary
Continuous Y and Discrete X
Test for MEDIAN- Moods Median Test – tests if medians from 2 or more sample means are equal
- HOV (Levene) – Compares 2 or more samples
Test for MEAN- 1-sample T test – tests if a sample means is equal to a known mean or target- 2-sample T test – tests if 2 sample means are equal- One way ANOVA – tests if 2 or more samples have equal means- Two way ANOVA – tests if means from samples classified by 2 categories are equal
- F test – Compares 2 samples- HOV (Levene) – Compares 2 or more samples- χ2 Test – Compares a sample variance to a known population variance
Central TendencyVariation
Test for MEDIAN- Moods Median Test – tests if medians from 2 or more sample means are equal
- HOV (Levene) – Compares 2 or more samples
Test for MEAN- 1-sample T test – tests if a sample means is equal to a known mean or target- 2-sample T test – tests if 2 sample means are equal- One way ANOVA – tests if 2 or more samples have equal means- Two way ANOVA – tests if means from samples classified by 2 categories are equal
- F test – Compares 2 samples- HOV (Levene) – Compares 2 or more samples- χ2 Test – Compares a sample variance to a known population variance
Central TendencyVariation
Y =
Nor
mal
Y =
Non
Nor
mal
Treqna Base Manual
First Edition 247 © All Rights Reserved TreQna 2005
Normal (z) Distribution
When sample size is infinite uncertainty prediction becomes very small, hence we can use Normal distribution for predictions.
df = ∞
σμXZ −=
μ
Sample size infinite
Treqna Base Manual
First Edition 248 © All Rights Reserved TreQna 2005
While working on project if our metric is related to controlling or decreasing spread, we might be interested to know if factor X has an impact on variance in Y.
To compare variance of two or multiple data samples we use Homogeneity of Variance
(HOV) Test.
The hypothesis for the test is:
HO = Variance for all the groups are equal
Ha = At least one group variance is different
HOV is also prerequisite to testing for group means. Some tools which are used to test
means of groups use pooled deviation and so HOV is a tool which used before we
decide which tool we will use before testing for means.
Homogeneity of Variance (HOV)
Treqna Base Manual
First Edition 249 © All Rights Reserved TreQna 2005
HOV - Minitab
If you wish to study variation across 2 or more Groups and you have established that each groups stable and normally distributed. Perform the Homogeneity of Variance Test.
Minitab File: HOV.mtw
Treqna Base Manual
First Edition 250 © All Rights Reserved TreQna 2005
HOV - Input
22
21a
22
21o
σ σ :H
σ σ :H
≠
=
.05 α =
Treqna Base Manual
First Edition 251 © All Rights Reserved TreQna 2005
In the output window the graph on left side represent confidence interval around standard deviation of each group.On the right hand side we get P-values for the test.The test results which we can read from left side window are:F-Test: We get this result only when there are two groups and normal data. For more than two groups and normal data F-test is replaced by Bartlett’s test.Bartlett’s Test: When your data is normal and have more than two groups under evaluation. Do not use this test result when your data is non- normal as Bartlett’s test is not robust to non-normality.Levene’s Test: Use this test result when data is continuous and not necessary normal.
HOV Test - Output
p < a: Reject Ho
p > a: Accept Ho
22
21a
22
21o
σ σ :H
σ σ :H
≠
=.05 α =
Treqna Base Manual
First Edition 252 © All Rights Reserved TreQna 2005
2 Sample t test
HO (Null Hypothesis) : μ1 = μ2
HA (Alternate Hypothesis) : μ1 = μ2
Treqna Base Manual
First Edition 253 © All Rights Reserved TreQna 2005
2 Sample t test - Minitab
Minitab File: HOV.mtw
Treqna Base Manual
First Edition 254 © All Rights Reserved TreQna 2005
Run Homogeneity of Variance test before you indicate to assume equal variances in the input window.
2 sample t - Input
Click ‘Options’ to change your Alternate Hypothesis
Check this box if HOV Indicates that there is no statistically significant difference in Variation among the groups being tested
Treqna Base Manual
First Edition 255 © All Rights Reserved TreQna 2005
2 Sample t -Output
Two-Sample T-Test and CI: Agent A, Agent B
Two-sample T for Agent A vs Agent B
N Mean StDev SE Mean
Agent A 40 70.20 2.19 0.35
Agent B 40 49.58 2.85 0.45
Difference = mu (Agent A) - mu (Agent B)
Estimate for difference: 20.6250
95% CI for difference: (19.4920, 21.7580)
T-Test of difference = 0 (vs not =): T-Value = 36.28
P-Value = 0.000 DF = 73
Accept HA (Alternate Hypothesis) : μ1 = μ2
The p value indicates that the two agents are statically different in their performance.
Treqna Base Manual
First Edition 256 © All Rights Reserved TreQna 2005
Two Sample T-test and One Way ANOVA test the same thing, however the difference comes from the fact that Two Sample T-test can analyze only two sample at a time and so running it formultiple groups is a logistical nightmare. Though Two Sample T-test gives us flexibility of not assuming equal variances
ANOVA v/s Regression:
Analysis of variance (ANOVA) is similar to regression in that it is used to investigate and model the relationship between a response variable and one or more independent variables. There are two ways ANOVA is different from Regression Analysis:•The independent variables are categorical•It doesn’t make any assumption about the the nature of relationship between Y and independent variable.
ANOVA stand for Analysis of Variance. Though the name is misleading
One Way ANOVA is used to compare means of two or more groups. The hypothesis
used is:
Ho: μ1 = μ2 = μ3 = μ4
Ha: At least one m different from the others
Since ANOVA uses pooled standard deviation it is essential to run HOV before this test.
In case HOV fails than one may use Two Sample T-Test.
One Way Analysis of Variance (ANOVA)
Treqna Base Manual
First Edition 257 © All Rights Reserved TreQna 2005
A Test to determine if the means of the number of Queries Resolved by 4 agents is statistically differentHo: μ1 = μ2 = μ3 = μ4Ha: at least one m different from the others
One Way ANOVA - Minitab
Minitab File: Anova.mtw
Treqna Base Manual
First Edition 258 © All Rights Reserved TreQna 2005
One-Way ANOVA – Input/ Output
One-way ANOVA: Average Queries Resolved versus Agent
Source DF SS MS F P
Agent 3 9998.6 3332.9 287.01 0.000
Error 196 2276.1 11.6
Total 199 12274.6
S = 3.408 R-Sq = 81.46% R-Sq(adj) = 81.17%
Individual 95% CIs For Mean Based on
Pooled StDev
Level N Mean StDev --------+---------+---------+---------+-
Agent A 50 19.940 2.234 (*-)
Agent B 50 29.908 2.779 (-*)
Agent C 50 39.874 4.049 (*-)
Agent D 50 28.596 4.164 (-*)
--------+---------+---------+---------+-
24.0 30.0 36.0 42.0
Pooled StDev = 3.408
An ANOVA output gives us:1. P-value for significance2. Confidence band to understand which ones are different
Treqna Base Manual
First Edition 259 © All Rights Reserved TreQna 2005
Non–Parametric Test - Objectives
Understand distribution - Normality Test
Possible reasons for not normal data
Understand interpretation of Mood’s Median Test
Understand HOV Test
Treqna Base Manual
First Edition 260 © All Rights Reserved TreQna 2005
Non-parametric Tests
What if I don’t have normal data?
When we don’t have Normal data we use non-parametric tests. Nonparametric
means that no assumption for distribution of data exist.
Tools used for analysis are:
Stability : Run Chart
Centre tendency : Mood’s Median Test
Spread (Variance) : HOV – Use Levine's test
Capability : Use DPMO to calculate sigma
It is common to have data which is not not normal. Care should be taken before reaching that conclusion.
Distribution should be check for multiple populationCheck for outliers
Normal data gives flexibility in terms of number of tools available for analysis. However sufficient tools are available for not normal data.
Treqna Base Manual
First Edition 261 © All Rights Reserved TreQna 2005
Non-parametric Tests
The term nonparametric means that no assumption is made about the distribution of
data.
– Non-parametric tests are used when we have no assumption for distribution
of the data to work with.
– Mood’s Median Test is a nonparametric test which is frequently used to
compare medians of two or more groups. One way ANOVA is the
counterpart of this test for Normal data.
Examples of non-normal data:
Cycle time data is usually non normal data. Usually a skewed distribution.
Metric data where only one sided specifications limits are defined.
Treqna Base Manual
First Edition 262 © All Rights Reserved TreQna 2005
Dealing With Non-Normal Data
Do some checks before concluding that data is non normal:
– Normality Test
– Look for resolution of data. The measurement System should be able to capture the unit of data in an adequate fashion.
– Investigate outliers(A word of caution: sometimes while taking out outliers we end up taking out a considerable amount of data from sample, this is not recommended, limit your removal of data to only a few data points).
– Check your sample size. Small sample size (<30 points) might not be able to predict population characteristics correctly.
– Attempt to transform the data. However it should not be used as a compulsion. Consult with MBB/BB before you use transformation. Common transforms include:
– finding the square root of all data points
– finding the log of all data points
– finding the square of all data points
– Box-cox Transformation
Care is required before concluding that data is non-normal. Primarily because more often than not data is supposed to be normal. However we still encounter situations when we come across non-normal data.Use the non-normal tests!
Treqna Base Manual
First Edition 263 © All Rights Reserved TreQna 2005
E.g. File: MoodsMedianStat > Nonparametrics > Mood’s Median Test
– Response = Average Call Handle Time– Factor = Team
Answer the following questions:1. Are the medians different? (interpret the p-value)2. Which medians are different? (interpret the confidence intervals)
Mood’s Median Test
HO = Medians for all the groups are equalHA = At least one median is different
Hypothesis for the test:
To perform a Mood’s Median Test on collected data:Enter Y data in one column in Minitab Second column should define subgroupsChoose test from Stat > Nonparametric > Mood’s Median Test
Treqna Base Manual
First Edition 264 © All Rights Reserved TreQna 2005
Mood’s Median Test
Select Stat > Nonparametrics > Moods Median Test
Treqna Base Manual
First Edition 265 © All Rights Reserved TreQna 2005
Mood’s Median Test – Input
Response data or Y
Factors or X’s
Treqna Base Manual
First Edition 266 © All Rights Reserved TreQna 2005
Mood’s Median Test - Output
Mood Median Test: AHT versus Team
Mood median test for AHT
Chi-Square = 26.32 DF = 2 P = 0.000
Individual 95.0% CIs
Team N<= N> Median Q3-Q1 -----+---------+---------+---------+-
A 7 8 280.0 14.0 (----*----)
B 15 0 263.0 10.0 (---*--)
C 1 14 296.0 13.0 (-----*--)
-----+---------+---------+---------+-
264 276 288 300
Overall median = 279.0
P value
Confidence interval
Mood’s Median test output gives a table where it defines by group number of data items below and above group’s median. It then performs a simple Chi-square test on this summary table. Based on the test it gives P value.If you would manually run Chi-square test on the table, you would get the same result.On the right side of table we get a visual of confidence band around group medians. When the test is significant you would see that one of the group confidence band far from others.
Treqna Base Manual
First Edition 267 © All Rights Reserved TreQna 2005
Hypothesis for various tests
Run Charts
Ho : There are no trends, cluster oscillations or mixtures
Ha : There exist trends, clusters, oscillations or mixtures
Normality Test
Ho : Data is normally distributed
Ha : Data is not normally distributed
Mood’s Median Test
Ho : X1 = X2 = X3 =…..= Xn
Ha : At least one median is different from the others
Homogeneity of variance
Ho : σ12 = σ2
2 =….. = σn2
Ha : σ12 ≠ σ2
2 ≠ ….. ≠ σn2
∼ ∼ ∼ ∼
Treqna Base Manual
First Edition 268 © All Rights Reserved TreQna 2005
1-sample t test
Ho : μ = T
Ha : μ ≠ T
2-sample t test
Ho : μ1 = μ2
Ha : μ1 ≠ μ2
Paired t test
Ho : Difference between the two means = 0
Ha : Difference between the two means ≠ 0
ANOVA
Ho : μ1 = μ2 = μ3 =….= μn
Ha : At least one mean is different from the others
Hypothesis for various tests
Treqna Base Manual
First Edition 269 © All Rights Reserved TreQna 2005
Continuous Y and X
Continuous Y and Continuous X
• Correlation : Tests Linear relationship between two variables
• Regression : Defines linear relationship between a dependent and an independent
variable
Correlation assumes that both variables can change while regression (typically) assumes that the independent variable is measured without error. A significant correlation means that 2 variables tend to go up or down together (or change in opposite directions), for whatever reason. A significant regression means that the dependent variable can be predicted by the independent variable, again, for whatever reason. Causality is not tested by either method. Though theoretically different, both tests usually give the same answer (significant or not).
Treqna Base Manual
First Edition 270 © All Rights Reserved TreQna 2005
E.g. File: Grades.MTWStat > Basic Stats > Correlation
Correlation - Minitab
Treqna Base Manual
First Edition 271 © All Rights Reserved TreQna 2005
Correlation – Inputs/Outputs
Treqna Base Manual
First Edition 272 © All Rights Reserved TreQna 2005
Regression Minitab
Treqna Base Manual
First Edition 273 © All Rights Reserved TreQna 2005
Regression Minitab Inputs
Treqna Base Manual
First Edition 274 © All Rights Reserved TreQna 2005
Regression Minitab Inputs
Click the ‘Graphs’ button
Treqna Base Manual
First Edition 275 © All Rights Reserved TreQna 2005
Regression Minitab Inputs
Click the ‘Results’ button
Treqna Base Manual
First Edition 276 © All Rights Reserved TreQna 2005
Regression Minitab Outputs
Regression Analysis: Score1 versus Score2
The regression equation isScore1 = - 4.67 + 4.40 Score2
Predictor Coef SE Coef T PConstant -4.6674 0.8572 -5.44 0.001Score2 4.3975 0.3514 12.51 0.000
S = 0.572711 R-Sq = 95.7% R-Sq(adj) = 95.1%
Analysis of Variance
Source DF SS MS F PRegression 1 51.353 51.353 156.56 0.000Residual Error 7 2.296 0.328Total 8 53.649
Treqna Base Manual
First Edition 277 © All Rights Reserved TreQna 2005
Discrete Y & Discrete X
Two Chi-Square Tests:o Goodness-of-Fit Test to test if observed data has expected distributiono Test of Independence to test if two or more proportions are different
• Chi-Square statistic is calculated using formula:
• Critical Value depends on two factors:o a - the confidenceo df – degrees of freedom for data
e
eon
i fff 2
1
⎟⎠⎞
⎜⎝⎛ −
= Σ=
2χ
Treqna Base Manual
First Edition 278 © All Rights Reserved TreQna 2005
Chi-Square Test–χ2
Tool Utility:
Chi-square tests enables us to test whether two or more population can be
considered equal.
There are two types of Chi-square χ2 tests:
- Goodness-of-Fit Test: To test if an observed set of data fits an expected
distribution of data.
- Test of Independence: To test if two or more proportions are associated (Belong
to similar distribution)
Treqna Base Manual
First Edition 279 © All Rights Reserved TreQna 2005
Calculation for Chi-square statistic is:
Where:n = number of categoriesfo = observed frequency of datafe = expected frequency of data
e
eon
i fff 2
1
⎟⎠⎞
⎜⎝⎛ −
= Σ=
2χ
Chi-Square Statistic
Treqna Base Manual
First Edition 280 © All Rights Reserved TreQna 2005
The chi-square uses “degrees of freedom” of data under scrutiny as a parameter while calculating Chi-Square critical value.
For any statistic being computed, the degrees of freedom of a sample are the number of values in the sample which are free to vary without changing the statistic being measured.
Degrees Of Freedom
Treqna Base Manual
First Edition 281 © All Rights Reserved TreQna 2005
The statistical test can be done both manually or using minitab. Steps involved in doing manual calculation are:
• Create frequency table and see if any of the frequency is less than 5(as we have mentioned earlier that avoid small frequency)
• Calculate Chi-square statistic using the formula
• Calculate critical value using the table. Critical value depends on two things, a -the confidence and degrees of freedom(df).
• If the chi-square statistic is less than the critical value in the table, the null hypothesis (good fit) is accepted. If not, the alternative hypothesis (that the data are not typical of the expected distribution) is accepted.
Chi Square Stat Test
A B
Critical Value
Area of Rejection
Chi-Square Distribution
P-value can also be calculated using formula in excel:CHIDIST(statistic,df)
Treqna Base Manual
First Edition 282 © All Rights Reserved TreQna 2005
Chi-Square Distribution
d f .2 5 0 .1 0 0 .0 5 0 .0 2 5 .0 1 0 .0 0 5 .0 0 11 1 .3 2 3 2 .7 0 6 3 .8 4 1 5 .0 2 4 6 .6 3 5 7 .8 7 9 1 0 .8 2 82 2 .7 7 3 4 .6 0 5 5 .9 9 1 7 .3 7 8 9 .2 1 0 1 0 .5 9 7 1 3 .8 1 63 4 .1 0 8 6 .2 5 1 7 .8 1 5 9 .3 4 8 1 1 .3 4 5 1 2 .8 3 8 1 6 .2 6 64 5 .3 8 5 7 .7 7 9 9 .4 8 8 1 1 .1 4 3 1 3 .2 7 7 1 4 .8 6 0 1 8 .4 6 75 6 .6 2 6 9 .2 3 6 1 1 .0 7 0 1 2 .8 3 2 1 5 .0 8 6 1 6 .7 5 0 2 0 .5 1 5
6 7 .8 4 1 1 0 .6 4 5 1 2 .5 9 2 1 4 .4 4 9 1 6 .8 1 2 1 8 .5 4 8 2 2 .4 5 87 9 .0 3 7 1 2 .0 1 7 1 4 .0 6 7 1 6 .0 1 3 1 8 .4 7 5 2 0 .2 7 8 2 4 .3 2 28 1 0 .2 1 9 1 3 .3 6 2 1 5 .5 0 7 1 7 .5 3 5 2 0 .0 9 0 2 1 .9 5 5 2 6 .1 2 59 1 1 .3 8 9 1 4 .6 8 4 1 6 .9 1 9 1 9 .0 2 3 2 1 .6 6 6 2 3 .5 8 9 2 7 .8 7 7
1 0 1 2 .5 4 9 1 5 .9 8 7 1 8 .3 0 7 2 0 .4 8 3 2 3 .2 0 9 2 5 .1 8 8 2 9 .5 8 8
1 1 1 3 .7 0 1 1 7 .2 7 5 1 9 .6 7 5 2 1 .9 2 0 2 4 .7 2 5 2 6 .7 5 7 3 1 .2 6 41 2 1 4 .8 4 5 1 8 .5 4 9 2 1 .0 2 6 2 3 .3 3 7 2 6 .2 1 7 2 8 .3 0 0 3 2 .9 0 91 3 1 5 .9 8 4 1 9 .8 1 2 2 2 .3 6 2 2 4 .7 3 6 2 7 .6 8 8 2 9 .8 1 9 3 4 .5 2 81 4 1 7 .1 1 7 2 1 .0 6 4 2 3 .6 8 5 2 6 .1 1 9 2 9 .1 4 1 3 1 .3 1 9 3 6 .1 2 31 5 1 8 .2 4 5 2 2 .3 0 7 2 4 .9 9 6 2 7 .4 8 8 3 0 .5 7 8 3 2 .8 0 1 3 7 .6 9 7
1 6 1 9 .3 6 9 2 3 .5 4 2 2 6 .2 9 6 2 8 .8 4 5 3 2 .0 0 0 3 4 .2 6 7 3 9 .2 5 21 7 2 0 .4 8 9 2 4 .7 6 9 2 7 .5 8 7 3 0 .1 9 1 3 3 .4 0 9 3 5 .7 1 8 4 0 .7 9 01 8 2 1 .6 0 5 2 5 .9 8 9 2 8 .8 6 9 3 1 .5 2 6 3 4 .8 0 5 3 7 .1 5 6 4 3 .3 1 21 9 2 2 .7 1 8 2 7 .2 0 4 3 0 .1 4 4 3 2 .8 5 2 3 6 .1 9 1 3 8 .5 8 2 4 3 .8 2 02 0 2 3 .8 2 8 2 8 .4 1 2 3 1 .4 1 0 3 4 .1 7 0 3 7 .5 6 6 3 9 .9 9 7 4 5 .3 1 5
2 1 2 4 .9 3 5 2 9 .6 1 5 3 2 .6 7 1 3 5 .4 7 9 3 8 .9 3 2 4 1 .4 0 1 4 6 .7 9 72 2 2 6 .0 3 9 3 0 .8 1 3 3 3 .9 2 4 3 6 .7 8 1 4 0 .2 8 9 4 2 .7 9 6 4 8 .2 6 82 3 2 7 .1 4 1 3 2 .0 0 7 3 5 .1 7 2 3 8 .0 7 6 4 1 .6 3 8 4 4 .1 8 1 4 9 .7 2 82 4 2 8 .2 4 1 3 3 .1 9 6 3 6 .4 1 5 3 9 .3 6 4 4 2 .9 8 0 4 5 .5 5 8 5 1 .1 7 92 5 2 9 .3 3 9 3 4 .3 8 2 3 7 .6 5 2 4 0 .6 4 6 4 4 .3 1 4 4 6 .9 2 8 5 2 .6 2 0
2 6 3 0 .4 3 4 3 5 .5 6 3 3 8 .8 8 5 4 1 .9 2 3 4 5 .6 4 2 4 8 .2 9 0 5 4 .0 5 22 7 3 1 .5 2 8 3 6 .7 4 1 4 0 .1 1 3 4 3 .1 9 4 4 6 .9 6 3 4 9 .6 4 5 5 5 .4 7 62 8 3 2 .6 2 0 3 7 .9 1 6 4 1 .3 3 7 4 4 .4 6 1 4 8 .2 7 8 5 0 .9 9 3 5 6 .8 9 22 9 3 3 .7 1 1 3 9 .0 8 7 4 2 .5 5 7 4 5 .7 2 2 4 9 .5 8 8 5 2 .3 3 6 5 8 .3 0 23 0 3 4 .8 0 0 4 0 .2 5 6 4 3 .7 7 3 4 6 .9 7 9 5 0 .8 9 2 5 3 .6 7 2 5 9 .7 0 3
4 0 4 5 .6 1 6 5 1 .8 0 5 5 5 .7 5 8 5 9 .3 4 2 6 3 .6 9 1 6 6 .7 6 6 7 3 .4 0 25 0 5 6 .3 3 4 6 3 .1 6 7 6 7 .5 0 5 7 1 .4 2 0 7 6 .1 5 4 7 9 .4 9 0 8 6 .6 6 16 0 6 6 .9 8 1 7 4 .3 9 7 7 9 .0 8 2 8 3 .2 9 8 8 8 .3 7 9 9 1 .9 5 2 9 9 .6 0 7
7 0 7 7 .5 7 7 8 5 .5 2 7 9 0 .5 3 1 9 5 .0 2 3 1 0 0 .4 2 5 1 0 4 .2 1 5 1 1 2 .3 1 78 0 8 8 .1 3 0 9 6 .5 7 8 1 0 1 .8 7 9 1 0 6 .6 2 9 1 1 2 .3 2 9 1 1 6 .3 2 1 1 2 4 .8 3 99 0 9 8 .6 5 0 1 0 7 .5 6 5 1 1 3 .1 4 5 1 1 8 .1 3 6 1 2 4 .1 1 6 1 2 8 .2 9 9 1 3 7 .2 0 8
1 0 0 1 0 9 .1 4 1 1 1 8 .4 9 8 1 2 4 .3 4 2 1 2 9 .5 6 1 1 3 5 .8 0 7 1 4 0 .1 6 9 1 4 9 .4 4 9
A lp h a
Treqna Base Manual
First Edition 283 © All Rights Reserved TreQna 2005
d f . 9 9 5 . 9 9 0 . 9 7 5 . 9 5 0 . 9 0 0 . 7 5 0 . 5 0 01 . 0 0 0 0 3 9 . 0 0 0 1 6 0 . 0 0 0 9 8 0 . 0 0 3 9 3 0 . 0 1 5 8 0 0 . 1 0 1 5 0 0 . 4 5 5 0 0 02 0 . 0 1 0 0 . 0 2 0 0 . 0 5 1 0 . 1 0 3 0 . 2 1 1 0 . 5 7 5 1 . 3 8 63 0 . 0 7 2 0 . 1 1 5 0 . 2 1 6 0 . 3 5 2 0 . 5 8 4 1 . 2 1 3 2 . 3 6 64 0 . 2 0 7 0 . 2 9 7 0 . 4 8 4 0 . 7 1 1 1 . 0 6 4 1 . 9 2 3 3 . 3 5 75 0 . 4 1 2 0 . 5 5 4 0 . 8 3 1 1 . 1 4 5 1 . 6 1 0 2 . 6 7 5 4 . 3 5 1
6 0 . 6 7 6 0 . 8 7 2 1 . 2 3 7 1 . 6 3 5 2 . 2 0 4 3 . 4 5 5 5 . 3 4 87 0 . 9 8 9 1 . 2 3 9 1 . 6 9 0 2 . 1 6 7 2 . 8 3 3 4 . 2 5 5 6 . 3 4 68 1 . 3 4 4 1 . 6 4 6 2 . 1 8 0 2 . 7 3 3 3 . 4 9 0 5 . 0 7 1 7 . 3 4 49 1 . 7 3 5 2 . 0 8 8 2 . 7 0 0 3 . 3 2 5 4 . 1 6 8 5 . 8 9 9 8 . 3 4 3
1 0 2 . 1 5 6 2 . 5 5 8 3 . 2 4 7 3 . 9 4 0 4 . 8 6 5 6 . 7 3 7 9 . 3 4 2
1 1 2 . 6 0 3 3 . 0 5 3 3 . 8 1 6 4 . 5 7 5 5 . 5 7 8 7 . 5 8 4 1 0 . 3 4 11 2 3 . 0 7 4 3 . 5 7 1 4 . 4 0 4 5 . 2 2 6 6 . 3 0 4 8 . 4 3 8 1 1 . 3 4 01 3 3 . 5 6 5 4 . 1 0 7 5 . 0 0 9 5 . 8 9 2 7 . 0 4 2 9 . 2 9 9 1 2 . 3 4 01 4 4 . 0 7 5 4 . 6 6 0 5 . 6 2 9 6 . 5 7 1 7 . 7 9 0 1 0 . 1 6 5 1 3 . 3 3 91 5 4 . 6 0 1 5 . 2 2 9 6 . 2 6 2 7 . 2 6 1 8 . 5 4 7 1 1 . 0 3 6 1 4 . 3 3 9
1 6 5 . 1 4 2 5 . 8 1 2 6 . 9 0 8 7 . 9 6 2 9 . 3 1 2 1 1 . 9 1 2 1 5 . 3 3 81 7 5 . 6 9 7 6 . 4 0 8 7 . 5 6 4 8 . 6 7 2 1 0 . 0 8 5 1 2 . 7 9 2 1 6 . 3 3 81 8 6 . 2 6 5 7 . 0 1 5 8 . 2 3 1 9 . 3 9 0 1 0 . 8 6 5 1 3 . 6 7 5 1 7 . 3 3 81 9 6 . 8 4 4 7 . 6 3 3 8 . 9 0 7 1 0 . 1 1 7 1 1 . 6 5 1 1 4 . 5 6 2 1 8 . 3 3 82 0 7 . 4 3 4 8 . 2 6 0 9 . 5 9 1 1 0 . 8 5 1 1 2 . 4 4 3 1 5 . 4 5 2 1 9 . 3 3 7
2 1 8 . 0 3 4 8 . 8 9 7 1 0 . 2 8 3 1 1 . 5 9 1 1 3 . 2 4 0 1 6 . 3 4 4 2 0 . 3 3 72 2 8 . 6 4 3 9 . 5 4 2 1 0 . 9 8 2 1 2 . 3 3 8 1 4 . 0 4 1 1 7 . 2 4 0 2 1 . 3 3 72 3 9 . 2 6 0 1 0 . 1 9 6 1 1 . 6 8 8 1 3 . 0 9 1 1 4 . 8 4 8 1 8 . 1 3 7 2 2 . 3 3 72 4 9 . 8 8 6 1 0 . 8 5 6 1 2 . 4 0 1 1 3 . 8 4 8 1 5 . 6 5 9 1 9 . 0 3 7 2 3 . 3 3 72 5 1 0 . 5 2 0 1 1 . 5 2 4 1 3 . 1 2 0 1 4 . 6 1 1 1 6 . 4 7 3 1 9 . 9 3 9 2 4 . 3 3 7
2 6 1 1 . 1 6 0 1 2 . 1 9 8 1 3 . 8 4 4 1 5 . 3 7 9 1 7 . 2 9 2 2 0 . 8 4 3 2 5 . 3 3 62 7 1 1 . 8 0 8 1 2 . 8 7 9 1 4 . 5 7 3 1 6 . 1 5 1 1 8 . 1 1 4 2 1 . 7 4 9 2 6 . 3 3 62 8 1 2 . 4 6 1 1 3 . 5 6 5 1 5 . 3 0 8 1 6 . 9 2 8 1 8 . 9 3 9 2 2 . 6 5 7 2 7 . 3 3 62 9 1 3 . 1 2 1 1 4 . 2 5 6 1 6 . 0 4 7 1 7 . 7 0 8 1 9 . 7 6 8 2 3 . 5 6 7 2 8 . 3 3 63 0 1 3 . 7 8 7 1 4 . 9 5 3 1 6 . 7 9 1 1 8 . 4 9 3 2 0 . 5 9 9 2 4 . 4 7 8 2 9 . 3 3 6
4 0 2 0 . 7 0 7 2 2 . 1 6 4 2 4 . 4 3 3 2 6 . 5 0 9 2 9 . 0 5 1 3 3 . 6 6 0 3 9 . 3 3 55 0 2 7 . 9 9 1 2 9 . 7 0 7 3 2 . 3 5 7 3 4 . 7 6 4 3 7 . 6 8 9 4 2 . 9 4 2 4 9 . 3 3 56 0 3 5 . 5 3 5 3 7 . 4 8 5 4 0 . 4 8 2 4 3 . 1 8 8 4 6 . 4 5 9 5 2 . 2 9 4 5 9 . 3 3 5
7 0 4 3 . 2 7 5 4 5 . 4 4 2 4 8 . 7 5 8 5 1 . 7 3 9 5 5 . 3 2 9 6 1 . 6 9 8 6 9 . 3 3 48 0 5 1 . 1 7 2 5 3 . 5 4 0 5 7 . 1 5 3 6 0 . 3 9 1 6 4 . 2 7 8 7 1 . 1 4 5 7 9 . 3 3 49 0 5 9 . 1 9 6 6 1 . 7 5 4 6 5 . 6 4 7 6 9 . 1 2 6 7 3 . 2 9 1 8 0 . 6 2 5 8 9 . 3 3 4
1 0 0 6 7 . 3 2 8 7 0 . 0 6 5 7 4 . 2 2 2 7 7 . 9 2 9 8 2 . 3 5 8 9 0 . 1 3 3 9 9 . 3 3 4
A l p h a
Chi-Square Distribution
Treqna Base Manual
First Edition 284 © All Rights Reserved TreQna 2005
This Test is used to check independence of two or more proportions. The hypothesis used is:H0 = all proportions are equalHa = at least one proportion is different from rest
Again as a caution frequency less than 5 not to be used for the test.
To do the test we use Contingency tables. The rows and the columns define the classification of data.
Following is an example of contingency table:
Degrees of Freedom:
Degrees of freedom for the test are calculated using formula:
(M-1)*(N-1)
Where M and N are number of columns and rows respectively.
Test of Independence
fo = 37
fo = 63
fo = 113
fo = 167
Leads
Dead Ends
Tele sales Team A
Total = 150
Total = 230
Total = 100 Total = 280
Tele sales Team A
Treqna Base Manual
First Edition 285 © All Rights Reserved TreQna 2005
The test of Independence includes calculating Chi-Square statistic and Critical value from chi-square table using degrees of freedom and confidence.However, doing the test on minitab is easier:
Test of Independence – Minitab Input
Treqna Base Manual
First Edition 286 © All Rights Reserved TreQna 2005
Test of IndependenceMinitab Input And Results
Ho: No correlation between team and leads generated.
Ha: Correlation between team and leads generated
Treqna Base Manual
First Edition 287 © All Rights Reserved TreQna 2005
Tool picking guide
Treqna Base Manual
First Edition 288 © All Rights Reserved TreQna 2005
Statistical Test Choices
Treqna Base Manual
First Edition 289 © All Rights Reserved TreQna 2005
DEFINE
MEASURE
ANALYZE
IMPROVE
CONTROL
Map project
CTQ Characteristics and standards
Measurement System Analysis
Baseline process
Screen for vital X’s
Defined Improve Process
MSA on X’s
Approve project
Data collection
Performance objective
Identify drivers of variation
Study interaction between X’s
Improved Process capability
Establish control plan
Treqna Base Manual
First Edition 290 © All Rights Reserved TreQna 2005
Improve Phase Objectives
• Develop and propose a solution
• Identify Vital X’s
• Develop transfer function
• Experiment with possibilities and develop solution
• Confirm that proposed solution will improve the process
• Pilot solution with small scale tests in real business environment
• Run statistical tests on the results to confirm the solution
• Identify and deploy resources to implement the solution
• Deploy solution on the process
Treqna Base Manual
First Edition 291 © All Rights Reserved TreQna 2005
Experiment with various solutions
Develop a transfer functions
Identify Vital Xs
Activities
Control Impact matrix, Regression
Interaction between Xs, Transfer function
Study Interaction between Xs
Step I.2
Tools
Deliverables
Objective
Step I.3Step I.1
FMEA, Process Map, Pareto, Fishbone analysis
Pilot SolutionVital Xs
Define Improved ProcessScreen for Vital Xs
Improve Phase
DEFINE MEASURE ANALYZE IMPROVE CONTROL
Treqna Base Manual
First Edition 292 © All Rights Reserved TreQna 2005
Characterization of Xs
Vital X’s or key elements which impact project Y have following characteristics:
• Independent variables
• Can be set at different levels to manipulate the response variable or dependent variable Y.
• Variation in independent variable contributes to variation in dependent variable Y
• Different X’s will have different impact on Y depending on the transfer function
• May be continuous and/or discrete
• Independent variables might not be measurable (E.g. alternative work-flow sequences)
Treqna Base Manual
First Edition 293 © All Rights Reserved TreQna 2005
Screening of X’s is done for various reasons:
• Focus on few factors which make most impact
• Ensure best utilization of resources available for the project
Tools Used:
• FMEA
• Pareto
• Fishbone analysis
• Screening DOE if advanced cases when X’s have interaction
Screening of Xs
Treqna Base Manual
First Edition 294 © All Rights Reserved TreQna 2005
Project Metric: Errors in Application form
Analyze result gave us seven factors which are significant. Now the pareto chart can be used to prioritize or screen further these factors. The output clearly says that Name, DOB and Country field contribute around 80% errors. So working on rest of the factors might not be as beneficial.
Using Pareto
Treqna Base Manual
First Edition 295 © All Rights Reserved TreQna 2005
When all the significant factors are available from Analyze step A3. One might use
FMEA to evaluate corresponding risk associated i.e. Risk priority number.
E.g.
On insurance application form where premium calculation is based on age risk
associated with error in DOB field would be more than error associated with wrong
address as it might result in financial loss to the insurance company.
Using FMEA
Treqna Base Manual
First Edition 296 © All Rights Reserved TreQna 2005
Experiment with various solutions
Develop a transfer functions
Identify Vital Xs
Activities
Control Impact matrix, Regression
Interaction between Xs, Transfer function
Study Interaction between Xs
Step I.2
Tools
Deliverables
Objective
Step I.3Step I.1
FMEA, Process Map, Pareto, Fishbone analysis
Pilot SolutionVital Xs
Define Improved ProcessScreen for Vital Xs
Improve Phase
DEFINE MEASURE ANALYZE IMPROVE CONTROL
Treqna Base Manual
First Edition 297 © All Rights Reserved TreQna 2005
Problem Solving Process
Practical Problem
Statistical Problem
Statistical Solution
Practical Solution
Transfer Function Y=F(X)
Treqna Base Manual
First Edition 298 © All Rights Reserved TreQna 2005
Transfer Function
The goal of improve is to develop a solution using the transfer function
Y = {X1, X2, X3…Xn}
The transfer function:
• Relates the vital X’s to the project Y
• Predicts the effect and direction of changes on Y due to changes in X’s
• Helps in fitting a solution which maximizes the effect on Y using all or few of the X’s
Treqna Base Manual
First Edition 299 © All Rights Reserved TreQna 2005
The goal of improve is to develop a solution using the transfer function
Y = {X1, X2, X3…Xn}
Tools that give mathematical transfer function:
• DOE
• Regression
• GLM (General Linear Model)
Transfer Function
Treqna Base Manual
First Edition 300 © All Rights Reserved TreQna 2005
Using the Interaction Matrix – In the following example training has high and positive impact on call quality, while length of the call has low but negative effect on the call quality.
In absence of a transfer function it is a very handy tool to get insight into development of possible solution
Interaction Matrix
High
Low
Influence on Call Quality (Project Y) + -
Subject Knowledge
Training
Length of the call
Finance Background
Treqna Base Manual
First Edition 301 © All Rights Reserved TreQna 2005
Experiment with various solutions
Develop a transfer functions
Identify Vital Xs
Activities
Control Impact matrix, Regression
Interaction between Xs, Transfer function
Study Interaction between Xs
Step I.2
Tools
Deliverables
Objective
Step I.3Step I.1
FMEA, Process Map, Pareto, Fishbone analysis
Pilot SolutionVital Xs
Define Improved ProcessScreen for Vital Xs
Improve Phase
DEFINE MEASURE ANALYZE IMPROVE CONTROL
Treqna Base Manual
First Edition 302 © All Rights Reserved TreQna 2005
Tolerance means the allowed variation in a variable
Once we have a a transfer function:
Y = F (X)
We would like to know for given allowed variation in Y what is the variation one would expect for X’s
This process of identifying variation limits for X’s using LSL and USL for Y is called Statistical Tolerancing
Tolerancing
Treqna Base Manual
First Edition 303 © All Rights Reserved TreQna 2005
Assume given transfer function Y= aX + b
Concept of Tolerancing
Y
X
Y = aX + b
USL
LSL
Tolerance for Y
Tolerance for X
The concept can be extrapolated to transfer function involving two X’s by imagining a three dimensions (three axes)
Treqna Base Manual
First Edition 304 © All Rights Reserved TreQna 2005
• Set Tolerance based on customer VOC. E.g. Specification limits defined by
customer
• When multiple X’s are involved. Set Tolerance for all simultaneously, specially when
the interaction between X’s are involved
• Adjust tolerance for measurement variation
More on Tolerancing
Treqna Base Manual
First Edition 305 © All Rights Reserved TreQna 2005
This step may be broken into three steps:
• Ideas generation
• Solution design
• Pilot solution
Solution Identification
Treqna Base Manual
First Edition 306 © All Rights Reserved TreQna 2005
By this step we have established:
• Vital X’s – prioritized
• Transfer function – Either we have a mathematical transfer function or we know how are X’s related to Y, i.e. what should we do to better Y – increase X or decrease X and will it make more impact on Y than any other X ?
• Tolerancing – Based on the transfer function we know how much variation we can allow for each of the X’s
Ideas Generation
While transfer function can exactly tells us the optimum setting for X’s, in its absence one has fair idea about what needs to be done with X’s in order to influence Y positively.
Treqna Base Manual
First Edition 307 © All Rights Reserved TreQna 2005
The project team and process experts can together participate in a team effort to generate ideas for solution.
Following may be used for generating ideas:
• Brainstorming – team tool for generating lots of ideas
• Problem Analogy – use an appropriate analogy for the current problem and try to generate solution for the same
• Best Practices – team might also look for similar processes and make a list of best practices which might be useful
Ideas Generation
Treqna Base Manual
First Edition 308 © All Rights Reserved TreQna 2005
All the ideas and best practices should be than ranked based on following factors in the same priority:
• Impact on Y
• Resource Requirement for implementation
Select ideas which would have maximum impact on Y and minimum resource requirement
Solution Design Selecting the best Ideas
A solution could be:
• One solution
• Set of solution, addressing different X’s or supplementing each other
Treqna Base Manual
First Edition 309 © All Rights Reserved TreQna 2005
The Pilot – Proof of Concept
Pilot is a process in which a solution is tested on existing process with real business environment. The proof of concept – it is important to prove that a hypothetical solution actually delivers what is expected. Helps surface challenges in full scale implementation
The objective of running a pilot is to:
• Analyze the effects of your solution on the process and plan for a successful full-scale implementation
• Establish that the solution would achieve what is desired.
• It is the proof of concept and can help in getting buy in from project champion for resource mobilization required for full scale implementation.
• To identify if there would be any challenges in full scale implementation of solution.
Treqna Base Manual
First Edition 310 © All Rights Reserved TreQna 2005
Running a Pilot
• Prepare a plan to execute pilot
• Do risk assessment of doing the pilot, use FMEA
• Discuss and establish test population, resources, location, duration and timing.
• Create data collection plan
• Pilot solution and collect data
• Analyze data and establish if the solution meets expectation using statistical test
• Identify resources required for full scale implementation of the solution
Treqna Base Manual
First Edition 311 © All Rights Reserved TreQna 2005
Improve Summary
• Vital X’s – select few which make big impact on Y
• Tools used are Pareto, FMEA, Fishbone Analysis
• Transfer function – what can one do in absence of mathematical transfer function
• Statistical Tolerancing – translating allowed variation in Y to X’s
• Solution Design
• Generating Ideas, prioritizing ideas
• Pilot solution – steps in running of a pilot
Treqna Base Manual
First Edition 312 © All Rights Reserved TreQna 2005
DEFINE
MEASURE
ANALYZE
IMPROVE
CONTROL
Map project
CTQ Characteristics and standards
Measurement System Analysis
Baseline process
Screen for vital X’s
Defined Improve Process
MSA on X’s
Approve project
Data collection
Performance objective
Identify drivers of variation
Study interaction between X’s
Improved Process capability
Establish control plan
Treqna Base Manual
First Edition 313 © All Rights Reserved TreQna 2005
Control Phase Objectives
• Ensure that the process stays in control after the solution is implemented
• Detect and identify any potential deviation from control state and ensure that a
corresponding correcting procedure is established
• Establish plan to ensure sustenance of improved process performance
Treqna Base Manual
First Edition 314 © All Rights Reserved TreQna 2005
What Does Control Mean
• Monitor X’s using control charts for variation
• Identify potential variation causes and prepare an action plan
• Establish plan to ensure sustenance of improved process performance
Treqna Base Manual
First Edition 315 © All Rights Reserved TreQna 2005
Refer to Step A.1 and A.3
Process capability, Pre-Post data comparison
Improved Process Capability
Step C.2
Tools
Activities
Deliverables
Objective
Step C.3Step C.1
Mistake proofing, Risk management, Statistical process control
FMEA & Control ChartsRefer to step M.3
Control PlanAcceptable gage
Establish Control PlanMSA on Xs
Control Phase
DEFINE MEASURE ANALYZE IMPROVE CONTROL
Treqna Base Manual
First Edition 316 © All Rights Reserved TreQna 2005
MSA on Xs
Deliverable for this step:
• Validate measurement system for Xs
• Refer to step M3
Treqna Base Manual
First Edition 317 © All Rights Reserved TreQna 2005
Refer to Step A.1 and A.3
Process capability, Pre-Post data comparison
Improved Process Capability
Step C.2
Tools
Activities
Deliverables
Objective
Step C.3Step C.1
Mistake proofing, Risk management, Statistical process control
FMEA & Control ChartsRefer to step M.3
Control PlanAcceptable gage
Establish Control PlanMSA on Xs
Control Phase
DEFINE MEASURE ANALYZE IMPROVE CONTROL
Treqna Base Manual
First Edition 318 © All Rights Reserved TreQna 2005
Improved Process Capability
Deliverable for this step:
• Calculate capability for new process
• Compare pre and post data for project Y using tools described in step A.3
Treqna Base Manual
First Edition 319 © All Rights Reserved TreQna 2005
Refer to Step A.1 and A.3
Process capability, Pre-Post data comparison
Improved Process Capability
Step C.2
Tools
Activities
Deliverables
Objective
Step C.3Step C.1
Mistake proofing, Risk management, Statistical process control
FMEA & Control ChartsRefer to step M.3
Control PlanAcceptable gage
Establish Control PlanMSA on Xs
Control Phase
DEFINE MEASURE ANALYZE IMPROVE CONTROL
Treqna Base Manual
First Edition 320 © All Rights Reserved TreQna 2005
Statistical Process Control (SPC)
• Control charts are used to monitor variation on X’s and process performance over
time. Control charts help in identifying variation caused due to special cause.
• Control charts are not proactive tool as mistake proofing is. They are useful when
X’s can not be mistake proofed and controlled within tolerance.
Treqna Base Manual
First Edition 321 © All Rights Reserved TreQna 2005
Statistical Process Control
• The purpose of using statistical purpose control is to monitor X’s and process
performance over time for variation.
• Post improvements a stable process working at its new capability would have
common cause variation.
• If the process has changed and has special cause variation than action needs
to be taken.
• Control charts are used as tools for statistical process control. They tell us when
X’s go out of control and surface the special cause performance.
Treqna Base Manual
First Edition 322 © All Rights Reserved TreQna 2005
Variation
• Common Cause Variation – is the natural or true variation of process and is inherent
in the process
• Special Cause Variation- is usually of larger magnitude compared to common cause
variation. It is usually due to special causes and happens once in a while.
• Precautionary measure should be taken to avoid recurrence of special causes
Treqna Base Manual
First Edition 323 © All Rights Reserved TreQna 2005
Common cause, Special cause
• Common cause
• Natural, Random
• Inherent to any process
• Variation caused by common cause can be reduced by making changes to the system. This would result in improved capability of the system.
• Special Cause
• Unusual
• Specific and once in a while (sporadic)
• Captured in control charts
• Process stability may be explained as:
• In Control, Stable - Only common cause
• Out of Control, Unstable - Common cause with special cause is present
Treqna Base Manual
First Edition 324 © All Rights Reserved TreQna 2005
Differentiating Variation
Perceived Variation
Common Causes Special Causes
CommonCauses
SpecialCauses
Tampering of process
Focus on process change
Under-reactingOpportunity missed for
taking precautions
Focus on mistake proofing, taking precautions.
True Variation
Precaution should be taken while studying variation, the mistakes in treating common cause as special cause or special cause as common cause causes problem for the process. We usually end up in either creating more variation in the process or not able to rectify the cause of variation.
Treqna Base Manual
First Edition 325 © All Rights Reserved TreQna 2005
Control Charts
A control chart:
• Is a time ordered plot of the data
• Plots the variation of the data in expected range also called control limits
• Identifies when there is a special cause present
Upper Control Limit
Lower Control Limit
Example of Control Chart
Control charts use +/- 3 sigma for control limits. One should calculate the control limits when the process is stable and freeze them. Control charts in the future should be used using the same limits, this would keep an eye on special cause as it should and would also tell if the capability of the process is going down.
Treqna Base Manual
First Edition 326 © All Rights Reserved TreQna 2005
Control Charts – Minitab Tests
Area of any control chart can be divided into different segments in terms of distance from centre line in multiples of standard deviation. Each segment becomes a zone in separated from centre line by multiples of s.
The eight tests defined in Minitab look for position of data points on in the following zones of control charts.
-3σ
-2σ
+2σ
+3σ
-1σ
+1σ
Zone A
Zone A
Zone B
Zone B
Zone C
Zone C
For any control chart, you can find the Minitab Rules under the “test” Button.You can select the rules you want Minitab to use, but you should select all eight tests.Note: Minitab rules only should be used in the exam to evaluate special causes of variation.
Treqna Base Manual
First Edition 327 © All Rights Reserved TreQna 2005
Minitab Tests
Test 1: One point or more points more than 3s from the center line.
Test 2: 9 data points in a row on the same side of center line.
UCL
LCL
Average
ABCCBA
1
UCL
LCL
Average
ABCCBA
2
•Test 1 is positive if there is a shift in the process mean if there is an increase in the process standard deviation, or if there is a “single aberration in the process such as a mistake in calculation, an error in measurement, process breakdown etc
•Test 2 signals a shift in the process mean.
Treqna Base Manual
First Edition 328 © All Rights Reserved TreQna 2005
Minitab Tests
Test 3: Six points in a row, either increasing or decreasing
Test 4: Fourteen points in row, alternating up and down
UCL
LCL
Average
ABCCBA
4
UCL
LCL
Average
ABCCBA
3
•Test 3 signals a drift in the process mean. The causes can include improvement in skill, any kind of deterioration in the process.
•Test 4 signals a systematic effect produced by two different population. The causes could be due to two different operator, two different process etc.
Treqna Base Manual
First Edition 329 © All Rights Reserved TreQna 2005
Minitab Tests
Test 5: Two out of three points greater than 2s on the same side
Test 6: Four out of five points greater than 1s on the same side
UCL
LCL
Average
ABCCBA
5
UCL
LCL
Average
ABCCBA 6
•In the case of charts for variables, the first four tests should be augmented by Tests 5 and 6 when earlier warning is desired.
Treqna Base Manual
First Edition 330 © All Rights Reserved TreQna 2005
Minitab Tests
Test 7: Fifteen points in a row within 1s from center line on either side.
Test 8: Eight points in a row greater than 1s from centre line on either side
UCL
LCL
Average
ABCCBA
8
UCL
LCL
Average
ABCCBA
7
•Tests 7 and 8 indicate stratification (observations in a subgroup have multiple sources with different means). Test 7 is positive when the observations in the subgroup always have multiple sources. Test 8 is positive when the subgroups are taken from one source at a time.
Treqna Base Manual
First Edition 331 © All Rights Reserved TreQna 2005
Control Limits vs. Specification Limits
Control Limits: Control limits are calculated by Minitab based on the data input. They are internal to the process and are calculated using standard deviation. Control limits refer to stability of the process at a given capability.
Specification Limits: Specification limits are defined by the customer. It defines the desired level of performance by customer. Specification limits are not based on any statistic, but is based on what customer expects. Specification limits are used to derive the defect rate of process and hence the capability of the process.
PROCESS A
Process A under control limit and under customer specification limit
LowerSL
UpperSL
PROCESS B
Process B under control limit but has unacceptable variation when evaluated against customer specification limits.
LowerSL
UpperSL
Treqna Base Manual
First Edition 332 © All Rights Reserved TreQna 2005
Process States
The Four possibilities of any Process
The State of Chaos
• Non conforming and non predictable
• Non conformity is also unstable
• Random changes to process would not fix the process
• Need to eliminate effect of assignable causes
The Brink of Chaos
• Confirming yet unpredictable
• Process output is influenced by assignable causes
The Threshold State
• Non Confirming and predictable
• Periodic instability
• Stable but not able to deliver on customer standards fully
The Ideal State
• Confirming and predictable
• Process control limits fall within specification limits
• Process stays inherently stable over time
Treqna Base Manual
First Edition 333 © All Rights Reserved TreQna 2005
Types of Control Charts
Variable Control Chart
• Uses continuous metric like Cycle Time, call handle time, Lengths, Diameters, Drops, etc.
• Generally one characteristic per chart
Attribute Control Chart
• Uses attribute or ordinal data like Pass/Fail, Good/Bad, Go/No-Go
• Could be many characteristics per chart
Variable = ContinuousAttribute = Discrete
Treqna Base Manual
First Edition 334 © All Rights Reserved TreQna 2005
Selecting Control Charts
Get Data
Variable or Attribute
Continuous Data (cycle time, length etc
Attribute Data (Yes / No, Go / No-go, Good/ Bad
Rational Subgroups?
Constant Lot Size?Yes No
Yes
No
X-bar and R for sample of size <8
X-bar and S sample of larger size
I-MR
Defect or Defective
Defect orDefective
c nppu
Treqna Base Manual
First Edition 335 © All Rights Reserved TreQna 2005
Variable Control Charts
Variable Control Chart without rational Subgroups:
• Individuals Chart: a plot of the individual values over time
• Moving-Range Chart: a plot of the moving range over time (Minitab takes the range as two consecutive data points, it can be changed in chart options)
• I-MR Chart: plots individual chart and moving range chart in single window
Variable Control Chart with rational Subgroups:
• S Chart: a plot of sample standard deviations over time
• X-bar Chart: plots of the sample means over time
• R-Chart: a plot of the sample range of a sample over time
• X-bar and R: plots x-bar chart and r chart in the same chart window
• X-bar and S: plots x-bar chart and s chart in the same process window
Variable = ContinuousAttribute = Discrete
Treqna Base Manual
First Edition 336 © All Rights Reserved TreQna 2005
Assumptions for Variable Charts
• There are two assumptions about the plotted statistic which control charts make:
• Data is random, each data point is not influenced by past data point value.
• The data has normal probability distribution function
• Though control charts assume normality of data, they give reliable results for non normal data also. However if the data is extremely skewed one might like to transform data. Minitab provides Box-Cox transformation model for transforming data for control charts
Treqna Base Manual
First Edition 337 © All Rights Reserved TreQna 2005
Data Collection
• Data should be in time order as used for run charts
• Use rational sub grouping as discussed in step M.3
• Use team and subject matter experts for rational sub grouping
Treqna Base Manual
First Edition 338 © All Rights Reserved TreQna 2005
X Bar R Chart
• Used for data with subgroup size less than 10, for subgroup size more than 10 use X bar S, which plots subgroup standard deviations instead of range
Sample
Sa
mp
le M
ea
n
191715131197531
600.5
600.0
599.5
599.0
__X=599.548
UC L=600.321
LC L=598.775
Sample
Sa
mp
le R
an
ge
191715131197531
3
2
1
0
_R=1.341
UC L=2.835
LC L=0
Xbar-R Chart of Weight
Central line at Mean
Control Limits
Treqna Base Manual
First Edition 339 © All Rights Reserved TreQna 2005
I-MR-R/S (Between/Within)
Three charts and three perspectives for the process
Subg
roup
Mea
n
191715131197531
601
600
599
_X=599.548
UCL=600.763
LCL=598.333
MR
of
Subg
roup
Mea
n
191715131197531
1.6
0.8
0.0
__MR=0.457
UCL=1.493
LCL=0
Sample
Sam
ple
Ran
ge
191715131197531
3.0
1.5
0.0
_R=1.341
UCL=2.835
LCL=0
I-MR-R/S (Between/Within) Chart of Weight
Individual charts for sample means, control limits are based on moving range of means. Refers to stability of process location
Moving range charts subgroup means. Refers to between sample variation
R chart or S chart showing within sample variation
Treqna Base Manual
First Edition 340 © All Rights Reserved TreQna 2005
I-MR Chart
Used for data with no sub-grouping
Observation
Ind
ivid
ua
l V
alu
e
37332925211713951
0.120
0.115
0.110
0.105
0.100
_X=0.11025
UC L=0.11782
LC L=0.10268
Observation
Mo
vin
g R
an
ge
37332925211713951
0.016
0.012
0.008
0.004
0.000
__MR=0.00285
UC L=0.00930
LC L=0
1
1
1
1
I-MR Chart of Data
Individual values plotted with mean as centre line
Moving range using alternate data values, i.e. length of 2
Treqna Base Manual
First Edition 341 © All Rights Reserved TreQna 2005
Discrete Control Charts
Variable = ContinuousAttribute = Discrete
Treqna Base Manual
First Edition 342 © All Rights Reserved TreQna 2005
C Chart
• Charts defects when sub group size is constant and based on Poisson distribution
• C-bar is calculated as (Total defects / Total number of units)
Sample
Sam
ple
Coun
t
191715131197531
8
7
6
5
4
3
2
1
0
_C=2.55
UCL=7.341
LCL=0
C Chart of Data
Treqna Base Manual
First Edition 343 © All Rights Reserved TreQna 2005
u Chart
• Charts defects when sub group size is not constant and based on Poisson distribution
• U-bar is calculated as (Total defects / Total number of units)
Sample
Sam
ple
Coun
t Pe
r Un
it
151413121110987654321
0.14
0.12
0.10
0.08
0.06
0.04
0.02
0.00
_U=0.0478
UCL=0.1134
LCL=0
1
U Chart of Defects
Tests performed with unequal sample sizes
Treqna Base Manual
First Edition 344 © All Rights Reserved TreQna 2005
np Chart
• Charts defectives when sub group size is constant and based on Binomial distribution
Sample
Sam
ple
Coun
t
10987654321
30
25
20
15
10
5
0
__NP=10.8
UCL=20.39
LCL=1.21
1
NP Chart of Rejects
Treqna Base Manual
First Edition 345 © All Rights Reserved TreQna 2005
p Chart
• Charts defectives when sub group size is not constant and based on Binomial distribution
Sample
Prop
orti
on
151413121110987654321
0.14
0.12
0.10
0.08
0.06
0.04
0.02
0.00
_P=0.0478
UCL=0.1118
LCL=0
1
P Chart of Rejects
Tests performed with unequal sample sizes
Treqna Base Manual
First Edition 346 © All Rights Reserved TreQna 2005
Process Control System
• A process control system is a strategy to sustain the improvement in the process. It includes the action plan for any deviation or variation from the control state.
• A process control system includes:
• Risk Management
• Mistake-proofing
• Statistical process control (SPC)
Risk Management
Mistake Proofing
Statistical Process Control
Avoidance
Control
Process control system ensures ongoing process control, which also means that the process keeps delivering the improved capability.
Factors which make a process control system effective:•Clarity of requirements before making the control plan•Clear communication to involved parties•Good training•Buy-in from stakeholders – level of involvement and ownership from stakeholders can help
Treqna Base Manual
First Edition 347 © All Rights Reserved TreQna 2005
Risk Management
• Identify risk which can deteriorate the performance of the process
• Evaluate each risk for probability of happening and impact on the process
• Make action plan to limit or reduce the risk
Treqna Base Manual
First Edition 348 © All Rights Reserved TreQna 2005
Risk Management
• Identify risks in the following categories:
• Technology related
• Financial
• Decision making
• Business
• Use FMEA as a risk assessment tool. (Refer to M.1)
Treqna Base Manual
First Edition 349 © All Rights Reserved TreQna 2005
Mistake Proofing
• A technique to eliminate errors
• Also known as Poka Yoke (Japanese)
• Proactive approach to stop errors from happening
• Example:
• Fuel reserve indicator in vehicles
• Shape of SIM card allows only one way to place it in cell
Treqna Base Manual
First Edition 350 © All Rights Reserved TreQna 2005
Mistake Proofing
• Helps to limit X’s within the tolerance so that the process performance does not go out of control
• Identifies the loopholes and helps implementing proactive steps for warning or avoidance of situation which can lead to X’s stepping out of tolerance
Treqna Base Manual
First Edition 351 © All Rights Reserved TreQna 2005
Mistake Proofing
• How to mistake proof:
• Understand origin of defects
• Identify sources of errors resulting in defects
• Identify process steps where mistake proofing can be done
• Redesign process steps such that they are resistant to errors. The options to make errors are reduced.
Treqna Base Manual
First Edition 352 © All Rights Reserved TreQna 2005
Errors
• Sources of errors:
• Process Variation
• Procedural incorrectness
• Measurement accuracy
• Human errors
• One can use fishbone analysis to brain storm and identify the sources of errors
Treqna Base Manual
First Edition 353 © All Rights Reserved TreQna 2005
Control Plan
• Control plan for a project should include following things:
• CTQ
• Input and out variables
• Specifications and tolerances
• Control measures and tools
• Monitoring and sampling plan
Treqna Base Manual
First Edition 354 © All Rights Reserved TreQna 2005
Control Plan Sheet
• Following is a sample control plane sheet
Date Created:Date Revised:
Measure Process Step Input/Output Specifications TargetSample
Size
Data Collection Frequency
Control Tool
Reaction Plan
Control Plan
Project Y, X’s to be measured
Process step where measured
Whether input or output measure
Specifications and tolerances
Target for the measure
Sample size to be
collected
How often the data is to be
collected
Control charts, Mistake proofing
other tools
Actions to be taken when the measure goes out of control
Treqna Base Manual
First Edition 355 © All Rights Reserved TreQna 2005
Who owns it?
Responsibility plan:
• Who owns the activity?
• Who is the back up?
• Define procedures and operational definitions
• Escalation matrix:
• Escalation matrix as a part of reaction plan
Treqna Base Manual
First Edition 356 © All Rights Reserved TreQna 2005
Control - Summary
• Statistical process control
• Eight tests for control charts
• Variable and Attribute control charts
• Control limits vs. Specification limits
• Risk management and FMEA
• Mistake proofing
• Control plan
• Elements of control plan
• Additional details