Six Sigma Treqna Base Manual Ed1 PDF

Treqna Base Manual
treqna.com
First Edition
All Rights Reserved TreQna 2005
C ontents
DEFINE Step D1 Map Project Step D2 Approve Project MEASURE Step M1 CTQ Characteristics & Standards Step M2 Measurement System Analysis Step M3 Data Collection Plan ANALYZE Step A1 Baseline Process Step A2 Performance Objective Step A3 Identify drivers of Variation IMPROVE Improve Step I1 Screen for Vital Xs Improve Step I2 Study Interaction between Xs Improve Step I3 Define Improved Process CONTROL Control Step C1 MSA on Xs Control Step C2 Improved Process Capability Control Step C3 Establish Control Plan Page 4 Page 36 Page 45 Page 82 Page 116
Treqna Base Manual
Page 145 Page 191 Page 197
First Edition
Treqna Base Manual
DEFINE MEASURE ANALYZE IMPROVE CONTROL
Map project Approve project
CTQ Characteristics and standards Measurement System Analysis Data collection
Baseline process Performance objective Identify drivers of variation
Screen for vital Xs Study interaction between Xs Defined Improve Process
MSA on Xs Improved Process capability Establish control plan
First Edition
Treqna Base Manual
Define Phase
DEFINE MEASURE ANALYZE IMPROVE CONTROL Step D.2
Approve project Project feasibility CBA FMEA
Step D.1
Step Deliverables Map project Project Charter COPIS Survey, Focus Group, Interview, Charter, COPIS
Tools
First Edition
Treqna Base Manual
Topics
About Projects Customer, CTQs and VOC COPIS Project Charter
First Edition
Treqna Base Manual
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
First Edition
Treqna Base Manual
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
First Edition
Treqna Base Manual
Topics
About Projects Customer, CTQs and VoC COPIS Project Charter
First Edition
Treqna Base Manual
Identifying the Customer

Who is the customer? Internal Customer or External Customer ?
Process Customer Output Input Supplier
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
First Edition
Treqna Base Manual
Critical to Quality (CTQ)

CTQs from any process are translated from customer requirements
External Customer
Internal Customer
Customer driven CTQs
Process driven CTQs
Customer to CTQ
Identify Customer
Capture customer requirement
Derive and Define CTQs
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.
First Edition
10
Treqna Base Manual
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.
First Edition
11
Treqna Base Manual
VOC Tools Evaluation

Survey
Advantages: 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
First Edition
12
Treqna Base Manual
VOC Tools Evaluation

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
First Edition
13
Treqna Base Manual
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 sequence Sample Bias No-response bias Unclear questions Interviewer / facilitator induced bias
First Edition
14
Treqna Base Manual
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 needsthose 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.
First Edition
15
Treqna Base Manual
Topics
First Edition
16
Treqna Base Manual
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
Thought Process Process Flow

C Customer O Output P Process I Input S Supplier
COPIS and SIPOC are used interchangeably.
First Edition
17
Treqna Base Manual
COPIS - Components
C Customer O Output P Process I Input S Supplier
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
First Edition
18
Treqna Base Manual
COPIS - Components
Step1: Identify Customers, Outputs, Inputs, Suppliers and Process Name
C Customer
O Output
P Process
I Input
S Supplier
Step2: Identify customer requirements Step3: Identify process steps
First Edition
19
Treqna Base Manual
Topics
First Edition
20
Treqna Base Manual
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
First Edition
21
Treqna Base Manual
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
First Edition
22
Treqna Base Manual
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.
First Edition
23
Treqna Base Manual
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?
First Edition
24
Treqna Base Manual
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.
First Edition
25
Treqna Base Manual
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
First Edition
26
Treqna Base Manual
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.
First Edition
27
Treqna Base Manual
The Goal Statement Example

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.
First Edition
28
Treqna Base Manual
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
First Edition
29
Treqna Base Manual
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
First Edition
30
Treqna Base Manual
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
First Edition
31
Treqna Base Manual
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
First Edition
32
Treqna Base Manual
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 Belts Master 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 challenges Sponsor 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
First Edition
33
Treqna Base Manual
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)
Champions role is not clearly defined
First Edition
34
Treqna Base Manual
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
First Edition
35
Treqna Base Manual
Define Phase
DEFINE MEASURE ANALYZE IMPROVE CONTROL Step D.2
Approve project Project feasibility CBA FMEA
Step D.1
Step Deliverables Map project Project Charter COPIS Survey, Focus Group, Interview, Charter, COPIS
Tools
First Edition
36
Treqna Base Manual
Topics
Project Risks Cost Benefit Analysis Project Go/No Go decision
First Edition
37
Treqna Base Manual
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
First Edition
38
Treqna Base Manual
Topics
First Edition
39
Treqna Base Manual
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.
First Edition
40
Treqna Base Manual
Topics
First Edition
41
Treqna Base Manual
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 champions commitment to mitigate risks and drive contingency measures, if any require his influence
Financial commitment for the completion of the project (allocate budget)
First Edition
42
Treqna Base Manual
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 champions commitment to ensure the success of the project
First Edition
43
Treqna Base Manual
First Edition
44
Treqna Base Manual
Measure Phase
DEFINE MEASURE Step M.1
Step CTQ characteristics and standards Identify project Y metric Establish performance standards Process Map, QFD, FMEA, Pareto
ANALYZE
IMPROVE
CONTROL Step M.3

Data collection Sampling Strategy Segmentation factors
Step M.2
Measurement system analysis Identify gage error
Deliverables
Tools
Continuous Gage R&R, Short Form Gage, Test/ Retest Study, Attribute Gage R&R
First Edition
45
Treqna Base Manual
Topics
Selecting CTQ Characteristics Process mapping CTQ Drill down tree QFD FMEA
Performance Standards
First Edition
46
Treqna Base Manual
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.
First Edition
47
Treqna Base Manual
Understanding a process
A process is a series of logical steps that transform inputs (raw materials) in to customer defined outputs.
Process Customer Output Input Supplier
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
First Edition
48
Treqna Base Manual
Components of a process
Process Controls
------------------- Process Boundary -------------------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.
First Edition
49
Treqna Base Manual
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)
First Edition
50
Treqna Base Manual
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 processs key requirements
First Edition
51
Treqna Base Manual
Frequently used symbols
Process step
Process Start / End
Decision box
Input to a process
Off page connector
Process steps connector
Document
Multidocument
First Edition
52
Treqna Base Manual
Types of process maps

Three frequently used process maps are: Basic Process map: used when the process is small and simple
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 80% this map more informative.
20%
Cross functional process map: used when the process has many handoffs between different departments. It is also known as deployment map.
Human Resource Operations Finance
First Edition
53
Treqna Base Manual
Topics
First Edition
54
Treqna Base Manual
CTQ Drill down Tree

E.g. Ticket Counter at Movie theatre. From customer VOC to project metric
Big Y
Better Billing Process
Good Customer Service
Timely Process
Curt and Polite
Knowledgeable
Accurate Billing
Less queue time
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.
First Edition
55
Treqna Base Manual
Topics
First Edition
56
Treqna Base Manual
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
First Edition
57
Treqna Base Manual
QFD -Objectives
Understand Quality Function Deployment (QFD) as a tool.
Describe the steps of QFD.
Analyze output from QFD.
First Edition
58
Treqna Base Manual
Definition of QFD
QFD is a methodology and tool to identify and translate customer needs and wants into measurable features and requirements converting the whats in to hows
QFD links the needs of the customer (end user) with design, development, manufacturing, and service functions.
Used to identify Critical to Quality Characteristics (CTQs)
QFD is also known as the House of Quality
First Edition
59
Treqna Base Manual
QFD Matrix
Correlation between measures
9 9
Time to prepare
9
Temperature (cold) Ingredient cost
Measures that capture customers requirement
Customers importance rating Customers Expectations Customer Needs

Quenches thirst Should taste good Should be reasonably priced Good service 4 4 2 3
Overall CTQ impact on customer needs

9 9 Strong Positive Strong Negative 9 Positive Negative
Ingredient mix 9 36
Quantity
9 3 9
9 39 48 18 42
Impact a CTQ has on the customers expectations
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 customers 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 measures When 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 3050% 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.
First Edition
60
Treqna Base Manual
Analyzing the QFD matrix

Look for blank rows and columns blank row means that we dont 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 CTQs
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.
First Edition
61
Treqna Base Manual
Topics
First Edition
62
Treqna Base Manual
FMEA - Objectives
Understanding what an FMEA does
Types of FMEAs
When to use
Learn steps involved in making of FMEA
First Edition
63
Treqna Base Manual
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
First Edition
64
Treqna Base Manual
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.
First Edition
65
Treqna Base Manual
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 CTQs identification Improve Risk analysis on solution Control To develop process control plan
First Edition
66
Treqna Base Manual
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
First Edition
67
Treqna Base Manual
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)
First Edition
68
Treqna Base Manual
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?
First Edition
69
Treqna Base Manual
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
First Edition
70
Treqna Base Manual
Rating Scale
Many scales are available for rating severity, occurrence and detection. For simplicity (recommended for GBs) one can use the following scale:
Rating
1
Severity
Low- e.g. Low or no effect customer Medium
Occurrence
Low e.g. remote likelihood Medium-
Detection
High- e.g.. always detectable Medium
High- e.g. effecting customer
High- e.g. almost assured failure
Low- e.g. not detectable
First Edition
71
Treqna Base Manual
How does it look?
List process steps or product parts
List Failure Modes for each process step

Failure Modes and Effects analysis (FMEA )
Rate the Severity of the Effect to the customer
Rate the detection based on current controls
Date S Potential Potential Potential E Process Step Causes failure mode failure effect V.
Prepared By: O Current C Controls C D R S O D R E P Recommendations E C E P V. C T N T N
List the Effects of each Failure Mode
List the causes for each Failure Mode
Rate how often a particular cause or Failure Mode occurs
Calculate RPN and give recommendations for reducing RPN. Calculate new RPN post implementation of actions
First Edition
72
Treqna Base Manual
Topics
First Edition
73
Treqna Base Manual
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.
First Edition
74
Treqna Base Manual
Elements - CTQ performance standard

An Example CTQ
Project Y Turn Around Time
CTQ Definition
Time from application to new account opening
CTQ
Unit of Measure
Days
Performance Standard
CTQ Target
3 Days
Specification Limits
Not greater than 5 days
Defect Definition
Each new account Application which Takes more than 5 days
First Edition
75
Treqna Base Manual
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
First Edition
76
Treqna Base Manual
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!)
First Edition
77
Treqna Base Manual
Target, Specification limit and Defect

Defect Anything that results in customer dissatisfaction. Nonconformance to the CTQ.
Target Specification Limit Defect
First Edition
78
Treqna Base Manual
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
First Edition
79
Treqna Base Manual
Formulas
Defects per Unit Total 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 Opportunities Total number of units multiplied by opportunity per unit.
Defects per Opportunity Total Defects divided by total opportunities.
Defects per Million Opportunities Defects per opportunity multiplied by a million.
First Edition
80
Treqna Base Manual
Topics Covered
First Edition
81
Treqna Base Manual
Measure Phase
Step CTQ characteristics and standards Identify project Y metric Establish performance standards Process Map, QFD, FMEA, Pareto
ANALYZE
IMPROVE
CONTROL Step M.3

Step M.2
Deliverables
Tools
First Edition
82
Treqna Base Manual
Topics
Measurement System Analysis - Concepts
Types of MSA
First Edition
83
Treqna Base Manual

Measurement System Analysis - Objectives
Understand the terms: Precision , Accuracy, Resolution , Stability, Linearity, Repeatability, Reproducibility,
Evaluated the measurement system and established variation induced by it in process data.
Project teams consensus on measurement system
And define calibration standards
First Edition
84
Treqna Base Manual
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.
First Edition
85
Treqna Base Manual
Process Variation
Process True Variation producing parts different lengths for parts
Input Process Output
True Process Variation Variation Induced by measurement system 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.
First Edition
86
Treqna Base Manual
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`
First Edition
87
Treqna Base Manual
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.
First Edition
88
Treqna Base Manual
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.
First Edition
89
Treqna Base Manual
Understanding Precision and Accuracy
Target
Precise but away from target
Precise but not accurate
==
At target and precise
Precise and accurate
At target but not precise
Accurate but not precise
First Edition
90
Treqna Base Manual
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. ReproducibilityVariation when two or more people measure the same unit with the same measuring equipment. StabilityVariation obtained when the same person measures the same unit with the same equipment over a large gap of time. LinearityThe consistency of the accuracy of gage across the entire range of the measurement system.
First Edition
91
Treqna Base Manual
Topics
Measurement System Analysis - Concepts
Types of MSA
First Edition
92
Treqna Base Manual
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.
First Edition
93
Treqna Base Manual
Improving the Measurement System
NOT Precise and NOT Accurate
Precise but NOT Accurate
Precise and Accurate
First Edition
94
Treqna Base Manual
Attribute GAGE
Checks gage for: Repeatability
Reproducibility
Accuracy
First Edition
95
Treqna Base Manual
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
First Edition
96
Treqna Base Manual
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.
First Edition
97
Treqna Base Manual
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.
First Edition
98
Treqna Base Manual
Attribute gage sheet

TRUE Sam ple A nswer 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 N Y Y N N Y Y N N N Y Y Y N N N Y N Y Y Y N N N N Y N N N Y First Operator Trial1 N Y Y N N Y N N N N Y N Y N N N Y N Y N Y N N N N Y N N N N Trial2 N Y Y N N Y Y N N N Y N Y N N N Y N Y N Y N N N N Y N N N Y Trial3 N Y N N N Y N N N N Y N Y N N N Y N Y N Y N N N N Y N N N N Second Operator Trial1 N Y Y Y N Y Y Y N N Y Y Y N N N Y N Y Y Y Y Y N N Y N N N Y Trial2 N Y Y Y N Y N N N N Y Y Y N N N Y N Y Y Y Y Y N N Y N N N Y Trial3 N Y Y Y N Y Y Y N N Y Y Y N N N Y N Y Y Y Y Y N N Y N N N Y Third Operator Trial1 Y Y Y N N Y Y N N Y Y Y Y N Y N Y N Y Y Y N N Y N Y N N N Y Trial2 N Y Y N N Y Y N N N Y Y Y N Y N Y N Y Y Y N N Y N Y N N N Y Trial3 N Y Y N N N Y N N Y Y Y Y N Y N Y N Y Y Y N N Y N Y N N N Y
Repeatability: % times each operator matches on all three measures for the same unit. Total 90 opportunity exist in the example.
Reproducibility: % times all operators match on all repeated measures for the same unit. Total 30 opportunities exist in the example
Accuracy % times each individual measure matches with the standard. Since there are total 270 measures, there are 270 opportunities
First Edition
99
Treqna Base Manual
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).
First Edition
100
Treqna Base Manual
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.
First Edition
101
Treqna Base Manual
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
First Edition
102
Treqna Base Manual
Test re-test gage

Make Run chart to find out if the data is stable
Run Chart of C1
107.5 105.0 102.5 C1 100.0 97.5 95.0 2 4 6
13 11.00000 3 0.82094 0.17906
10 12 Observation
14
16
16 13.00000 2 0.95238 0.04762
18
20
Number of runs about median: Expected number of runs: Longest run about median: Approx P-Value for Clustering: Approx P-Value for Mixtures:
Number of runs up or down: Expected number of runs: Longest run up or down: A pprox P-Value for Trends: A pprox P-Value for O scillation:
First Edition
103
Treqna Base Manual
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
Summary for C1
A nderson-D arling N ormality T est A -S quared P -V alue M ean S tD ev V ariance S kew ness K urtosis N M inimum 1st Q uartile M edian 3rd Q uartile M aximum 97.505 96.235 9 5 % C onfide nce I nte r v a ls
Mean Median 96 97 98 99 100 101
0.36 0.425 99.200 3.622 13.116 0.367278 -0.959324 20 94.000 96.000 98.500 102.500 106.000 100.895 101.000 5.290
94
96
98
100
102
104
106
95% C onfidence I nterv al for M ean 95% C onfidence Interv al for M edian 95% C onfidence Interv al for S tD ev 2.754
Therefore, the measurement device has inadequate precision.
First Edition
104
Treqna Base Manual
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
First Edition
105
Treqna Base Manual
How a gage plays with Spec Limits

What is the challenge with having high R&R error ? Following scenario explains why..
The gage variation might cause this data to fail on the lower specification limit LSL The gage variation might cause this data to fail on the upper specification limit USL
The gage variation might cause this data to pass on lower specification limit
The gage variation might cause this data to pass on upper specification limit
Centre circle is the true data point and the limits on left and right show the measurement variation
First Edition
106
Treqna Base Manual
Gage ANOVA - Objectives

Gage ANOVA study: Quantify amount of measurement variability Identify amount of measurement variation from different sources
Data Type: Continuous
First Edition
107
Treqna Base Manual
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 Response

1 2 3 4 1 2 3 4 1 2 3 4 1 2 3 4 1 1 1 1 1 1 1 1 2 2 2 2 2 2 2 2 1 1 1 1 2 2 2 2 1 1 1 1 2 2 2 2 20.0 19.1 19.5 18.2 21.2 20.0 19.0 22.0 17.0 19.5 19.6 18.0 19.5 20.1 21.0 18.9
First Edition
108
Treqna Base Manual
Gage ANOVA Minitab

Choose Stat> Quality Tools> Gage Study> Gage R&R Study (Crossed).
First Edition
109
Treqna Base Manual
Gage ANOVA Minitab
Operator Part Trial Response

1 2 3 4 1 2 3 4 1 2 3 4 1 2 3 4 1 1 1 1 1 1 1 1 2 2 2 2 2 2 2 2 1 1 1 1 2 2 2 2 1 1 1 1 2 2 2 2 20.0 19.1 19.5 18.2 21.2 20.0 19.0 22.0 17.0 19.5 19.6 18.0 19.5 20.1 21.0 18.9
Select method of analysis as ANOVA
First Edition
110
Treqna Base Manual
Minitab Session Output

Gage R&R Study - ANOVA Method Two-Way ANOVA Table With Interaction Source Part Operator Part * Operator Repeatability Total Gage R&R Source Total Gage R&R Repeatability Reproducibility Operator Operator*Part Part-To-Part Total Variation VarComp 50.475 16.283 34.192 10.041 24.151 328.936 379.411 %Contribution (of VarComp) 13. 30 4.29 9.01 2.65 6.37 86.70 100.00 Study Var (6 * SD) 42.627 24.212 35.084 19.012 29.486 108.820 116.871 %Study Var (%SV) 36.47 20.72 30.02 16.27 25.23 93.11 100.00 DF 9 2 18 30 59 SS 18343.8 530.8 1162.5 488.5 20525.7 MS 2038.20 265.40 64.59 16.28 F 31.5583 4.1093 3.9663 P 0.000 0.034 0.000
P-value < 0.05 says significant variation. In the example Parts, operator and the interaction all are significant factors
Source Total Gage R&R Repeatability Reproducibility Operator Operator*Part Part-To-Part Total Variation
StdDev (SD) 7.1046 4.0353 5.8474 3.1687 4.9144 18.1366 19.4785
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%
Number of Distinct Categories = 3
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.
First Edition
111
Treqna Base Manual
Gage ANOVA Graphical Output

Components of variation graph shows percent contribution by category
Gage R&R (ANOVA) for Weight
G age name: Date of study :
Components of Variation
100 Percent
% Contribution % Study Var
In By part graph, there are large differences between different parts.
Reported by : Tolerance: M isc:

Weight by Part
100 75 50
50
Gage R&R
Repeat
Reprod
Part-to-Part
5 Part
10
R Chart by Operator
16 Sample Range 1 2 3 UCL=15.36 100 75 _ R=4.7 LCL=0 50 1
Weight by Operator
Xbar Chart by Operator

1 100 Sample Mean Average 75 _ UCL=89.69 _ X=80.85 LCL=72.01 2 3 100 75 50 1 2 3 4
2 Operator
Operator * Part Interaction

Operator 1 2 3
50
5 6 Part
10
X bar graph by operator shows most of the points out of control, which says most of the variation is due to parts
In by operator graph, there isn't much difference between operators
First Edition
112
Treqna Base Manual
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
First Edition
113
Treqna Base Manual
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
First Edition
114
Treqna Base Manual
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
First Edition
115
Treqna Base Manual
Measure Phase
Objective CTQ characteristics and standards Identify project Y metric Establish performance standards Process Map, QFD, FMEA, Pareto
ANALYZE
IMPROVE
CONTROL Step M.3

Step M.2
Deliverables
Tools
First Edition
116
Treqna Base Manual
Topics
Data Collection Plan
Data Segmentation
Sampling
First Edition
117
Treqna Base Manual

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
First Edition
118
Treqna Base Manual

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.
First Edition
119
Treqna Base Manual
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.
First Edition
120
Treqna Base Manual
Topics
Data Segmentation
Sampling
First Edition
121
Treqna Base Manual
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
First Edition
122
Treqna Base Manual
Data Segmentation
Example: A project is being done on Accuracy of Application processing, the possible segmentation factors are
Work type Touch time Vintage of processor
Accuracy
Day of Week
No of Processors
These segmentation factors are logical groups that may point to be significant xs during the analyze phase.
First Edition
123
Treqna Base Manual
Topics
Data Segmentation
Sampling
First Edition
124
Treqna Base Manual
Sampling - Objectives
Understand sampling and why is it required
Advantages of sampling
Different sampling techniques
First Edition
125
Treqna Base Manual
Population vs. Sample

Population: The entire set of data, the universal set. Symbol used is N The population mean represented by Standard deviation is represented by s
Sample: The part or subset of a population, the universal subset signifies the sample mean X s signifies standard deviation of a sample
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.
First Edition
126
Treqna Base Manual
Sampling
Sampling is the process of collecting subset or portion of data from population and predicting population characteristics
Population N=1000
Sample N=50
First Edition
127
Treqna Base Manual
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.
First Edition
128
Treqna Base Manual
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.
First Edition
129
Treqna Base Manual
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.
First Edition
130
Treqna Base Manual
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.
First Edition
131
Treqna Base Manual
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.
First Edition
132
Treqna Base Manual
Sampling Strategy Random Sampling

Random Sampling:
Sample
Production Line
Samples picked up from assembly line without any order or system. Each item has equal probability of getting selected in the 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.
First Edition
133
Treqna Base Manual
Stratified Random Sampling

Stratified Random Sampling
Sample 1.1 Sample 1.2
Production Line
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.
First Edition
134
Treqna Base Manual
Systematic Sampling
Systematic Sampling:
Sample Production Line 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.
First Edition
135
Treqna Base Manual
Rational Sub grouping

Random Sampling:
3rd hour 2nd hour 1st hour
Sample Production Line

Three items selected with a frequency of one hour. In this case sample has subgroup size of three.
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.
First Edition
136
Treqna Base Manual
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 collection The 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.
First Edition
137
Treqna Base Manual
Sample Size Calc - Continuous Data

Sample represents population characteristics. We can use confidence interval formula and solve it for n to get sample size:
n
Standard error
CI
X Z
/ 2
As standard error is also represented by delta, hence: Solving for n we get:

n =

/2
/2
The equation for 95% confidence becomes (using Z table):

x 1.96 n =
2
Where,
error or precision required from sample in representing the population is standard deviation N is sample size The value of Z/2 depends on confidence level we choose, the corresponding value can taken out from Z table. Finite Population Correction The 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 n N (finite) = n(1 + n/N) FPC =
N 1
First Edition
138
Treqna Base Manual
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)
First Edition
139
Treqna Base Manual
Estimate
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 in minutes or hours Collect some data from population and calculate using the sample
First Edition
140
Treqna Base Manual
Sample Size Calc - Discrete Data

The formula to calculate sample for proportions data is:
=
/2
1 P
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:
P
/2
1 P n
Again if the sample is more than 5% of population finite population correction should be used
First Edition
141
Treqna Base Manual
Estimating p and
Estimate p
Calculate proportionate defective using small sample Use historical control charts to estimate Use subject matter expertise to estimate
Estimate
Use subject matter expertise to estimate Estimate using formula: = Z
P
B
/2
(1 P
n
First Edition
142
Treqna Base Manual
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
First Edition
143
Treqna Base Manual
Screen for vital Xs Study interaction between Xs Define Improved Process
First Edition
144
Treqna Base Manual
Analyze Phase
DEFINE MEASURE
Step A.1
Objective Deliverables Baseline Process Process capability for project Y -Study data for shape, stability and normality -Capability analysis -Short Term vs. Long Term sigma -Understanding shift, Common cause variation, special cause variation Descriptive test, run chart, capability pack Tools
ANALYZE
Step A.2
IMPROVE
CONTROL
Step A.3
Performance Objective Improvement goal for project Y -Concept of Benchmarking -Benchmarking Types -Benchmarking as process
Identify drivers of variation List all statistically significant Xs -Root Cause Analysis (RCA) -Process map analysis -Graphical analysis -Statistical Analysis
Activities
Benchmarking
1-Sample T-test, 2-Sample T-test, Oneway ANOVA, Moods Median, Homogeneity of Variance, Simple Linear Regression, Correlation/Scatter Diagrams, Chi SquareTest of Independence, Chi SquareTest for Goodness of Fit, Pareto, Cause and Effect diagram, VA/NVA analysis
First Edition
145
Treqna Base Manual
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
First Edition
146
Treqna Base Manual
Stability, Shape and Normality

Stability
Tool Used: Run Chart Stability is lack of excessive variation. There should be no trends, clusters, mixtures or oscillations.
Normality
Tool Used: Descriptive test, Normality Test Check if the data follows a normal distributions
Shape
Tool Used: Descriptive Stats, Histogram Study distribution characteristics
First Edition
147
Treqna Base Manual
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
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 values Means 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.
First Edition
148
Treqna Base Manual
Run Chart - Minitab

MINITAB FILE: RunChart.mtw
First Edition
149
Treqna Base Manual
Run Chart - Subgroup Size 1
Choose C2 for Data Set Subgroup Size to 1 Click Ok
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 expected Ha: Fewer runs observed than expected A 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 expected Oscillation occurs when the data fluctuates up and down rapidly, indicating that the process is not steady Clusters: Ho: No fewer runs observed than expected. Ha: Fewer runs observed than expected Clusters 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 expected Mixture 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.`
First Edition
150
Treqna Base Manual
Studying Stability
Ho: Data is random, special causes not present Ha: Data is not random, special causes present
First Edition
151
Treqna Base Manual
Studying Stability
First Edition
152
Treqna Base Manual
Run Chart Subgroup Size more than 1
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.
First Edition
153
Treqna Base Manual
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 ?
First Edition
154
Treqna Base Manual
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 Variable Maximum Agent A 71.000
0.535 8.460 19.000 45.750 51.000 56.000
First Edition
155
Treqna Base Manual
Graphical Output
Click on Stat > Basic Statistics > Graphical Summary

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 value Mean Median Standard Deviation Variance Kurtosis (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.
First Edition
156
Treqna Base Manual
Different Shapes
Central Tendency: Mean
Normal
Dispersion: Standard Deviation
Central Tendency: Q1 or Q3 Dispersion: Stability Factor, defined as Q1/Q3
Skewed
First Edition
157
Treqna Base Manual
Different Shapes
Central Tendency: Median Dispersion: Range or Span
Long Tailed
Bi-modal
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.
First Edition
158
Treqna Base Manual
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
First Edition
159
Treqna Base Manual
Area under a Normal Curve
68.26%
95.46%
99.73%
- 3
- 2
- 1
0
Standard Deviations
The probability of a randomly selected value of being within Deviations from the mean is 95.46%
+ -2
Standard
First Edition
160
Treqna Base Manual
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. Statistically,
Z =
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.
First Edition
161
Treqna Base Manual
Z - Value
LSL LSL
USL USL
ZLSL
ZUSL
-4 Standard Deviations 2 Units of Measure
-3 4
-2 6
-1 8
0 10
1 12
2 14
3 16 18
X = 10
= 2
ZUSL =
USL - X s
X - LSL s
14 - 10 4 = =2 2 2
= 10 - 4 6 = =3 2 2
ZLSL =
A Simple Z Capability Calculation For this calculation assume USL = 14 LSL = 4 Nominal = 10 Note : Nominal is also called Target Value How 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.
First Edition
162
Treqna Base Manual
Calculating Z Bench value
P(d)
LSL
P(d)
USL
LSL
X T
USL
P(d)
Total
= P(d)
LSL
+ P(d)
USL
ZLT = from Z table ZST = ZLT + Z shift
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 1.5 If 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.
First Edition
163
Treqna Base Manual
Single-Tail Z Table A
(Values of Z from 0.00 to 4.99)
Z
0.00 0.10 0.20 0.30 0.40 0.50 0.60 0.70 0.80 0.90 1.00 1.10 1.20 1.30 1.40 1.50 1.60 1.70 1.80 1.90 2.00 2.10 2.20 2.30 2.40 2.50 2.60 2.70 2.80 2.90 3.00 3.10 3.20 3.30 3.40 3.50 3.60 3.70 3.80 3.90 4.00 4.10 4.20 4.30 4.40 4.50 4.60 4.70 4.80 4.90
0.00
5.00e-001 4.60e-001 4.21e-001 3.82e-001 3.45e-001 3.09e-001 2.74e-001 2.42e-001 2.12e-001 1.84e-001 1.59e-001 1.36e-001 1.15e-001 9.68e-002 8.08e-002 6.68e-002 5.48e-002 4.46e-002 3.59e-002 2.87e-002 2.28e-002 1.79e-002 1.39e-002 1.07e-002 8.20e-003 6.21e-003 4.66e-003 3.47e-003 2.56e-003 1.87e-003 1.35e-003 9.68e-004 6.87e-004 4.83e-004 3.37e-004 2.33e-004 1.59e-004 1.08e-004 7.23e-005 4.81e-005 3.17e-005 2.07e-005 1.33e-005 8.54e-006 5.41e-006 3.40e-006 2.11e-006 1.30e-006 7.93e-007 4.79e-007
0.01
0.02
0.03
0.04
0.05
0.06
0.07
0.08
0.09
First Edition
164
Treqna Base Manual
Single-Tail Z Table A
(Values of Z from 5.00 to 9.99)
Z
5.00 5.10 5.20 5.30 5.40 5.50 5.60 5.70 5.80 5.90 6.00 6.10 6.20 6.30 6.40 6.50 6.60 6.70 6.80 6.90 7.00 7.10 7.20 7.30 7.40 7.50 7.60 7.70 7.80 7.90 8.00 8.10 8.20 8.30 8.40 8.50 8.60 8.70 8.80 8.90 9.00 9.10 9.20 9.30 9.40 9.50 9.60 9.70 9.80 9.90
0.00
0.01
0.02
0.03
0.04
0.05
0.06
0.07
0.08
0.09
First Edition
165
Treqna Base Manual
Standard Normal Distribution Table B
This Is Known As The Standard Normal Distribution Which Has A Mean Of Zero (= 0) And A Standard Deviation of One (s = 1)
First Edition
166
Treqna Base Manual
Standard Normal Distribution Table B
This Is Known As The Standard Normal Distribution Which Has A Mean Of Zero (= 0) And A Standard Deviation of One (s = 1)
First Edition
167
Treqna Base Manual
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
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 1.5 If 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.
First Edition
168
Treqna Base Manual
Choosing a Capability Tool

Start Is data normal ? Is data discrete ? No Yes Run Capability Analysis (normal data) Read ZLT from Capability Analysis report and 1.5 to arrive at ZST No Treat data as discrete for this step and run Discrete Data Process Capability report (see notes)
Yes Run Discrete Data Process Capability report (excel spreadsheet)
Is target specified? Yes No Is sub group size > 1?
No
ZST = ZLT + 1.5

(ZShift = 1.5)
Yes
Run Capability Sixpack (see next page for interpretation)
Read ZLT from Capability Analysis report and 1.5 to arrive at ZST
To convert continuous data to discrete follow the following steps 1. Minitab>Stat>Tables>Tally Individual variables 2. Under Variable select column that has data on project Y 3. Select Count and Cumulative percent 4. Look for yield @ USL or LSL in session window 5. Calculate Z value using Abridge table
First Edition
169
Treqna Base Manual
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
First Edition
170
Treqna Base Manual
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.
First Edition
171
Treqna Base Manual
First Edition
172
Treqna Base Manual
First Edition
173
Treqna Base Manual
First Edition
174
Treqna Base Manual
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.
First Edition
175
Treqna Base Manual
Abridged Yield to Sigma Conversion Table

Long-Term Yield
99.99966%
99.9995% 99.9992% 99.9990% 99.9980% 99.9970% 99.9960% 99.9930% 99.9900% 99.9850% 99.9670% 99.9520% 99.9320% 99.9040% 99.8650% 99.8140% 99.7450% 99.6540% 99.5340% 99.1810% 98.930% 98.610% 98.220% 97.730% 97.130% 96.410% 95.540% 94.520% 91.920% 90.320% 88.50% 86.50% 84.20% 81.60% 78.80% 75.80% 72.60% 65.60% 61.80% 58.00% 54.00% 50% 46% 43% 39% 35%
ST Process Defects Per Defects Per 1,000,000 Sigma 100,000

6.0
5.9 5.8 5.7 5.6 5.5 5.4 5.3 5.2 5.1 4.9 4.8 4.7 4.6 4.5 4.4 4.3 4.2 4.1 3.9 3.8 3.7 3.6 3.5 3.4 3.3 3.2 3.1 2.9 2.8 2.7 2.6 2.5 2.4 2.3 2.2 2.1 1.9 1.8 1.7 1.6 1.5 1.4 1.3 1.2 1.1 0.9 0.8 0.7 0.6 0.5 0.4 0.3 0.2 0.1
Defects Per 10,000

0.034
0.05 0.08 0.1 0.2 0.3 0.4 0.7 1.0 1.5
Defects Per 1,000

0.0034
0.005 0.008 0.01 0.02 0.03 0.04 0.07 0.1 0.15
Defects Per 100

0.00034
0.0005 0.0008 0.001 0.002 0.003 0.004 0.007 0.01 0.015 0.033 0.048 0.068 0.096 0.135 0.186 0.255 0.346 0.466 0.819 1.07 1.39 1.78 2.27 2.87 3.59 4.46 5.48
3.4
5 8 10 20 30 40 70 100 150
0.34
0.5 0.8 1 2 3 4 7 10 15
99.9770%
5.0
99.3790%
4.0
6,210
8,190 10,700 13,900 17,800 22,700 28,700 35,900 44,600 54,800
330 480 680 960 1,350 1,860 2,550 3,460 4,660
230
93.320%
3.0
66,800
80,800 96,800 115,000 135,000 158,000 184,000 212,000 242,000 274,000
6,680
8,080 9,680 11,500 13,500 15,800 18,400 21,200 24,200 27,400
819 1,070 1,390 1,780 2,270 2,870 3,590 4,460 5,480
621
33 48 68 96 135 186 255 346 466
23
62.1
81.9 107 139 178 227 287 359 446 548
3.3 4.8 6.8 9.6 13.5 18.6 25.5 34.6 46.6
2.3
0.23
0.33 0.48 0.68 0.96 1.35 1.86 2.55 3.46 4.66 8.19 10.7 13.9 17.8 22.7 28.7 35.9 44.6 54.8 80.8 96.8 115 135 158 184 212 242 274 344 382 420 460 500 540 570 610 650 720 750 780 810 840 860 880 900 920
0.023
6.21
0.621
69.20%
2.0
308,000
344,000 382,000 420,000 460,000 500,000 540,000 570,000 610,000 650,000 720,000 750,000 780,000 810,000 840,000 860,000 880,000 900,000 920,000
30,800
34,400 38,200 42,000 46,000 50,000 54,000 57,000 61,000 65,000 72,000 75,000 78,000 81,000 84,000 86,000 88,000 90,000 92,000
3,080
3,440 3,820 4,200 4,600 5,000 5,400 5,700 6,100 6,500 7,200 7,500 7,800 8,100 8,400 8,600 8,800 9,000 9,200
808 968 1,150 1,350 1,580 1,840 2,120 2,420 2,740
668
66.8
6.68
8.08 9.68 11.5 13.5 15.8 18.4 21.2 24.2 27.4 34.4 38.2 42 46 50 54 57 61 65
308
30.8
31%
28% 25% 22% 19% 16% 14% 12% 10% 8%
1.0
690,000
69,000
6,900
690
69
72 75 78 81 84 86 88 90 92
First Edition
Note: Subtract 1.5 to get long-term Sigma level
176
Treqna Base Manual
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.
First Edition
177
Treqna Base Manual
The Shift
Centered Distribution
LSL
6s
m m
USL
Distribution Shifted 1.5
LSL
4.5s
USL
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.
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.
First Edition
178
Treqna Base Manual
Visual Representation of ST & LT

Short-Term capability
Time = t
Time = t + x
Long-Term capability
LSL
USL
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.
First Edition
179
Treqna Base Manual
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
Non-Random Variation: Shifts Trends Cycles
First Edition
180
Treqna Base Manual
Rational Subgrouping
Rational Subgroups Common Cause Variation - Within Group
Special Cause Variation between groups
Time
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.
To create rational subgroups find:

A time period that is short enough such that you do not have This time period will differ from process to process
special causes of variation
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
First Edition
181
Treqna Base Manual
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.
First Edition
182
Treqna Base Manual
Components of Variation
[
SS T SS T Total Total
SSW
}
SST
SSB
[
SS W SS W Within Within
2 +
Samples size n=3 at different times

SS B SS B Between Between
X
(X
K N
j=1
i=1
ij
j=1
(X
K N
j=1
i=1
ij
)2
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
SSTotal : deviation of each individual point from the overall mean SSBetween: 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 X : Subgroup mean : Overall Mean
First Edition
183
Treqna Base Manual
Standard Deviation & Sum of Squares

One Sample
=
(X
n i
X 1
Where n is the sample size
(n
g
Long-Term

LT

j= 1
i= 1
(X
gn
ij
Where n is the subgroup size and g is the number of subgroups. Numerator = denotes SST. X represents overall mean.
Short-Term
=
(X
g n j= 1 i= 1
ij
ST
g (n 1
Where n is the subgroup size and g is the number of subgroups. Numerator denotes SSW
First Edition
184
Treqna Base Manual
Z Equation
The Equation
Z = Specification Limit Central Tendency Standard Deviation Variations Z Short Term(ZST ) or Long Term(ZLT) Specification Limit USL or LSL Central Tendency Mean(m) or Target(T) Standard Deviation ST or LT
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.
ZLT =
SL LT
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 =
SL T ST
ZST is bigger than ZLT, so the shift is
ZShift=ZST-ZLT
Higher the shift bigger the control issue in process. Typical assumed shift when rational subgrouping not present is 1.5 sigma.
First Edition
185
Treqna Base Manual
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.
2.5 2.0
Zshift
1.5 1.0
Poor Control; Poor Technology
Poor Control;
Poor Technology
0.5 1 2 3 4
Good Control; Good Technology

5 6
Zst
The graph to your right helps identify the issue in your process. Once you baseline your process and understand its 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.
First Edition
186
Treqna Base Manual
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 6 means ZST = 6
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 control problem.
First Edition
187
Treqna Base Manual
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 one Sample: Subset of population which has same characteristics as population.
First Edition
188
Treqna Base Manual

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
First Edition
189
Treqna Base Manual

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
First Edition
190
Treqna Base Manual
Analyze Phase
DEFINE MEASURE
Step A.1
ANALYZE
Step A.2
IMPROVE
CONTROL
Step A.3
Performance Objective Set Improvement goal for project Y -Concept of Benchmarking -Benchmarking Types -Benchmarking as process
Activities
Benchmarking
First Edition
191
Treqna Base Manual
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
First Edition
192
Treqna Base Manual
Benchmark and Baseline
Benchmark
Z Short Term
Z Long Term
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
Benchmark is the best in industry, process baseline tells where the process stands against Industry
First Edition
193
Treqna Base Manual
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 industrys 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 .
First Edition
194
Treqna Base Manual
Types Of Benchmarking
Strategic Benchmarking It 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 Benchmarking It 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 Benchmarking It 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 Benchmarking It 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.
First Edition
195
Treqna Base Manual
Types Of Benchmarking
Internal Benchmarking It 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 Benchmarking External 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 Benchmarking It 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
First Edition
196
Treqna Base Manual
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 rationale for selection initial contact development of detailed questions
Benchmarking partners
First Edition
197
Treqna Base Manual
Benchmarking Plan
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
First Edition
198
Treqna Base Manual
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)
First Edition
199
Treqna Base Manual
Summary Performance Objective

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
First Edition
200
Treqna Base Manual
Analyze Phase
DEFINE MEASURE
Step A.1
ANALYZE
Step A.2
IMPROVE
CONTROL
Step A.3
Performance Objective Set Improvement goal for project Y -Concept of Benchmarking -Benchmarking Types -Benchmarking as process
Activities
Benchmarking
First Edition
201
Treqna Base Manual
Bridging the Gap -Practical to Statistical

D M
Practical Problem
Project Y Project Charter MSA Data Collection Process Characteristics This section is the beginning of statistical problem solving process. By the end of this section we would have established Critical Xs
Statistical Problem
- Centering and Variation Establish Process Capability
Determine Statistical Goal (Benchmarking) Critical Xs
Statistical Solution
I C
Transfer function Y=f(X) Pilot/Implement Solution MSA on Vital Xs
Practical Solution
Control Plan
First Edition
202
Treqna Base Manual
Tools to be used in this section

Graphical Tools
1. 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 Xs which affect project Y
First Edition
203
Treqna Base Manual
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
First Edition
204
Treqna Base Manual
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.
First Edition
205
Treqna Base Manual
How to work with a Fishbone

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:
S a m p le F is h b o n e S a m p le F is h b o n e
M e a s u r e m e n ts M e a s u r e m e n ts M a t e r ia l M a t e r ia l P e r so n n e l P e r so n n e l
P u t in y o u r your QP uu et s tin io n Qr u he e e s t io n h ere
E n v ir o n m e n t E n v ir o n m e n t
M e th o d s M e th o d s
M a c h in e s M a c h in e s
The attached diagram is an example to the structure, which once filled gives complete picture of problem and causes.
First Edition
206
Treqna Base Manual
Prioritization Matrix
The matrix only prioritizesthe output still needs to be ratified with data and an objective Few prioritization matrices which can be used are: outlook Criterion v/s criterion
QFD Control/Impact
Why use it ?
To narrow down systematically to critical Xs by weighing each option w.r.t. its importance and effect on Y
How does it help?
Helps six sigma team to focus on Xs on priority Systemizes the approach of prioritizing as it based on on Xs weighed against each other w.r.t. to there effect on Y Quickly surfaces critical Xs from disagreements within the team
First Edition
207
Treqna Base Manual
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
First Edition
208
Treqna Base Manual
Pareto Chart
Use the tool: To filter vital few from trivial many Xs 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
First Edition
209
Treqna Base Manual
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
First Edition
210
Treqna Base Manual
Pareto Chart - Minitab

1) Collect data
P ro c e s s P ro c e s s P ro c e s s P ro c e s s P ro c e s s P ro c e s s P ro c e s s P ro c e s s P ro c e s s P ro c e s s
A A B B C C D D E E
S o ft S k ills Is s u e S o ft S k ills Is s u e 585 585 821 821 149 149 946 946 415 415
P o o r K n o w le d g e P o o r K n o w le d g e 3949 3949 5408 5408 994 994 6307 6307 3112 3112
C ll ss M tt oor reedd C aal l M oonni i
N o P r o b le m R e s o lu tio n N o P r o b le m R e s o lu tio n 390 390 548 548 99 99 631 631 277 277
2). Total results and arrange data by Segmentation Factors
S S oo f ft t S S kk i il ll lss I Iss ss uu ee P o o r K P o o r K nn oo w w l lee dd gg ee N o P r o b l e m R N o P r o b l e m R ee ss oo l luu t ti ioo nn
TT oo t taa l l dd ee f fee cc t tss 22 99 11 77 11 99 22 77 66 22 00 44 55
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 defects Frequency of defects For 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
First Edition
211
Treqna Base Manual
Pareto
Tool Utility: Useful tool identify factors/segments which contribute to the majority of the
effect
How: Minitab > Quality Tools > Pareto Chart
First Edition
212
Treqna Base Manual
Pareto Minitab Inputs

Choose Chart defects table Labels in Defects and Frequencies in Counts
First Edition
213
Treqna Base Manual
Pareto Minitab Output
First Edition
214
Treqna Base Manual
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.
First Edition
215
Treqna Base Manual
Value Analysis
Value-Added Work
For every activity in the process ask the question Is Customer Willing To Pay For Them?
Value-Enabling Work
Process steps which are non essential but enable or support the Value-Adding Tasks to be done better/faster.
Nonvalue-Added Work
This is part of process steps which is Non-Essential for production and delivery product/ service.
Answers to this question amount to Value added work which changes the product/ services physically and so is essential to the process.
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 customers 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 departments Too many supervisors Multiple 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.
First Edition
216
Treqna Base Manual
Types Of Non-Value-Added Work
Inspection (Quality Check) Failure (Internal/External) Wait / Delay Setup Initialization Inventory Movement
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.
First Edition
217
Treqna Base Manual
Workflow and Time allocation
Wait / Delay Time
Processing Time (Product Touch Time)
T1 T2+T2
Cycle Time
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 Time Total 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 dont capture. Process Time It 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.
First Edition
218
Treqna Base Manual
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.
First Edition
219
Treqna Base Manual
Analyzing Workflow: Track the unit flow

Agent fills app Agent fills app with customer with customer Agent mails app Agent mails app using USPS using USPS App received in App received in mail room mail room Mail room sorts Mail room sorts by department by department Data entry apps Data entry apps placed in data placed in data entry tub entry tub Apps transferred Apps transferred to data entry to data entry every two hours every two hours Data entry operator Data entry operator screens app for screens app for completeness completeness Is app complete ?
Yes No
Procure missing Procure missing information from information from agent agent Route to Route to Resolution team Resolution team
Order print Order print
Complete data entry Complete data entry and forward to audit and forward to audit team team Is information entered correct ?
Yes
Sort issues and Sort issues and declines declines
No
Correct Correct information information
Create policy package for Create policy package for Issued policy and letter for Issued policy and letter for declined policies declined policies
Route to Route to underwriter underwriter Can we issue policy ?

Yes
Apply postage and mail to Apply postage and mail to agent using SPS agent using SPS
No
Decline policy Decline policy
Agent receives letter // Agent receives letter package and notifies client package and notifies client
Issue policy Issue policy
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.
First Edition
220
Treqna Base Manual Value Analysis
Process 11 22 33 44 55 66 77 88 99 10 ProcessStep Step 10 11 11 12 12 13 13 14 14 15 15 16 16 17 17 18 18 Total TotalTime Time % %Total Total Time 475 100% TimeTaken Taken(min) (min) 22 22 10 10 21 21 120 120 33 44 55 33 77 99 12 12 11 22 33 21 21 240 240 10 10 475 100% U U U Value 24 5% UU U U U ValueAdd Add 24 5% Value Valueenabling enabling Rework Rework Transportation Transportation Inventory Inventory/ /queue queue Audit Audit/ /Inspection Inspection U Prep / Set U Prep / Setup up
U U U U U U U U U U U U U U U U U U U U U U U U U U
99 66 261 261 147 147 24 24 44
2% 2% 1% 1% 55% 55% 31% 31% 5% 5% 1% 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 time Note: 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: Example 4 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.
First Edition
221
Treqna Base Manual
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 1 0
# of Students # of Students
0 5 5 5 5 6 0 6 0 6 5 6 5 7 0 7 0
T e s t G ra d e s T e s t G ra d e s
7 5 7 5
8 0 8 0
8 5 8 5
9 0 9 0
9 5 9 5
1 0 0 1 0 0
First Edition
222
Treqna Base Manual
Histogram Minitab
Click on Graph > Histograms MINITAB FILE: DescriptiveStats.mtw 1.Click on Simple 2.Click Ok
First Edition
223
Treqna Base Manual
Histogram Output
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.
First Edition
224
Treqna Base Manual
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 > DOTPLOT Click on Simple, Click Ok Double Click on C1, Click ok
First Edition
225
Treqna Base Manual
Dot Plot - Minitab

Dot plot can be plotted from GRAPHS> DOTPLOT In Minitab Release 14 the Dotplot from Graphs menu gives multiple option for Single Y, Multiple Y, Single group, Multiple group and there combination.
First Edition
226
Treqna Base Manual
Dot Plot Output
When Dotplot plotted for multiple operators (groups) gives us comparative visual representation of their: Variation Center and spread
First Edition
227
Treqna Base Manual
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
First Edition
228
Treqna Base Manual
Box plot - How to Read ?

*
Outlier any point outside the lower or upper limit 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)
First Edition
229
Treqna Base Manual
Box plot - Minitab

MINITAB FILE: BoxPlots.mtw
Minitab > Graph > Boxplot
Choose One Y With Groups, Click Ok. Use C6 for Data, C7 for Subscripts Click ok
First Edition
230
Treqna Base Manual
Statistical Hypothesis Testing
First Edition
231
Treqna Base Manual
The Nature Of Statistical Problems
Desired Process
Current Process
LSL
USL
Poor Precision, Focus on reducing Variation
Desired Process
Current Process
T LSL USL
Poor Accuracy, Focus on Centering the process
While working on Six Sigma Projects we come across two types of problems with data. Variation (Spread) Centering Though 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 Xs have an effect on the Y. These tests may be testing for variance or for mean depending on what is our Y metric.
First Edition
232
Treqna Base Manual
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)
First Edition
233
Treqna Base Manual
Concept Of Hypothesis Testing

1. 2. All processes have variation. Different Samples from the same process may vary
3. Hypothesis testing differentiates between sampling error (Precision) and true process differences.
Are these Samples from the same distribution or are they truly from different processes.
First Edition
234
Treqna Base Manual
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.
First Edition
235
Treqna Base Manual
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.)
First Edition
236
Treqna Base Manual
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 >
First Edition
237
Treqna Base Manual
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: B A Ha: B > A The statement is about the population means, not the sample means.
First Edition
238
Treqna Base Manual
Decision Error
Based on the evidence there are four possible outcomes of decision we make:
Truth Truth Ho Innocent Set Free Innocent, Set Free Ha Guilty Guilty, Set Free
Verdict Verdict
Ho
Ha
First Edition
Jailed
Innocent, Jailed
Guilty, Jailed
239
Treqna Base Manual
Decision Error - Types

Truth Truth Ho Innocent Set Free Verdict Verdict Ho
Correct Decision 1 - = Confidence
Ha Guilty
Type II Error
Jailed
Ha
Type I Error
Correct Decision 1 - = Power
1 - = Probability of correctly accepting the null hypothesis (detecting no change when there is none). 1 - = 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
For any statistical testing procedure define

1. = P[Rejecting the null hypothesis when it is true] = P[ type I error] 2. = P[accepting the null hypothesis when it is false] = P[ type II error]
The symbol is called the level of significance. Usually we have set at .05, which means we have 95%(1- ) 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.
First Edition
240
Treqna Base Manual

Types of Hypothesis One Sample Hypothesis
1. Ho : = constant = T Ha : constant = T
Ho
Ha
2. Ho : 2 = constant Ha : 2 constant
In this and following slides read:

as mean and as standard deviation
First Edition
241
Treqna Base Manual

Types of Hypothesis Two Sample Hypotheses
3. Ho : 1 = 2
Ha : 1 2
4. Ho : 1 2 Ha : 1 > 2
5. Ho : 12 = 22
Ha : 12 22
1 2
First Edition
242
Treqna Base Manual

Types of Hypothesis Multi Sample Hypotheses
6. Ho : 1 = 2 = . . . = n Ha : at least one not equal
7. Ho : 12 = 22 = . . . = n2 Ha : at least one not equal
First Edition
243
Treqna Base Manual
One-sided & Two-sided Hypothesis Tests
One-Sided Test Ex. Ho : 1 < 10 Ha : 1 > 10

4 5 6
Acceptance Region 1-
7 8 9
/2
12 13 14 15 16
= 10
10
11
Two-Sided Test Ex. Ho : 1 = 10 Ha : 1 10

4 5
Acceptance Region 1- /2
6 7 8 9
/2
=101011
12
13
14
15
16
The symbols used for two-sided tests are = and . The symbols used for one-sided tests are , , < and >
First Edition
244
Treqna Base Manual
The P-Value
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
p p< < a: a: Reject Reject H H o o p p> > a: a: Accept Accept H H o o
First Edition
245
Treqna Base Manual
Hypothesis Tests Summary

Continuous Y and Discrete X
Variation - F test Compares 2 samples - HOV (Levene) Compares 2 or more samples - 2 Test Compares a sample variance to a known population variance Central Tendency 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 Test for MEDIAN - Moods Median Test tests if medians from 2 or more sample means are equal
Y = Normal
First Edition
Y = Non Normal
- HOV (Levene) Compares 2 or more samples
246
Treqna Base Manual
Normal (z) Distribution

When sample size is infinite uncertainty prediction becomes very small, hence we can use Normal distribution for predictions.
Sample size infinite
Z = X
df =
First Edition
247
Treqna Base Manual
Homogeneity of Variance (HOV)

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.
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.
First Edition
248
Treqna Base Manual
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
First Edition
249
Treqna Base Manual
HOV - Input
H H
o a
: :
2 1 2 1
=
= .05
2 2 2 2
First Edition
250
Treqna Base Manual
HOV Test - Output

H H
o a
: :
2 1 2 1
2 2 2 2
= .05
p < a: Reject Ho p > a: Accept Ho
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 Bartletts test. Bartletts 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 Bartletts test is not robust to nonnormality. Levenes Test: Use this test result when data is continuous and not necessary normal.
First Edition
251
Treqna Base Manual
2 Sample t test
HO (Null Hypothesis) : 1 = 2 HA (Alternate Hypothesis) : 1 = 2
First Edition
252
Treqna Base Manual
2 Sample t test - Minitab

Minitab File: HOV.mtw
First Edition
253
Treqna Base Manual
2 sample t - Input
Check this box if HOV Indicates that there is no statistically significant difference in Variation among the groups being tested
Click Options to change your Alternate Hypothesis
Run Homogeneity of Variance test before you indicate to assume equal variances in the input window.
First Edition
254
Treqna Base Manual
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 Agent B 40 49.58 2.85 0.35 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) : 1 = 2
The p value indicates that the two agents are statically different in their performance.
First Edition
255
Treqna Base Manual
One Way Analysis of Variance

(ANOVA)
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: 1 = 2 = 3 = 4 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.
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 for multiple 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 doesnt make any assumption about the the nature of relationship between Y and independent variable.
First Edition
256
Treqna Base Manual
One Way ANOVA - Minitab

A Test to determine if the means of the number of Queries Resolved by 4 agents is statistically different Ho: 1 = 2 = 3 = 4 Ha: at least one m different from the others
Minitab File: Anova.mtw
First Edition
257
Treqna Base Manual
One-Way ANOVA Input/ Output
One-way ANOVA: Average Queries Resolved versus Agent Source DF Agent SS MS 11.6 F P 3 9998.6 3332.9 287.01 0.000
Error 196 2276.1 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 --------+---------+---------+---------+(-*) (*-) (-*) 30.0 36.0 42.0 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 Pooled StDev = 3.408
--------+---------+---------+---------+-
An ANOVA output gives us: 1. P-value for significance 2. Confidence band to understand which ones are different
First Edition
258
Treqna Base Manual
NonParametric Test - Objectives

Understand distribution - Normality Test Possible reasons for not normal data Understand interpretation of Moods Median Test Understand HOV Test
First Edition
259
Treqna Base Manual
Non-parametric Tests
What if I dont have normal data?
When we dont 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 : Moods Median Test Spread (Variance) Capability : HOV Use Levine's test : 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 population Check for outliers
Normal data gives flexibility in terms of number of tools available for analysis. However sufficient tools are available for not normal data.
First Edition
260
Treqna Base Manual
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. Moods 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.
First Edition
261
Treqna Base Manual
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 nonnormal data. Use the non-normal tests!
First Edition
262
Treqna Base Manual
Moods Median Test

Hypothesis for the test: HO = Medians for all the groups are equal HA = At least one median is different
E.g. File: MoodsMedian Stat > Nonparametrics > Moods 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)
To perform a Moods Median Test on collected data:

Enter Y data in one column in Minitab Second column should define subgroups Choose test from Stat > Nonparametric > Moods Median Test
First Edition
263
Treqna Base Manual
Moods Median Test

Select Stat > Nonparametrics > Moods Median Test
First Edition
264
Treqna Base Manual
Moods Median Test Input

Response data or Y
Factors or Xs
First Edition
265
Treqna Base Manual
Moods Median Test - Output
Mood Median Test: AHT versus Team
Mood median test for AHT Chi-Square = 26.32 DF = 2 P = 0.000
P value
Individual 95.0% CIs Team N<= N> Median Q3-Q1 -----+---------+---------+---------+A B C 7 8 280.0 14.0 15 0 263.0 10.0 (---*--) 1 14 296.0 13.0 (-----*--) (----*----)
Confidence interval
-----+---------+---------+---------+264 276 288 300
Overall median = 279.0
Moods Median test output gives a table where it defines by group number of data items below and above groups 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.
First Edition
266
Treqna Base Manual
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
Moods Median Test
Ho : X1 = X2 = X3 =..= Xn
Ha : At least one median is different from the others
Homogeneity of variance
Ho : 12 = 22 =.. = n2 Ha : 12 22 .. n2
First Edition
267
Treqna Base Manual
Hypothesis for various tests

1-sample t test Ho : = T Ha : T 2-sample t test Ho : 1 = 2 Ha : 1 2 Paired t test Ho : Difference between the two means = 0 Ha : Difference between the two means 0 ANOVA Ho : 1 = 2 = 3 =.= n Ha : At least one mean is different from the others
First Edition
268
Treqna Base Manual
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).
First Edition
269
Treqna Base Manual
Correlation - Minitab
E.g. File: Grades.MTW Stat > Basic Stats > Correlation
First Edition
270
Treqna Base Manual
Correlation Inputs/Outputs
First Edition
271
Treqna Base Manual
Regression Minitab
First Edition
272
Treqna Base Manual
Regression Minitab Inputs
First Edition
273
Treqna Base Manual

Click the Graphs button
First Edition
274
Treqna Base Manual

Click the Results button
First Edition
275
Treqna Base Manual
Regression Minitab Outputs
Regression Analysis: Score1 versus Score2 The regression equation is Score1 = - 4.67 + 4.40 Score2 Predictor Coef SE Coef T P Constant -4.6674 0.8572 -5.44 0.001 Score2 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 P Regression 1 51.353 51.353 156.56 0.000 Residual Error 7 2.296 0.328 Total 8 53.649
First Edition
276
Treqna Base Manual
Discrete Y & Discrete X

Two Chi-Square Tests: o Goodness-of-Fit Test to test if observed data has expected distribution o Test of Independence to test if two or more proportions are different
Chi-Square statistic is calculated using formula:
2 =
i=1
f
e
Critical Value depends on two factors: o a - the confidence o df degrees of freedom for data
First Edition
277
Treqna Base Manual
Chi-Square Test2
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 2 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)
First Edition
278
Treqna Base Manual
Chi-Square Statistic
Calculation for Chi-square statistic is:
2 =
i=1
f
e
Where: n = number of categories fo = observed frequency of data fe = expected frequency of data
First Edition
279
Treqna Base Manual
Degrees Of Freedom
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.
First Edition
280
Treqna Base Manual
Chi Square Stat Test

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 Distribution Critical Value A B Area of Rejection
P-value can also be calculated using formula in excel: CHIDIST(statistic,df)
First Edition
281
Treqna Base Manual
Chi-Square Distribution
df 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 40 50 60 70 80 90 100 .2 5 0 1 .3 2 3 2 .7 7 3 4 .1 0 8 5 .3 8 5 6 .6 2 6 7 .8 4 1 9 .0 3 7 1 0 .2 1 9 1 1 .3 8 9 1 2 .5 4 9 1 3 .7 0 1 1 4 .8 4 5 1 5 .9 8 4 1 7 .1 1 7 1 8 .2 4 5 1 9 .3 6 9 2 0 .4 8 9 2 1 .6 0 5 2 2 .7 1 8 2 3 .8 2 8 2 4 .9 3 5 2 6 .0 3 9 2 7 .1 4 1 2 8 .2 4 1 2 9 .3 3 9 3 0 .4 3 4 3 1 .5 2 8 3 2 .6 2 0 3 3 .7 1 1 3 4 .8 0 0 4 5 .6 1 6 5 6 .3 3 4 6 6 .9 8 1 7 7 .5 7 7 8 8 .1 3 0 9 8 .6 5 0 1 0 9 .1 4 1 .1 0 0 2 .7 0 6 4 .6 0 5 6 .2 5 1 7 .7 7 9 9 .2 3 6 1 0 .6 4 5 1 2 .0 1 7 1 3 .3 6 2 1 4 .6 8 4 1 5 .9 8 7 1 7 .2 7 5 1 8 .5 4 9 1 9 .8 1 2 2 1 .0 6 4 2 2 .3 0 7 2 3 .5 4 2 2 4 .7 6 9 2 5 .9 8 9 2 7 .2 0 4 2 8 .4 1 2 2 9 .6 1 5 3 0 .8 1 3 3 2 .0 0 7 3 3 .1 9 6 3 4 .3 8 2 3 5 .5 6 3 3 6 .7 4 1 3 7 .9 1 6 3 9 .0 8 7 4 0 .2 5 6 5 1 .8 0 5 6 3 .1 6 7 7 4 .3 9 7 8 5 .5 2 7 9 6 .5 7 8 1 0 7 .5 6 5 1 1 8 .4 9 8 .0 5 0 3 .8 4 1 5 .9 9 1 7 .8 1 5 9 .4 8 8 1 1 .0 7 0 1 2 .5 9 2 1 4 .0 6 7 1 5 .5 0 7 1 6 .9 1 9 1 8 .3 0 7 1 9 .6 7 5 2 1 .0 2 6 2 2 .3 6 2 2 3 .6 8 5 2 4 .9 9 6 2 6 .2 9 6 2 7 .5 8 7 2 8 .8 6 9 3 0 .1 4 4 3 1 .4 1 0 3 2 .6 7 1 3 3 .9 2 4 3 5 .1 7 2 3 6 .4 1 5 3 7 .6 5 2 3 8 .8 8 5 4 0 .1 1 3 4 1 .3 3 7 4 2 .5 5 7 4 3 .7 7 3 5 5 .7 5 8 6 7 .5 0 5 7 9 .0 8 2 9 0 .5 3 1 1 0 1 .8 7 9 1 1 3 .1 4 5 1 2 4 .3 4 2 A lp h a .0 2 5 5 .0 2 4 7 .3 7 8 9 .3 4 8 1 1 .1 4 3 1 2 .8 3 2 1 4 .4 4 9 1 6 .0 1 3 1 7 .5 3 5 1 9 .0 2 3 2 0 .4 8 3 2 1 .9 2 0 2 3 .3 3 7 2 4 .7 3 6 2 6 .1 1 9 2 7 .4 8 8 2 8 .8 4 5 3 0 .1 9 1 3 1 .5 2 6 3 2 .8 5 2 3 4 .1 7 0 3 5 .4 7 9 3 6 .7 8 1 3 8 .0 7 6 3 9 .3 6 4 4 0 .6 4 6 4 1 .9 2 3 4 3 .1 9 4 4 4 .4 6 1 4 5 .7 2 2 4 6 .9 7 9 5 9 .3 4 2 7 1 .4 2 0 8 3 .2 9 8 9 5 .0 2 3 1 0 6 .6 2 9 1 1 8 .1 3 6 1 2 9 .5 6 1 .0 1 0 6 .6 3 5 9 .2 1 0 1 1 .3 4 5 1 3 .2 7 7 1 5 .0 8 6 1 6 .8 1 2 1 8 .4 7 5 2 0 .0 9 0 2 1 .6 6 6 2 3 .2 0 9 2 4 .7 2 5 2 6 .2 1 7 2 7 .6 8 8 2 9 .1 4 1 3 0 .5 7 8 3 2 .0 0 0 3 3 .4 0 9 3 4 .8 0 5 3 6 .1 9 1 3 7 .5 6 6 3 8 .9 3 2 4 0 .2 8 9 4 1 .6 3 8 4 2 .9 8 0 4 4 .3 1 4 4 5 .6 4 2 4 6 .9 6 3 4 8 .2 7 8 4 9 .5 8 8 5 0 .8 9 2 6 3 .6 9 1 7 6 .1 5 4 8 8 .3 7 9 1 0 0 .4 2 5 1 1 2 .3 2 9 1 2 4 .1 1 6 1 3 5 .8 0 7 .0 0 5 7 .8 7 9 1 0 .5 9 7 1 2 .8 3 8 1 4 .8 6 0 1 6 .7 5 0 1 8 .5 4 8 2 0 .2 7 8 2 1 .9 5 5 2 3 .5 8 9 2 5 .1 8 8 2 6 .7 5 7 2 8 .3 0 0 2 9 .8 1 9 3 1 .3 1 9 3 2 .8 0 1 3 4 .2 6 7 3 5 .7 1 8 3 7 .1 5 6 3 8 .5 8 2 3 9 .9 9 7 4 1 .4 0 1 4 2 .7 9 6 4 4 .1 8 1 4 5 .5 5 8 4 6 .9 2 8 4 8 .2 9 0 4 9 .6 4 5 5 0 .9 9 3 5 2 .3 3 6 5 3 .6 7 2 6 6 .7 6 6 7 9 .4 9 0 9 1 .9 5 2 1 0 4 .2 1 5 1 1 6 .3 2 1 1 2 8 .2 9 9 1 4 0 .1 6 9 .0 0 1 1 0 .8 2 8 1 3 .8 1 6 1 6 .2 6 6 1 8 .4 6 7 2 0 .5 1 5 2 2 .4 5 8 2 4 .3 2 2 2 6 .1 2 5 2 7 .8 7 7 2 9 .5 8 8 3 1 .2 6 4 3 2 .9 0 9 3 4 .5 2 8 3 6 .1 2 3 3 7 .6 9 7 3 9 .2 5 2 4 0 .7 9 0 4 3 .3 1 2 4 3 .8 2 0 4 5 .3 1 5 4 6 .7 9 7 4 8 .2 6 8 4 9 .7 2 8 5 1 .1 7 9 5 2 .6 2 0 5 4 .0 5 2 5 5 .4 7 6 5 6 .8 9 2 5 8 .3 0 2 5 9 .7 0 3 7 3 .4 0 2 8 6 .6 6 1 9 9 .6 0 7 1 1 2 .3 1 7 1 2 4 .8 3 9 1 3 7 .2 0 8 1 4 9 .4 4 9
First Edition
282
Treqna Base Manual
Chi-Square Distribution
d f 1 2 3 4 5 6 7 8 9 10 1 1 1 1 1 1 1 1 1 2 2 2 2 2 2 2 2 2 2 3 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 .9 9 5 .0 0 0 0 3 9 0 .0 1 0 0 .0 7 2 0 .2 0 7 0 .4 1 2 0 0 1 1 2 2 3 3 4 4 5 5 6 6 7 .6 .9 .3 .7 .1 .6 .0 .5 .0 .6 .1 .6 .2 .8 .4 7 8 4 3 5 0 7 6 7 0 4 9 6 4 3 6 9 4 5 6 3 4 5 5 1 2 7 5 4 4 .9 9 0 .0 0 0 1 6 0 0 .0 2 0 0 .1 1 5 0 .2 9 7 0 .5 5 4 0 1 1 2 2 3 3 4 4 5 5 6 7 7 8 .8 .2 .6 .0 .5 .0 .5 .1 .6 .2 .8 .4 .0 .6 .2 7 3 4 8 5 5 7 0 6 2 1 0 1 3 6 2 9 6 8 8 3 1 7 0 9 2 8 5 3 0 .9 7 5 .0 0 0 9 8 0 0 .0 5 1 0 .2 1 6 0 .4 8 4 0 .8 3 1 1 1 2 2 3 3 4 5 5 6 6 7 8 8 9 1 1 1 1 1 1 1 1 1 1 .2 .6 .1 .7 .2 .8 .4 .0 .6 .2 .9 .5 .2 .9 .5 3 9 8 0 4 1 0 0 2 6 0 6 3 0 9 7 0 0 0 7 6 4 9 9 2 8 4 1 7 1 3 2 8 1 0 4 3 8 7 1 A lp h a .9 5 0 .0 0 3 9 3 0 0 .1 0 3 0 .3 5 2 0 .7 1 1 1 .1 4 5 1 2 2 3 3 4 5 5 6 7 .6 .1 .7 .3 .9 .5 .2 .8 .5 .2 3 6 3 2 4 7 2 9 7 6 5 7 3 5 0 5 6 2 1 1 .9 0 0 .0 1 5 8 0 0 0 .2 1 1 0 .5 8 4 1 .0 6 4 1 .6 1 0 2 2 3 4 4 5 6 7 7 8 .2 .8 .4 .1 .8 .5 .3 .0 .7 .5 0 3 9 6 6 7 0 4 9 4 4 3 0 8 5 8 4 2 0 7 .7 5 0 .1 0 1 5 0 0 0 .5 7 5 1 .2 1 3 1 .9 2 3 2 .6 7 5 3 4 5 5 6 .4 .2 .0 .8 .7 5 5 7 9 3 5 5 1 9 7 .5 0 0 .4 5 5 0 0 0 1 .3 8 6 2 .3 6 6 3 .3 5 7 4 .3 5 1 5 6 7 8 9 1 1 1 1 1 1 1 1 1 1 2 2 2 2 2 2 2 2 2 2 .3 .3 .3 .3 .3 4 4 4 4 4 8 6 4 3 2 1 0 0 9 9 8 8 8 8 7 7 7 7 7 7 6 6 6 6 6
7 .5 8 4 8 .4 3 8 9 .2 9 9 1 0 .1 6 5 1 1 .0 3 6 1 1 1 1 1 1 1 1 1 1 2 2 2 2 2 1 2 3 4 5 6 7 8 9 9 0 1 2 3 4 .9 .7 .6 .5 .4 .3 .2 .1 .0 .9 .8 .7 .6 .5 .4 1 9 7 6 5 4 4 3 3 3 4 4 5 6 7 2 2 5 2 2 4 0 7 7 9 3 9 7 7 8
0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9
.3 .3 .3 .3 .3 .3 .3 .3 .3 .3 .3 .3 .3 .3 .3 .3 .3 .3 .3 .3
4 4 4 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3
7 .9 6 2 8 .6 7 2 9 .3 9 0 1 0 .1 1 7 1 0 .8 5 1 1 1 1 1 1 1 1 1 1 1 1 2 3 3 4 5 6 6 7 8 .5 .3 .0 .8 .6 .3 .1 .9 .7 .4 9 3 9 4 1 7 5 2 0 9 1 8 1 8 1 9 1 8 8 3
9 .3 1 2 1 0 .0 8 5 1 0 .8 6 5 1 1 .6 5 1 1 2 .4 4 3 1 1 1 1 1 1 1 1 1 2 3 4 4 5 6 7 8 8 9 0 .2 .0 .8 .6 .4 .2 .1 .9 .7 .5 4 4 4 5 7 9 1 3 6 9 0 1 8 9 3 2 4 9 8 9
8 .0 3 4 8 .6 4 3 9 .2 6 0 9 .8 8 6 1 0 .5 2 0 1 1 1 1 1 1 1 2 3 3 .1 .8 .4 .1 .7 6 0 6 2 8 0 8 1 1 7
8 .8 9 7 9 .5 4 2 1 0 .1 9 6 1 0 .8 5 6 1 1 .5 2 4 1 1 1 1 1 2 2 3 4 4 .1 .8 .5 .2 .9 9 7 6 5 5 8 9 5 6 3
0 0 1 2 3 3 4 5 6 6
.2 .9 .6 .4 .1 .8 .5 .3 .0 .7
8 8 8 0 2 4 7 0 4 9
40 50 60 70 80 90 100
2 0 .7 0 7 2 7 .9 9 1 3 5 .5 3 5 4 5 5 6 3 1 9 7 .2 .1 .1 .3 7 7 9 2 5 2 6 8
2 2 .1 6 4 2 9 .7 0 7 3 7 .4 8 5 4 5 6 7 5 3 1 0 .4 .5 .7 .0 4 4 5 6 2 0 4 5
2 4 .4 3 3 3 2 .3 5 7 4 0 .4 8 2 4 5 6 7 8 7 5 4 .7 .1 .6 .2 5 5 4 2 8 3 7 2
2 6 .5 0 9 3 4 .7 6 4 4 3 .1 8 8 5 6 6 7 1 0 9 7 .7 .3 .1 .9 3 9 2 2 9 1 6 9
2 9 .0 5 1 3 7 .6 8 9 4 6 .4 5 9 5 6 7 8 5 4 3 2 .3 .2 .2 .3 2 7 9 5 9 8 1 8
3 3 .6 6 0 4 2 .9 4 2 5 2 .2 9 4 6 7 8 9 1 1 0 0 .6 .1 .6 .1 9 4 2 3 8 5 5 3
3 9 .3 3 5 4 9 .3 3 5 5 9 .3 3 5 6 7 8 9 9 9 9 9 .3 .3 .3 .3 3 3 3 3 4 4 4 4
First Edition
283
Treqna Base Manual
Test of Independence
This Test is used to check independence of two or more proportions. The hypothesis used is: H0 = all proportions are equal Ha = 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:
Tele sales Team A Leads Dead Ends fo = 37 fo = 63 Total = 100 Tele sales Team A fo = 113 fo = 167 Total = 280 Total = 150 Total = 230
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.
First Edition
284
Treqna Base Manual
Test of Independence Minitab Input

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:
First Edition
285
Treqna Base Manual
Test of Independence
Minitab Input And Results
Ho: Ha:
No correlation between team and leads generated. Correlation between team and leads generated
First Edition
286
Treqna Base Manual
Tool picking guide
First Edition
287
Treqna Base Manual
Statistical Test Choices
First Edition
288
Treqna Base Manual
First Edition
289
Treqna Base Manual
Improve Phase Objectives

Develop and propose a solution

Identify Vital Xs 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
First Edition
290
Treqna Base Manual
Improve Phase
DEFINE MEASURE
Step I.1 Objective Screen for Vital Xs Vital Xs Deliverables Identify Vital Xs Activities
ANALYZE
IMPROVE
CONTROL
Step I.3 Define Improved Process Pilot Solution
Step I.2 Study Interaction between Xs Interaction between Xs, Transfer function Develop a transfer functions
Experiment with various solutions
Tools
FMEA, Process Map, Pareto, Fishbone analysis
Control Impact matrix, Regression
First Edition
291
Treqna Base Manual
Characterization of Xs
Vital Xs 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 Xs 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)
First Edition
292
Treqna Base Manual
Screening of Xs
Screening of Xs 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 Xs have interaction
First Edition
293
Treqna Base Manual
Using Pareto
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.
First Edition
294
Treqna Base Manual
Using FMEA
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.
First Edition
295
Treqna Base Manual
Improve Phase
DEFINE MEASURE
ANALYZE
IMPROVE
CONTROL
Tools
First Edition
296
Treqna Base Manual
Problem Solving Process
Practical Problem
Statistical Problem
Statistical Solution
Transfer Function Y=F(X)
Practical Solution
First Edition
297
Treqna Base Manual
Transfer Function
The goal of improve is to develop a solution using the transfer function Y = {X1, X2, X3Xn} The transfer function: Relates the vital Xs to the project Y Predicts the effect and direction of changes on Y due to changes in Xs Helps in fitting a solution which maximizes the effect on Y using all or few of the Xs
First Edition
298
Treqna Base Manual
Transfer Function
The goal of improve is to develop a solution using the transfer function Y = {X1, X2, X3Xn} Tools that give mathematical transfer function: DOE Regression GLM (General Linear Model)
First Edition
299
Treqna Base Manual
Interaction Matrix
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
Influence on Call Quality (Project Y)
+
Subject Knowledge Training
High
Finance Background
Length of the call
Low
First Edition
300
Treqna Base Manual
Improve Phase
DEFINE MEASURE
ANALYZE
IMPROVE
CONTROL
Tools
First Edition
301
Treqna Base Manual
Tolerancing
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 Xs
This process of identifying variation limits for Xs using LSL and USL for Y is called Statistical Tolerancing
First Edition
302
Treqna Base Manual
Concept of Tolerancing
Assume given transfer function Y= aX + b
Y USL LSL Y = aX + b
Tolerance for Y
X
The concept can be extrapolated to transfer function involving two Xs by imagining a three dimensions (three axes)
Tolerance for X
First Edition
303
Treqna Base Manual
More on Tolerancing
Set Tolerance based on customer VOC. E.g. Specification limits defined by customer When multiple Xs are involved. Set Tolerance for all simultaneously, specially when the interaction between Xs are involved Adjust tolerance for measurement variation
First Edition
304
Treqna Base Manual
Solution Identification
This step may be broken into three steps:
Ideas generation Solution design
Pilot solution
First Edition
305
Treqna Base Manual
Ideas Generation
By this step we have established: Vital Xs prioritized Transfer function Either we have a mathematical transfer function or we know how are Xs 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 Xs
While transfer function can exactly tells us the optimum setting for Xs, in its absence one has fair idea about what needs to be done with Xs in order to influence Y positively.
First Edition
306
Treqna Base Manual
Ideas Generation
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
First Edition
307
Treqna Base Manual
Solution Design
Selecting the best Ideas
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 A solution could be: One solution Set of solution, addressing different Xs or supplementing each other
First Edition
308
Treqna Base Manual
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 fullscale 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.
First Edition
309
Treqna Base Manual
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
First Edition
310
Treqna Base Manual
Improve Summary
Vital Xs 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 Xs
Solution Design
Generating Ideas, prioritizing ideas
Pilot solution steps in running of a pilot
First Edition
311
Treqna Base Manual
First Edition
312
Treqna Base Manual
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
First Edition
313
Treqna Base Manual
What Does Control Mean

Monitor Xs using control charts for variation Identify potential variation causes and prepare an action plan Establish plan to ensure sustenance of improved process performance
First Edition
314
Treqna Base Manual
Control Phase
DEFINE MEASURE
Step C.1 MSA on Xs Objective Acceptable gage Deliverables
ANALYZE
IMPROVE
CONTROL
Step C.3 Establish Control Plan
Step C.2 Improved Process Capability Process capability, PrePost data comparison
Control Plan
Activities
Mistake proofing, Risk management, Statistical process control Refer to step M.3 Refer to Step A.1 and A.3 FMEA & Control Charts
Tools
First Edition
315
Treqna Base Manual
MSA on Xs
Deliverable for this step: Validate measurement system for Xs Refer to step M3
First Edition
316
Treqna Base Manual
Control Phase
DEFINE MEASURE
ANALYZE
IMPROVE
CONTROL
Control Plan
Activities
Tools
First Edition
317
Treqna Base Manual
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
First Edition
318
Treqna Base Manual
Control Phase
DEFINE MEASURE
ANALYZE
IMPROVE
CONTROL
Control Plan
Activities
Tools
First Edition
319
Treqna Base Manual
Statistical Process Control (SPC)

Control charts are used to monitor variation on Xs 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 Xs can not be mistake proofed and controlled within tolerance.
First Edition
320
Treqna Base Manual
Statistical Process Control

The purpose of using statistical purpose control is to monitor Xs 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 Xs go out of control and surface the special cause performance.
First Edition
321
Treqna Base Manual
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
First Edition
322
Treqna Base Manual
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
First Edition
323
Treqna Base Manual
Differentiating Variation
Perceived Variation Common Causes Special Causes
Common Causes True Variation Special Causes
Focus on process change
Tampering of process
Under-reacting Opportunity missed for taking precautions
Focus on mistake proofing, taking precautions.
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.
First Edition
324
Treqna Base Manual
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.
First Edition
325
Treqna Base Manual
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.
+3 Zone A +2 +1 -1 -2 -3 Zone B Zone C Zone C Zone B Zone A
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.
First Edition
326
Treqna Base Manual
Minitab Tests
Test 1: One point or more points more than 3s from the center line.
1
A B C C B A UCL
Average
LCL
Test 2: 9 data points in a row on the same side of center line.

A B C C B A UCL
Average
2
LCL
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.
First Edition
327
Treqna Base Manual
Minitab Tests
Test 3: Six points in a row, either increasing or decreasing
A B C C B A UCL
Average
LCL
Test 4: Fourteen points in row, alternating up and down

A B C C B A
UCL
Average
LCL
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.
First Edition
328
Treqna Base Manual
Minitab Tests
Test 5: Two out of three points greater than 2s on the same side
A B C C B A
UCL
Average
LCL
Test 6: Four out of five points greater than 1s on the same side
A B C C B A UCL
Average
6
LCL
In the case of charts for variables, the first four tests should be augmented by Tests 5 and 6 when earlier warning is desired.
First Edition
329
Treqna Base Manual
Minitab Tests
Test 7: Fifteen points in a row within 1s from center line on either side.
A B C C B A UCL
7
Average
LCL
Test 8: Eight points in a row greater than 1s from centre line on either side
A B C C B A UCL
Average
LCL
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.
First Edition
330
Treqna Base Manual
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
Lower SL Upper SL
PROCESS B
Lower SL Upper SL
Process A under control limit and under customer specification limit
Process B under control limit but has unacceptable variation when evaluated against customer specification limits.
First Edition
331
Treqna Base Manual
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
First Edition
332
Treqna Base Manual
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 = Continuous Attribute = Discrete
First Edition
333
Treqna Base Manual
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 X-bar and R for sample of size <8 X-bar and S sample of larger size c I-MR Defect or Defective np u Defect or Defective
Yes
No
No
First Edition
334
Treqna Base Manual
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
First Edition
335
Treqna Base Manual
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
First Edition
336
Treqna Base Manual
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
First Edition
337
Treqna Base Manual
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
Xbar-R Chart of Weight

600.5 U C L=600.321 Sample M ean 600.0 _ _ X=599.548
Central line at Mean
599.5
599.0 LC L=598.775 1 3 5 7 9 11 Sample 13 15 17 19
Control Limits
U C L=2.835
Sample Range
2 _ R=1.341 1
0 1 3 5 7 9 11 Sample 13 15 17 19
LC L=0
First Edition
338
Treqna Base Manual
I-MR-R/S (Between/Within)
Three charts and three perspectives for the process
I-MR-R/S (Between/Within) Chart of Weight

601 Subgroup Mean UCL=600.763
600
_ X=599.548
599 LCL=598.333 1 1.6 3 5 7 9 11 13 15 17 19
Individual charts for sample means, control limits are based on moving range of means. Refers to stability of process location
MR of Subgroup Mean
UCL=1.493
0.8 __ MR=0.457 0.0 1 3.0 3 5 7 9 11 13 15 17 19 UCL=2.835 LCL=0
Moving range charts subgroup means. Refers to between sample variation
Sample Range
1.5
_ R=1.341
R chart or S chart showing within sample variation
0.0 1 3 5 7 9 11 Sample 13 15 17 19
LCL=0
First Edition
339
Treqna Base Manual
I-MR Chart
Used for data with no sub-grouping
I-MR Chart of Data

1
Individual V alue
0.120 U C L=0.11782 0.115 0.110 0.105 LC L=0.10268 0.100 1 5 9

1
_ X=0.11025
Individual values plotted with mean as centre line
13
17
21 O bser vation
25
29
33
37
0.016 0.012 0.008 0.004 0.000 1
M oving Range
U C L=0.00930
Moving range using alternate data values, i.e. length of 2
__ M R=0.00285 LC L=0 5 9 13 17 21 O bser vation 25 29 33 37
First Edition
340
Treqna Base Manual
Discrete Control Charts
First Edition
341
Treqna Base Manual
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)
C Chart of Data
8 7 6 Sample Count 5 4 3 2 1 0 1 3 5 7 9 11 Sample 13 15 17 19 LCL=0 _ C=2.55 UCL=7.341
First Edition
342
Treqna Base Manual
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)
U Chart of Defects
0.14 0.12 Sample Count Per Unit 0.10 0.08 0.06 0.04 0.02 0.00 1 2 3 4 5 6 7 8 9 Sample 10 11 12 13 14 15 LCL=0 _ U=0.0478
1
UCL=0.1134
Tests performed with unequal sample sizes
First Edition
343
Treqna Base Manual
np Chart
Charts defectives when sub group size is constant and based on Binomial distribution
NP Chart of Rejects
30
1
25 20 15 10 5 0 1 2 3 4 5 6 Sample 7 8 9 10 LCL=1.21 __ NP=10.8 UCL=20.39
First Edition
Sample Count
344
Treqna Base Manual
p Chart
Charts defectives when sub group size is not constant and based on Binomial distribution
P Chart of Rejects
0.14 0.12 0.10 Proportion 0.08 0.06 0.04 0.02 0.00 1 2 3 4 5 6 7 8 9 Sample 10 11 12 13 14 15 LCL=0 _ P=0.0478
1
UCL=0.1118
Tests performed with unequal sample sizes
First Edition
345
Treqna Base Manual
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)
Avoidance Risk Management Mistake Proofing Statistical Process Control 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
First Edition
346
Treqna Base Manual
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
First Edition
347
Treqna Base Manual
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)
First Edition
348
Treqna Base Manual
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
First Edition
349
Treqna Base Manual
Mistake Proofing
Helps to limit Xs 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 Xs stepping out of tolerance
First Edition
350
Treqna Base Manual
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.
First Edition
351
Treqna Base Manual
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
First Edition
352
Treqna Base Manual
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
First Edition
353
Treqna Base Manual
Control Plan Sheet

Following is a sample control plane sheet
Control Plan
Date Created: Date Revised: Data Control Reaction Sample Collection Tool Plan Size Frequency
Measure
Process Step Input/Output Specifications Target
Project Y, Xs to be measured Process step where measured
Whether input or output measure
Target for the measure Specifications and tolerances
How often the data is to be collected
Actions to be taken when the measure goes out of control
Sample size to be collected
Control charts, Mistake proofing other tools
First Edition
354
Treqna Base Manual
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
First Edition
355
Treqna Base Manual
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
First Edition
356

Six Sigma Treqna Base Manual Ed1 PDF

Caricato da

Informazioni sul documento

Titolo originale

Copyright

Formati disponibili

Condividi questo documento

Condividi o incorpora il documento

Opzioni di condivisione

Hai trovato utile questo documento?

Questo contenuto è inappropriato?

Copyright:

Formati disponibili

Six Sigma Treqna Base Manual Ed1 PDF

Caricato da

Copyright:

Formati disponibili

Treqna Base Manual

All Rights Reserved TreQna 2005

Treqna Base Manual

Page 145 Page 191 Page 197

Page 293 Page 298 Page 303

Page 317 Page 319 Page 321

All Rights Reserved TreQna 2005

Treqna Base Manual

DEFINE MEASURE ANALYZE IMPROVE CONTROL

Map project Approve project

CTQ Characteristics and standards Measurement System Analysis Data collection

Baseline process Performance objective Identify drivers of variation

Screen for vital Xs Study interaction between Xs Defined Improve Process

MSA on Xs Improved Process capability Establish control plan

All Rights Reserved TreQna 2005

Treqna Base Manual

All Rights Reserved TreQna 2005

Treqna Base Manual

All Rights Reserved TreQna 2005

Treqna Base Manual

Characteristics of a good project

All Rights Reserved TreQna 2005

Treqna Base Manual

Sources for project ideas ?

All Rights Reserved TreQna 2005

Treqna Base Manual

All Rights Reserved TreQna 2005

Treqna Base Manual

Identifying the Customer

Process Customer Output Input Supplier

External customer: E.g. Buyer of the product or services

All Rights Reserved TreQna 2005

Treqna Base Manual

Critical to Quality (CTQ)

Customer driven CTQs

Process driven CTQs

Capture customer requirement

Derive and Define CTQs

All Rights Reserved TreQna 2005

Treqna Base Manual

Voice of the Customer (VoC)

All Rights Reserved TreQna 2005

Treqna Base Manual

VOC Tools Evaluation

All Rights Reserved TreQna 2005

Treqna Base Manual

VOC Tools Evaluation

All Rights Reserved TreQna 2005

Treqna Base Manual

All Rights Reserved TreQna 2005

Treqna Base Manual

Understanding Customer Needs

All Rights Reserved TreQna 2005

Treqna Base Manual

All Rights Reserved TreQna 2005

Treqna Base Manual

Thought Process Process Flow

COPIS and SIPOC are used interchangeably.

All Rights Reserved TreQna 2005

Treqna Base Manual

All Rights Reserved TreQna 2005

Treqna Base Manual

Step2: Identify customer requirements Step3: Identify process steps