Professional Documents
Culture Documents
Symbiosis Institute of
Operations Management
“
A product found defective • 1987 - Motorola officially launched it’s Six Sigma
program as follows:
and corrected during
- Improve quality 10 times by 1989.
manufacturing had high
- Improve quality 100 times by 1991.
probability of failing during
- Achieve six sigma (3.4 DPMO) performance by 1992.
early use by customer
- Motorola won the first Malcolm Baldridge National
……1985, Bill Smith Quality Award in 1988.
9
Origin of Six Sigma
2 3,08,000
3 66,800
4 6,210
5 230
20,000 lost articles of Unsafe drinking water Seven articles One unsafe minute
mail/ hour for almost 15 minutes lost per hour every seven months
each day
5,000 incorrect Two short or long 1.7 incorrect One short or long
surgical operations per landings at most major operations per week landing every five years
week airports each day
200,000 wrong drug No electricity for almost 68 wrong prescriptions One hour without
prescriptions each year seven hours each month per year electricity every 34 years
14
Multiple meanings of Sigma
Apex Council
Champion or Sponsor
Process owner
Black Belt
Green Belt
Team Members
17
Apex Council
Top Management
• Accountable for Six Sigma business results.
• Develop a strong case for Six Sigma.
• Plan and actively participate in implementation.
• Create a vision and market “change”; Become a
powerful advocate.
• Set clear (SMART) objectives.
• Hold itself and others accountable.
• Demand specific measures of results.
• Communicate results (including setbacks).
• Helping to quantify the impact of Six Sigma
efforts on bottom line.
18
Champion or Sponsor
Senior Manager
• Oversees a Six Sigma project.
• Is accountable to the Apex Council.
• Sets rationale and goal for project.
• Be open to changes in project definitions.
• Find resources (time, support, money) for team.
• Help the team overcome roadblocks; smoothen
implementation.
• Focus on data-driven management.
• Identify and recruit other key players.
• Assist in identifying and developing training
materials.
19
Process Owner
Functional Head
• Implements solutions through Black Belts and
project teams.
• Communicate with champions and apex council. • Plan and execute training.
• Establish and adhere to a schedule for projects. • Help teams promote and celebrate successes.
• Deal with resistance to Six Sigma. • Document overall progress of Six Sigma.
Facilitator
• Six Sigma implementation experts with the ability to
develop, coach, and lead multiple cross-functional
process improvement teams.
• Use tools to quickly and efficiently drive improvement.
• Facilitate to keep team focused on the project objective.
• Ensure that the Six Sigma methods are followed.
• Help teams learn and understand Six Sigma tools and
techniques through regular project reviews.
• Responsible for the ultimate success of the project.
• Trains and develops Green Belts.
• Spread Six Sigma awareness throughout the
organization.
22
Green Belt
Process Experts
• Team members are vital for success.
• High Participation.
24
Typical Project Life Cycle and Effort
Sponsor/Champion Black Belt/Green Beltand Six Sigma Team Functional Team and Process Owner
Level of Effort
1 2
• Understanding
Organization Culture Generation of • Project identification
Organizational project ideas
• Creating the culture for Culture and selection
• Selection of projects
Excellence of Projects
• Risk Identification
• Risk assessment Risk Project • Effective ways of
Management Timeline
• Risk response managing time
• Risk Control
4 3
27
Understanding
Organization
Culture
28
Understanding organization culture
1 3 5
Drive for
Data and
Proactive perfection;
fact- driven
management Tolerate
management
Failure
30
Genuine focus on customers
Questioning why
Focusing on we do things
Proactive problem
Defining instead of blindly
management prevention rather
ambitious goals defending them.
means making than firefighting.
Being proactive habits out of what and reviewing
means acting in are, too often, them frequently.
advance of neglected
events rather business
than reacting to practices:
them.
35
Proactive management - Quiz
However, the improvement projects should be aligned to organization’s Vision, strategic goal. In order to achieve that during the early stage for
selection of lean six sigma projects and cross functional team member, it is recommended to involve the leadership team (Steering Committee)
40
Project selection
“
Some feel that project selection activities faced by Six Sigma companies.
is the Achilles heel of Six Sigma.
One quickly learns that if • Selecting appropriate projects is one of the biggest
projects are not selected challenges facing Six Sigma Teams
properly the Six Sigma initiative • Most organizations are able to identify a host of project
can be at risk: Projects don't opportunities, but the difficulty arises in sizing and
deliver the expected bottom-line
results, the organization
becomes frustrated with the
effort, and slowly but surely,
“ packaging those opportunities to create meaningful
projects. To be successful, the project selection process
must be well defined and disciplined. One process that
has proven successful incorporates a three-step
attention and resources are approach.
focused on other initiatives. • Step 1 – Establish a Project Selection Steering
Source: Ronald D. Snee “Dealing with the Committee
Achilles’ Heel of Six Sigma Initiatives,”
Quality Progress, March 2001 • Step 2 – Generate Project Ideas
• Step 3 – Assess and Prioritize Projects
41
Project selection steps
Identify Value levers in • Initial health checks • Score each project as • Assign • Evaluate projects using
the business High/ Med/Low for opportunities to evaluation criteria.
• Review of performance
benefit and effort project sponsors
• Strategic data- identification of gaps • Update Benefit / Effort.
for project
• Review plotted results
• Financial • Collection of data to definition. • Review plotted results.
quantify opportunities • Select highest priority
• Customer • Complete draft • Prioritize projects based on
opportunities for further
• Translate opportunity project charters techniques.
• Operational analysis
areas into project ideas
(Process) • Schedule project launch
43
Project selection roadmap
Strategy
Translation opportunities
Identify Policy
Initiatives
Screen Spot-
Creating Implementation
check Rank
Customer Business case
Enrich
requirements
Industry trends
Defining ROI
Business objectives Translation opportunities
Put On Hold /
Rejected
1. Divide weightage by 3 and put it 2. Multiply each of the summed weighted by the number at the
Step 2
into the answered box top. In this case 1*1,4.3*2,4.7*3,3.3*4, 4.7*5
Step 3 – Add up the total answer. In this case it is 60.33. Then divide the sum by the total of weighted sum i.e 60.33/18 which is 3.351.
The answer is the score of the project.
46
Selecting the right project
Risk
manage
ment
process
Project Risk
Maintaining a
formal document Communicate the Update the Risk
from all the above Risk to the register for any
steps - Risk stakeholders new risk identified.
Register
59
DMAIC vs DMADV
60
DMAIC Vs. DMADV
Define
project
scope
Does
Process
exist?
Measure Measure
Verify Control
61
DMAIC methodology
62
DMADV methodology
63
DMAIC approach
Measure
2
What are customer Why, when and where do
expectations of the process?
Define 1 3 Analyze
defects occur?
5 4
How can we make the Control Improve How can we fix the process?
process stay fixed?
64
Introduction to
Lean
65
History of Lean
66
The pioneer of Lean
The
TheFord
FordModel
ModelTTwas wasan anAmerican
American
car
carbuilt
builtbetween
between1908
1908and and1928
1928
bybythe
theFord
FordMotor
MotorCompany
Companyofof
Detroit,
Detroit,Michigan.
Michigan.ItItisisone
oneofofthe
the
most important cars in history
most important cars in history
because
becauseititwas
wasone
oneofofthethefirst
firstcars
cars
totobe
besold
soldfor
forvery
verylittle
littlemoney,
money,
making
makingititeasy
easyfor
forpeople
peopletototravel
travel
from
fromplace
placetotoplace.
place.
HenryFord’s
Henry Ford’srevolutionary
revolutionary
advancementsininassembly-line
advancements assembly-line
automobilemanufacturing
automobile manufacturingmade made
theModel
the ModelTTthethefirst
firstcar
cartotobe
be
affordablefor
affordable foraamajority
majorityofof
Americans.
Americans.
67
Lean origin
1908
Before: cars were built at one spot and
the workers moved from car to car.
“
All we are doing is looking
at a timeline from the
moment the customer gives
us an order to the point
when we collect the cash.
And we are reducing that
“
time line by removing the
non-value added steps
Taiichi Ohno,
Toyota Production System 1978
It is all about
waste elimination and flow creation!
70
Lean History
House of
Toyota Shingijutsu Womack
“Improved “Translated
“Flow” Flow” concepts”
1902 1930 1978
1900- 1940 -
1990
“Built in 1940 Toyota 1980 Toyota
Quality” Motor Production
Company System
Sakichi Ltd
Toyoda
71
3M – 3 Enemies (Losses)
MURI
Overburden
MURA 3M – 3 MUDA
Unevenness Enemies Waste
(Losses)
72
Waste: Muda (無駄)
are unnecessary for the customer. As a result, are unnecessary for the customer. As a result, Muda
Muda Type 2 should be eliminated. Type 2 should be eliminated.
73
The 8 Classical Wastes
Unevenness
Waiting Overproces
sing
Motion
W O Over
Production
M Types O
of
WASTE Inventory
I Waste
T S
D
Defects
Transport Skills
Overburden
Some
Effects:
Causes:
Movement of
information or
• Reverse Flow • Poor Ergonomics
materials that
• Zig-Zag process layout • Safety Hazards
does not add • Multi-level shop floors • Increased Lead time
value. This could • Non availability of proper MHE
be within or
outside the work
area.
75
Inventory
A company's Some
merchandise, Effects:
Causes:
raw materials,
and finished and
unfinished
• Bulk purchases • Huge capital investment
products which
• Huge variety of items (no • More space required
have not yet commonization)
been sold but • Improper line balancing
stored in different
stages. Having
more than the
minimum stocks
necessary for the
process.
76
Motion
Some
Effects:
Causes:
Operators /
persons making
• Zig-Zag process layout • Poor Ergonomics
movements that
• Non-availability of proper tools • Very Low productivity(silent killer)
are straining or • Work station not fit ergonomically
unnecessary,
such as looking
for parts, tools,
documents etc.
77
Waiting
Some
Effects:
Causes:
Operators /
persons standing
• No line balancing • Process delays
idle as machine
• Communication gaps • Productivity loss
cycles or • Lack of Andon systems
equipment failed
or required parts
have not arrived
yet..
78
Over Production
Some
Effects:
Causes:
Operators /
persons standing
• High setup change times • Erosion in Profitability
idle as machine
• Fluctuating demands • Increase in Inventory
cycles or • Improper line balancing
equipment failed • Inconsistency in raw material supplies
or required parts
have not arrived
yet..
79
Over Processing
Some
Effects:
Causes:
Providing or
• Lack of awareness or skills • Erosion in Profitability
creating MORE
• Process not studies • Increased process time
than the • Improper tools
customer
specification
requirement.
80
Defects
Some
Effects:
Causes:
Providing or
• Lack of skills • Increase in production cost
creating MORE
• Machine failures • Results in rework
than the • Quality errors • Results in shortage
customer • Increased scrap
specification
requirement.
81
8 kinds of waste (muda)
T Transport
I Inventory
D Defects
Unnecessary movement of items resulting in wasted efforts and Conveyor Belts for transporting raw material from warehouse to
Transportation
cost. plant
Human Intellect Overutilization or underutilization of talent. Engineer being used as operator for operating machine.
Waiting Waiting for the next process step. Waiting for your turn to meet the doctor at OPD in hospital
Toyota Production
System (TPS) was
developed by Taiichi
Ohno and Eiji Toyoda
between 1948-1975
Post World War II,
when Japanese Ohno drew lots of
industry was ideas from the West
decimated, the Toyoda and a lot of
family decided to start experimentation to
their automotive ultimately develop
company called In line with their new TPS
Toyota venture, Taiichi Ohno
was given the
assignment of scaling
up Toyota
automobiles to new
heights in line with the
American companies
87
Toyota Production system (TPS) : Pillars
TPS
Goal: Low Cost, High Quality, Short Lead Time, No Waste
Stability
88
TPS pillars
1 2 3
Just In Time
Jidoka Heijunka
(JIT)
• JIT is one of the pillars of TPS • Jidoka is one of the two pillars of TPS • Heijunka, in Japanese refers to levelization
• JIT refers to a system of production that • Jidoka highlights the causes of problems • Leveling the type and quantity of production over a
makes and delivers just what is needed, • Also referred as autonomation, i.e. fixed period of time
just when it is needed, and just in the automation with human intelligence. As it • Heijunka enables organization meet demand while
amount needed gives an equipment the ability to distinguish reducing wastes in the entire value chain
• JIT aims for complete elimination of waste between good and bad parts without operator • For instance: Assume that a toy producer receives
to achieve the best possible quality, intervention. orders for 500 of the same toy per week: 200
lowest possible cost and use of • Here, the machine detects a problem and orders on Monday, 100 on Tuesday, 50 on
resources, and the shortest possible communicates it via problem display board Wednesday, 100 on Thursday, and 50 on Friday.
production and delivery lead times Instead of trying to meet demand in sequence of
• The operators can confidently continue
performing work at another machine, as well the orders, the toy producer would use heijunka to
as easily identify the root cause and prevent level demand by producing an inventory of 100
recurrence toys near shipping to fulfill Monday’s orders.
Every Monday, 100 toys will be in inventory. The
• Once the root cause is rectified, the rest of the week, production will make a 100 toys
improvements are incorporated in the per day – a level amount.
standard workflow
89
TPS pillars
4 5
Standardized
Kaizen
Work
• The foundation of daily operation in TPS is • Kaizen is a Japanese word for the philosophy
Standardized Work that defines management’s role in
• Standardized Work sets up the best work continuously encouraging and implementing
sequence for each process. Once the best small improvements involving everyone.
sequence has been determined, it is • It is the process of continuous improvement in
always repeated in exactly the same way, small increments that make the process
thereby avoiding wastes, errors and more efficient, effective, under control, and
sustaining quality adaptable. (Refer Kaizen in detail under the
• It focuses on simplification by breaking down Lean ManagementTools)
complex processes into their sub processes
and then improving them
90
Goal of TPS
Goal of TPS
Design out
overburden Enable the Elimination of
Follow 5s at
(muri), production Participation of problems or
every step of the
inconsistency process as all resources in imperfection so
production
(mura), and flexible as the work process as to reduce
process
eliminate waste possible inventory
(muda)
91
Outcome of TPS
Outcome of TPS
Practiced by
one of the ten
Reduction in Improvement of largest Faster
Lower cost
lead time quality companies response
across the
world
92
Recognize Phase
93
Recognize Phase
Strategies for
Cost of poor
understanding Quality
customer (COPQ)
requirements Quality
Function
Deployment
(QFD)
Balance
Scorecard Voice of
(BSC) Customer
(VOC)
94
Balanced
Scorecard
(BSC)
95
What is Balanced Scorecard?
Balanced Scorecard is a
performance management
system for implementing strategic
planning and achieving
continuous improvement at all
levels of an organization. What is
Balanced
Scorecard?
BSC enables an
organization to
measure what is critical
to an organization i.e.
value drivers
96
Balanced Scorecard : An overview
Improves organization’s
performance by measuring
BSC increases focus on
strategy and results 01 what really matters Aligns 02
organization’s strategy with
employees on a daily basis
Improves communication of
the organization’s vision
Focuses on key drivers for
future performance 03 and strategy Prioritize 04
projects, products, and
services
98
Balanced Scorecard Model
The Balanced Scorecard model views the organization from four perspectives viz; Financial, Internal Business
Process, Learning & Growth, and Customer
This perspective views organizational financial performance and the use of financial resources
Return on Financial
Investment results
01 02
03
Return on Capital
Employed
100
Balanced Scorecard Model – Internal Process Perspective
This perspective views organizational performance through the lenses of the quality and efficiency related to our
product or services or other key business processes
COPQ Process
Alignment
01 02
03
Process Automation
/ Innovation
101
Balanced Scorecard Model – Learning & Growth Perspective
Employee
Turnover
This perspective views
organizational performance
through the lenses of human
capital, infrastructure, Training &
Job Learning
Learning &
technology, culture and other Expertise Opportunities
Growth
capacities that are key to
breakthrough performance
Job
Satisfaction
102
Balanced Scorecard Model – Customer Perspective
Cost
Quality
103
Balanced Scorecard Advantages
Aligns the
organization’s vision
and mission to
business strategy .
Enables better
decision making.
Balanced Improves internal and
Scorecard external
communication. Provides the
Advantages
management with a
comprehensive picture
of the processes.
Monitors organization
performance .
104
Balanced Scorecard Disadvantage
Time consuming
High
implementation
Balanced cost
Does not provide
Scorecard recommendations.
Disadvantage
creation!!!
spectacular
The ICEBERG
Nature’s most
The Iceberg of Cost Of
Poor Quality (COPQ)
Lean Six Sigma Black Belt training programme – Part 1
The ICEBERG
Nature’s most
Hidden 15-25%
spectacular
of Total Cost
creation!!!
109
Lean Six Sigma Black Belt training programme – Part 1
Only
Only 20% ofof the
the iceberg
iceberg isvisible.
isvisible.
• Hard Savings
• Soft Savings
Team Training, Inc (TTI) provides training services to a wide variety of service
and manufacturing companies. The person who answers the phone at TTI
doesn’t have the time or knowledge to answer questions from callers. He
transfers each caller to the appropriate person. Often, the “appropriate
person” is not in the office. Some clients or potential clients hang up rather
than leave voice mail messages because they need information right away.
TTI’s marketing representative prepares a contract to be signed by the client
requesting the training. Several iterations of the contract are often needed
until, finally, the contract matches the customer’s needs. Every version of the
contract is proofread by a TTI clerk before mailing it to the client.
Occasionally, a TTI representative misinterprets a client’s needs. In these
cases, the training misses the mark and the client is not pleased.
It is difficult to reach people at the client organizations. So, the TTI materials
representative estimates the number of course manuals needed for a given
training session. Of course, she over-estimates to make sure everyone has a
manual. Rush print jobs and overnight shipments of manuals are routine.
List the Failure Costs
115
Voice Of Customer
(VOC)
116
How do you know that there is a problem?
01 02 03 04
The machine was not
I am on hold for a long
working, and I have to wait long to
period before I get I am not happy with the
technician came after receive the goods that
connected to the right Quality of the product.
two hours of raising the I ordered.
executive.”
issue
05 06 07 08
There are many We had to pay a huge The cost of production
There is always a
rejections in the penalty for not being is too high; we need to
delay in fixing my
product lot you have complied to the do something to make
product.
sent to us. statutory requirement. it cost effective.
117
How do you know that there is a problem?
A customer
is someone
who
Uses your product or service A customer could be:
• External: Individuals or organizations
outside of your business who are usually
Decides to buy your product or service associated with paying money for your
products and services.
• Internal: Colleagues who receive products,
services, support or information from your
Pays for your product or service process – i.e., Engineering, Manufacturing,
Quality, Marketing.
• Regulatory: Any government agency that
has standards the process or product must
Gets impacted by your
conform to – i.e., OSHA, EPA, FDA etc.
product or service
118
How do you know that there is a problem?
Internal
Correct order delivered Customer who placed
Order Delivery
to the correct person order
External
Supplychain Finance
Collection of Customer who
Analyzed reports
information & analysis outsourced the work
119
How do you know that there is a problem?
Before you
How does
approach to a the
business customer
problem, ask view my
these questions process?
Define how the process must perform Define the operating parameters around the process
Examples: Lead time (e.g., time to get equipment), Defect Examples: Cost reduction, capital limitations, space limitations,
free (e.g., injury free, pain free) Low cost etc. development time limitations, supplier capability etc.
Validate the need for the feature or, better yet, gather the
base need.
Avoid solutions until the Improve phase.
121
Performance Need Categories:
VOC C T Qs
Key Considerations In
Collecting Customer Data:
Identify customer
Now? Gather verbatim VOC • Review existing VOC data
segments that need to and determine service • Decide on what to collect
be targeted quality issue
• Use appropriate tools to gather VOC
- Surveys
- Interviews
What? - Be a customer
- Focus group
• List your - Customer observation
customers
- Listening posts
• Define customer
segments - Competitive comparison
• Narrow the list of • Collect data
customers
• Gather VOC
125
Steps for Gathering VOC
• The first step in gathering the VOC, is The Greatest Value Can Come From a
customer segmentation. Small Portion of Your Customer Base
– All customers are not created
equal, and do not create equal value
– Avoid “squeaky wheel” syndrome
• If customers aren’t segmented, it may prove
impossible to get a single “voice,” and the
multiple voices may lead in opposite
directions.
• Customers should be segmented or
grouped according to their similar need for
products and services
• Identify and focus on the most important
segments
Total Customers Total Value
126
Customer Segmentation:
800.00%
Economic Descriptive Attitudinal
• Revenue • Geographic • Price 600.00%
• Frequency • Demographic • Value
400.00%
• Size of Customer • Product feature • Service
• Cost • Industry 200.00%
• Strategic goals
0.00%
01 02 03 # of Customers
Price Service
127
Select Sources of Customer Information
• Complaints / feedbacks /
Sources of • Client SPOC
Listening Social media handles
Customer 02 Post • Customer Service
•
•
Billing & collection
Accounts Receivable
Information Representatives
• Interviews
Research • Surveys – global survey/ tax department survey etc.
03 Methods • Focus Groups – dedicated department/ functions
• Observations from the business around
128
Pros and Cons of each method of data collection
Survey via • Can randomly select, which allows you to draw • Requires a lot of labor hours
High
phone conclusions for the entire population • Customers may be annoyed by unwanted phone calls
While we reviewed the customer feedback forms of few customers who availed our services of relocations, following were the VOCs.
Which VOC should we initiate the project for?
QFD is a
planning
QFD is a QFD provides a
QFD reduces QFD is a methodology
proven better tracking
the uncertainty technique that that organizes
technique that system for
involved in promotes relevant
can reduce future design or
product and cross-functional information to
start up time process
process design teamwork facilitate better
and costs improvements
decision
making
134
How does Quality Function Deployment works?
Relationship
Matrix
Customer
importance ratings How?
Customer wants
Importance
ratings
4 5
Establish importance
rating for each “what”
140
Quality Function Deployment : Example
7
The Weight,
Shape, Size
and Thickness
of pizza is
Evaluate competing importance to
the customer
products Here we are
considering our
competition two pizza
companies i.e. Company
A and Company B (G =
Good, F = Fair, P = Poor)
141
Benefits of Quality Function Deployment
Define
How do I analyze
these countless
VOCs ?
Low interest rate Easy application Will come to my facility Knowledgeable reps Knows about my finances
All charges clearly stated Quick decision Available when I need to talk Friendly Makes finance suggestions
Pay back when I want Can apply over phone Responsive to my calls Make me feel comfortable Cares about my business
How to built?
▪ Kano model helps to describe Categories:
which needs, if fulfilled contribute
Purpose
to customer dissatisfaction - Must be
neutrality or delight. - More the better
▪ Kano Model Identifies: - Delighters
- Must be needs – Critical to ▪ If there is insufficient data to
customer expectation. enable the classification, collect
- More is better – Critical to addition data on VOC.
customer satisfaction. ▪ Prioritize the customer needs to
- Delighter – Converting wants develop the CTQ.
to needs.
153
Kano Model
SATISFACTION
MORE THE BETTER
DELIGHTERS + (One-dimensional)
DYSFUNCTIONAL - + FUNCTIONAL
MUST BE
CRITICAL PRIORITY
• Safe arrival
• Accurate booking
• Baggage arrives with passenger
• 99 per cent system uptime
-
DISSATISFACTION
155
Application of the Kano Model
Frame the questions carefully while gathering information from the customers:
Dysfunctional
Customer Requirements
1. Like 2. Expect 3. Neutral 4. Live with 5. Dislike
1. Like Q A A A O
2. Expect R I I I M
Functional 3. Neutral R I I I M
4. Live with R I I I M
5. Dislike R R R R Q
Legends:
A: Attractive O: One-dimensional
M: Must Be Q: Questionable result
R: Reverse I: Indifferent
157
Example
158
Determine Critical To
Quality (CTQ)
159
Example: Translating VOC to CTQs
• Quick response.
Specific
need (CCR) • User friendly forms.
• System and training content availability.
01 03
If we can’t
accurately
If we don’t know If we can’t control
measure
much about it, we it, we are at the
something, we
can’t control it. mercy of chance.
really don’t know
much about it.
02
162
Examples
We are not profitable on the production; the costs are over running the
estimated budgets.
We could not process the loan documents within the time promised to
customer
Boundary - Boundary -
S (Start of (Completion
Process) of Process) C
U O U
P I
U S
P N
T T
L P PROCESS P O
I U
U M
E T
T E
R R
Supplier (S) Input (I) Process (P) Output (O) Customer (C)
User Department Stock List Generate Purchase Requisition Purchase Requisition User Department HOD
1. Approved Purchase
Procurement
Requisition Raise Request of Quotation Request for Quotation Approved Vendors
2. List of Approved Vendors
Procurement Shortlisted Quotation Select preferred vendor Quotation of preferred Vendor Procurement
Procurement Quotation of preferred Vendor Prepare purchase order Purchase order Signing Authority
Signing Authority Purchase order Approve purchase order Approved purchase order Preferred vendor
168
Project Charter
169
What is a Project Charter?
Does project “Y” What is the What is the Clearly defining Key milestones/ Who are the key
link to business problem/ estimated target project’s In timelines/detailed resources ?
Y’s? opportunity of for scope and Out plan What will be the
improvement? improvement? of Scope roles of BBs/
Sponsor/ MBB’s
171
Business Case
Frame the questions carefully while gathering information from the customers:
What is wrong or not When and where What’s the
How big is the
meeting our does the impact of the
problem?
customer’s needs? problem occur? problem?
Example 1
“There has been late
payments done by some of Poor Statement
our clients and its leading to Because our customers are dissatisfied with our service, they are late paying their bills.
mismanagement of the
Improved Statement
cashflows in the
organization.” In the last 6 months (when) 20% of our repeat customers – not first timers (where) – were over 60
days late (what) paying our invoices. When surveyed, all of these customers reported extreme
“As observed from the credit dissatisfaction with our service (what). The current rate of late payments is up from 10% in 1990 and
report (how did you know) represents 30% of our outstanding receivables (how big). This negatively affects our operating cash
flow (impact).
for the last 3 months (when),
12% of our customers are
late, by over 45 days in Example 2
paying their bills (what). This Poor Statement
represents 20% (magnitude)
Customers are unable to access the call center half the time leading to high revenue losses.
of our outstanding
receivables & negatively Improved Statement
affects our operating cash During the year 2003, (when) 40% of our customers (extent) were unable to access the call center at
flow (consequence).” the first attempt (what). This causes dissatisfaction to our customers and a loss of revenue
opportunities to the organization (impact).
174
Goal Statement
Example 1
“There has been late
payments done by some of Poor Statement
our clients and its leading to Because our customers are dissatisfied with our service, they are late paying their bills.
mismanagement of the
Improved Statement
cashflows in the
organization.” In the last 6 months (when) 20% of our repeat customers – not first timers (where) – were over 60
days late (what) paying our invoices. When surveyed, all of these customers reported extreme
“As observed from the credit dissatisfaction with our service (what). The current rate of late payments is up from 10% in 1990 and
report (how did you know) represents 30% of our outstanding receivables (how big). This negatively affects our operating cash
flow (impact).
for the last 3 months (when),
12% of our customers are
late, by over 45 days in Example 2
paying their bills (what). This Poor Statement
represents 20% (magnitude)
Customers are unable to access the call center half the time leading to high revenue losses.
of our outstanding
receivables & negatively Improved Statement
affects our operating cash During the year 2020, (when) 40% of our customers (extent) were unable to access the call center at
flow (consequence).” the first attempt (what). This causes dissatisfaction to our customers and a loss of revenue
opportunities to the organization (impact).
176
Project Scope
Project Scope helps us to understand the start and end point for the process and also gives an
insight on project constraints and dimensions. It’s an attempt to define what will be covered in the
project deliverables. Scoping sharpens the focus of the project team & sets the expectations right.
There are two types of scoping:
• Longitudinal Scoping
• Lateral Scoping
Absence of proper scoping may result in the team losing interest in the project. Project becomes
difficult to implement. Even after implementation, the desired benefits are not seen. Team focuses
on trivial pain areas and missing out the real ones. Process selected is too broad to handle.
• Longitudinal Scoping: Longitudinal scoping is done on the length of the process e.g. – From the
receipt of PO till the delivery at the distributor's go-down
• e.g. – From the time of customer reporting the complaint till final satisfaction confirmation.
Mostly the ‘start’ & ‘end’ points are fixed.
• Lateral Scoping is done on the breadth of the process e.g. – All dispatches from North & South
regions e.g. – Calls received during general shift
177
Project Timelines / Milestones / Project Plan
• Head of the business • Coaches the team • Project Leader • Participates in the
unit / function • Help champions • Selects team members meetings
• Project owner • Help plan meetings • Plans team meetings • Helps execute the project
• Barrier remover tasks
• Observe team’s • Leads meetings
• Funds project performance • Implements the
• Make recommendation
recommendations or
• Resource aligner • Reviews projects with or changes
changes
• Project Reviewer Business unit head
• Army of people driving
the project with Project
Leader
179
Roles and Responsibilities of Team Members
Six Sigma
Name 1 A A A A A
champion Approver of the team
A decisions
Six Sigma
Name 2 Deployment A A A A A
leader Resource to the team, one
whose expertise and skills
Name 3 Master Black Belt A A A A A R may be needed on an ad-
Name 4 Process Manager A I I A A hoc basis
Name 10 Trainer M M R M R
180
Sample Project Charter
From last 6 months the avg. TAT of discharge for TPA & corp. patients has Hospital earn 35 - 40% of the revenue from TPA and Corporate Patients. But from last 6
increased to 5 hours. This delay leads to huge customer dissatisfaction. If unsolved months the avg. TAT of discharge process for TPA & Corp. patients has increased up to
it will lead to a revenue loss of INR 40,00,000 per annum and loss of reputation. 288 mins. This delay has a cascading effect on the further patient transfers which leads to
a revenue loss of INR 40,00,000 per annum and huge customer dissatisfaction. By doing
this project of 50% reduction in avg. discharge TAT will lead to a potential profit of INR
40,00,000/- per annum.
To reduce the Average TAT for discharge of TPA & Corp. patients by 50% i.e., Start: Consultant / Medical Officer writes discharge Order in Patient's file (TPA &
from 288 mins. to 150 mins. by July 2010. Corp.)
Avg. TAT Hours 288 150 Stop: Patient vacates the Room
Completion Schedule
All other patient discharge process Define 25th April 2010 24th April 2010
• Project depends on a cross-functional team and assumes that team members Improve 28th July 2010 27th July 2010
would support the project.
• There would be no special causes like any kind of pandemic or so. Control 20th August 2010 28th August 2010
181
Summary
“
Lean Six Sigma Black Belt training programme – Part 1
A definition of
An what success will
A
An understanding of look like that has
comprehensive A Team Charter
understanding of the process with been agreed on
Project case with
the Voice of the a SIPOC and by the team
with a Project responsibilities members and
Customer scope
Charter Sponsors / Apex
boundaries
Council.
187
Define phase recap
1 2 3
When there is no Which of the following Process Mapping
process, which is not a part of Project during the define
approach should Charter? phase must be
you avoid? a) Project Scope mapped as the
a) DMAIC way you want the
b) Business Case process to be?
b) DMADV c) Potential Causes a) True
c) IDOV d) Problem / Opportunity b) False
d) None of the Statement
above e) Project Timelines
189
Measure phase
190
Measure Phase Roadmap
01 02 03 04 05
Map the Process Plan for data Collect the data Calculate the Estimate Potential
collection and and measure it Process Sigma Savings
Validate the Level
Measurement
System
191
Learning outcomes
Process is a collection of activities that takes one or more inputs and transforms them into outputs
that are of value to the customer.
Inputs outputs
Supplier(s) Customer(s)
194
Why map the process?
No
Yes
196
Standard symbols used in Process Flow Diagram
MS
MS Visio/Excel
Visio/Excel Gliffy
Gliffy yEd
yEd CADe
CADe OmniGraffle ––
OmniGraffle
/PowerPoint/
/PowerPoint/ Specially
Specially for
for
Word
Word Mac users
Mac users
01 02 03 04 05
It is very important Implementing quick These quick wins Important point to All such
for a Lean Six wins gives can be any one or remember is that observations need
Sigma Project confidence to more of the either; Quick Wins cannot to be recorded and
Team to identify management that • A standard change the work planned for
few “Quick Wins” team is doing work checklist for done nor can they implementation in
based on the “AS- and also they help ensuring nothing change the Improve Phase.
IS” process map. to silence the is left unnoticed sequence of doing
These quick wins critics. • A standard work, because that
need to be such template for any would impact the
that they can format that CTQ and in that
implemented records data or case the Define
immediately. information Phase becomes
• A SOP for doing redundant.
repetitive kind of
work by many
operators
202
Data Types
203
What is Data?
02 Percent cream content in milk bottles (comes in four bottle container sets)
04 Number of blemishes per square yard of cloth, where pieces of cloth may be of variable
size
Time spent on revising the Tax Time spent on end to end tax
returns returns calculation for correct files
How?
What?
What method of sampling to be used. (Simple
Operational definition should answer this.
Random/ Stratified/ Systematic)
When? Who?
Frequency of data collection. (once in a day/ per The operator who is responsible to collect the
shift/ once in a week) data for you.
215
Sampling techniques
The Six Sigma team would always face a question ‘How much data do we need to have a valid sample?’ Though an important
part of data collection is to obtain a sample of reasonable size, it is one of many questions to be addressed during the
planning and development of a data collection strategy. Sample size is just one aspect of a valid data collection activity.
The validity of the data is impacted by many things: For example, operational definitions, data collection procedures and
recording.
In process improvement, there are several questions to keep in mind relative to sampling:
• Is the data representative of the situation or is bias possible? • What are the key considerations for either a process or
• Why am I sampling? To improve or control a process or to population situation?
describe some characteristic of a population? • What is the approach to sampling (e.g., random, systematic,
etc.) and approximately how many to sample.
216
When to Sample?
Time availability.
What is Sample?
The small number of items taken from the population to
make a judgment of the population is called a Sample.
The numbers of samples taken to make this judgment is
called Sample size.
218
Sampling Techniques
• Our day-to-day working is being impacted by more and more data. We have become a data driven society.
• In business and industry, we are using data in more ways than ever before.
• Today manufacturing companies gather massive amounts of information through measurement and inspection.
When this measurement data is being used to make decisions regarding the process and the business in general it
is vital that the data is accurate. If there are errors in our measurement system, we will be making decisions based
on incorrect data. We could be making incorrect decisions or producing non-conforming parts. A properly planned
and executed Measurement System Analysis (MSA) can help build a strong foundation for any data-based decision-
making process.
Definition
Before we dive further into MSA, we should review the definition of a measurement system and some of the common
sources of variation. A measurement system has been described as a system of related measures that enables the
quantification of characteristics. It can also include a collection of gages, fixtures, software and personnel required to
validate a particular unit of measure or assess the feature or characteristic being measured.
• The sources of variation in a measurement process can include the following:
• Process – test method, specification
• Personnel – the operators, their skill level, training, etc.
• Tools / Equipment – gages, fixtures, test equipment used and their associated calibration systems
• Items to be measured – the part or material samples measured, the sampling plan, etc.
• Environmental factors – temperature, humidity, etc.
All these possible sources of variation should be considered during Measurement System Analysis. Evaluation of a
measurement system should include the use of specific quality tools to identify the most likely source of variation. Most
MSA activities examine two primary sources of variation, the parts and the measurement of those parts. The sum of
these two values represents the total variation in a measurement system.
Total Variation = Part to Part Variation + Measurement system Variation.
225
What is measurement system variation?
• Measurement system variation is the variation that occurs when you measure something. Specifically, measurement
system variation is the sum of variation from repeatability and reproducibility.
• Any component of a measurement system, such as a gage, a procedure, and software, can be a source of variation.
Measurement system
Repeatability Reproducibility
variation
226
Sources of Process Variation
Overall Variation
Part-to-Part Variation: The variability in measurements across different parts. Ideally, differences between parts (Part-
to-Part) should explain most of the variability.
Measurement System variation: Measurement system variation is all variation associated with a measurement
process. Potential sources of variation include gages, standards, procedures, software, environmental components,
and so on. Like any other process, a measurement system is subject to both common-cause and special-cause
variation. To control the measurement system variation, you must first identify the sources of the variation, then you
must either eliminate or reduce the various causes. Measurement system variation is split into two sources:
Repeatability: Repeatability is the variation due to the measurement device. It is the variation that is observed when
the same operator measures the same part many times, using the same gage, under the same conditions.
Reproducibility: Reproducibility is the variation due to the measurement system. It is the variation that is observed
when different operators measure the same part many times, using the same gage, under the same conditions.
Reproducibility can be split into two sources:
• Operator: The variability in measurements due to different operators.
• Operator-by-Part: The variability in measurements due to different operator/part combinations after considering part
and operator separately.
228
Why Perform MSA
• An effective MSA process can help assure that the data being
collected is accurate and the system of collecting the data is
appropriate to the process.
• Good reliable data can prevent wasted time, labor and scrap in a
manufacturing process.
• To verify the adequacy of measurement system for Y when
establishing process baseline.
• To verify the adequacy of measurement system when verifying
causes.
• To verify the adequacy of measurement system when verifying
solutions.
• To verify the adequacy of measurement system when controlling
the X’s.
229
Why Perform MSA
Example
• A major manufacturing company began receiving calls from several of their
customers reporting non-compliant materials received at their facilities sites.
The parts were not properly snapping together to form an even surface or
would not lock in place.
• The process was audited and found that the parts were being produced out of
spec. The operator was following the inspection plan and using the assigned
gages for the inspection. The problem was that the gage did not have
adequate resolution to detect the non-conforming parts.
• An ineffective measurement system can allow bad parts to be accepted and
good parts to be rejected, resulting in dissatisfied customers and excessive
scrap. MSA could have prevented the problem and assured that accurate
useful data was being collected.
230
Validate Measurement System - Objective
To determine what percentage the Total Observed Variation is due to Measurement Device and Measurement Method
in addition to the true Part to Part variation.
To statistically verify that the current measurement system provides:
• Unbiased results
• Minimal variability within the measurement system.
• True representative values of the factors being measured
Gage system error within a measurement system is the sum of:
Accurate
Inaccurate Bias
Data Classification:
• Before analyzing the data and the gages, tools, we
must determine the type of data being collected in
sample. The data could be Variable data or Discrete
one.
• Attribute data is classified into specific values, whereas
Variable or continuous data can have infinite number of
values.
235
Gage R & R
Number of Distinct
Less than 5 Not Acceptable
Categories
237
Variable Gage R&R using Minitab
The number of distinct categories is less Measurement System does not meet the
than 5, therefore it is not acceptable. criteria as the SV% and Contribution %
exceed the criteria
239
Breaking down overall variation
100%
Overall Variation (source 89.33+10.67)
89.33%
100%
Part-to-Part Variation Measurement System Variation (source 89.33+10.67)
2.19% 7.56%
Variation due to
Variation due to Repeatability
Reproducibility
2.19% 5.37%
While 100% is the most desirable result in Attribute Agreement Analysis , the following guidelines are frequently used:
Kappa Guideline
0.90 to 1.00 Acceptable
0.80 to 0.90 Marginal
Less than 0.80 Not Acceptable
Key Points
01 02 03
Before collecting new data, evaluate the Before using existing data, try to estimate Don’t delay a Six Sigma project due to a
gage using MSA for either variable or the trustworthiness of the data. poor MSA. Keep the project moving with
attribute data. non numerical analysis wherever possible,
such as Process Analysis, Waste
Identification, identification of Non-Value-
Added Activities, and improve the gage as
the project goes forward.
243
Measure the data
244
Measure the data
Data measurements
• Mean • Range
• Median • Variance
• Mode • Standard Deviation
245
Measures of Central tendency
22, 26, 26, 31, 33, 37, 37, 42, 52, 52, 52, 57, 59
Mean or Average
The sum of the values in a data set divided by the X = 40.5 days
number of values.
Mode
Mode = 52 days
The most frequently occurring data value.
Median
The middle observation in the data set that has Median = 37 days
been arranged in ascending or descending order
246
Measures of Central tendency
22, 26, 26, 31, 33, 37, 37, 42, 52, 52, 52, 57, 59
Range
The largest data value minus the smallest data Range = Max - Min Range = 37
value.
Variance (s2)
2
s2 = (x - x )
The average deviation of all data values s2 = 162.67
from the mean. n - 1
25
20
20
Frequency
15
12
10
10
5 4 3
1
0
250
Example: Histogram
For each of 50 placements, the time (in days) it took to place a person in the position was recorded.
22 26 30 19 22 31 34 29 28 18
16 22 31 24 26 36 28 33 36 24
26 27 35 14 26 30 33 26 31 36
28 33 18 26 29 30 22 30 24 31
27 21 28 35 32 28 33 28 23 25
251
Example: Histogram
A histogram is a bar graph in which data are grouped into classes. The height of each bar shows how
many data values fall in each class.
25
20
20
Frequency
15
12
10
10
5 4
3
1
0
10-15 15-20 20-25 25-30 30-35 35-40
Days
252
Histogram – Number of Bins
Bell Shaped
Normal Distribution. This what we want to see in our Histogram`s
Bell Shaped
Double-Peaked
Shows that data has been gathered from two different sources. These are two
Normal Distributions, but from two different machines, shifts, production lines or
operators. Also, if this is representative of Vendor data, then it is possible that
Double-Peaked previous rejected lots are being mixed with new lot or vendor is outsourcing
production.
Plateau
This type of distribution demonstrates lack of Standard Process.
Plateau
254
Shapes of Data Sets
Truncated Skewed
Comb
Isolated-Peaked
This type of distribution demonstrates the existence of special incidents or causes in the process.
The figure on the left signifies the normal everyday operations of the process illustrated through the
Normal Distribution. The smaller figure on the right demonstrates the excess variation in the
process on specific days only, due to some identifiable special condition that does not exist or
Isolated-Peaked occur every day. It is a sporadic incident that occurs once in a while and needs to be addressed
through structured Corrective and Preventive actions. The basic process need not be tampered.
Only the specific sporadic cause needs to be addressed.
255
What is Normal Distribution?
01 02 03
03 It is symmetric around the mean: two halves of the curve are the same (mirror images).
Since many process outputs have this shape, the properties of the normal curve can be used to make
predictions about the process population.
258
Properties of Normal Distribution
99.73%
259
Box plot
* Outlier
*
*
Upper Adjacent Value
Interpretation:
• Week 1 median is 4.985, and the interquartile range is
4.4525 to 5.5575.
• Week 2 median is 5.275, and the interquartile range is
5.08 to 5.6775. An outlier appears at 7.0.
• Week 3 median is 5.43, and the interquartile range is 4.99
to 6.975. The data are positively skewed.
• The medians for the three weeks are similar. However,
during Week 2, an abnormally wide pipe was created,
and during Week 3, several abnormally wide pipes were
created.
263
Check the data
distribution
264
Test for Normality
A normal curve originates from a histogram. A histogram is a frequency distribution chart showing the number of times
a given value of the parameter we are trying to measuring occurs.
Minitab uses the Anderson-Darling test to determine if a set of data can be treated as normal data.
Interpreting the P-value:
The P-value is the probability of getting the particular sample if the population is normal.
P-value < 0.05 means that the chance of getting this sample from a normal population is very small (less than 5%).
Save
Project Redo, Undo
Open
Project
269
Worksheet Elements
Stack Data
• Stack Data
• Go to Minitab
• Select: Data>Stack Columns
• Double click C1, C2 and C3 and the data is put in the
variables box
• Click OK
271
Combine data into single column
Stack Data
Unstack Data
Unstack Data
In a project, you can perform analyses, and generate graphs. Projects contain one or more worksheets.
Project (.MPX) files store the following Worksheet (.MWX) files store the
items: following items:
Also enables the six sigma team to determine the percent of the
product/service not meeting the customer requirement.
281
Measuring the Capability
An unstable process...
100% of the individual data values are NOT within the
spread of natural variation (+3 ) of the process.
Few data values fall beyond the +3 limits and are
referred to as “OUTLIERS”.
…is unpredictable
284
Common Cause vs. Special Cause
Lower Upper
Specification Specification
Limit Limit
287
Examples of Stable and Capable processes
LSL
Process is Stable and Capable
USL
Process is Stable
but not Capable
LSL
USL
288
Control Limits vs. Specification Limits
Upper Control Limit Control Limits are statistical bounds (the natural
bounds of the data) used to determine process
stability.
Statistically Control Limits are equivalent to + 3 s.
Control limits are determined by the data (voice of the
Lower Control Limit
process).
LSL USL
289
Methods to Calculate Process Capability
Process capability
LSL USL
VOC VOP
LSL USL
L J L
292
Stable and Capable Process
LSL USL
Individual
Measurements
6s
A B
C D
6s
6s 6s
LS L USL
L S L A Very U S L LSL Highly USL
Capable Process Capable Process
Capable Process
295
Steps for evaluating process capability for variable data
01 02 03 04
Assure the data is Estimate the average and Determine the process’ Quantify process
normally distributed. standard deviation of the potential capability. performance.
process.
296
Process variation vs Specification
Let us assume that for a normally distributed stable process, the average is 178.6 and the standard
deviation is 3.6. The process target is 171, USL = 182, LSL = 160.
Process Variation
USL = 182
LSL = 160
x − 3s x + 3s
297
Determine the potential process capability
LSL USL
Process is just about meeting the
Cp = 1 specification but centering needs
attention.
The Cp index reflects the potential of the process if the average were perfectly centered between the specification limits.
Cp = USL - LSL
6s
Cp = 182 - 160
6s
To estimate the percentage of product / process that falls outside the specification limits, we compute Cp (upper) and
Cp (lower).
USL = 182
LSL = 160 167.8 178.6 189.4
The Cp index reflects the potential of the process if the average were perfectly centered between the
specification limits.
Z = USL - X Z = X - LSL
(upper)
s (lower) s
The Cp index reflects the potential of the process if the average were perfectly centered between the
specification limits.
Z = 0.94 Z = 5.17
(upper)
(lower)
Unlike the Cp, the Cpk index takes into account off-centering of the process. The larger the Cpk index, the
better.
6s 6s
CP = 1 CP = 1
CPK = 1 Cpk < 1
303
What is Sigma?
Plus or minus one standard deviation around the mean is about 68% of the total process output.
304
What is a Six Sigma process?
LSL USL
MEAN 1 2
-3 -2 -1
3
• Conceptually, for Normal Distributions, the Sigma Level is equal to the number of std deviations that can be fit between
mean and either specifications.
• Since in this case we can fit only one std dev between mean and specs, the sigma level is 1 sigma.
• Correspondingly as per property of normal distribution, the percentage of output meeting customer
requirement is 68%
306
Properties of Normal Distribution
LSL USL
-6 -5 -4 -2 -1 1 2 4 5 6
-3
MEAN 3
• In this case , we can fit 3 std dev between mean and secs, so sigma level = 3 sigma.
• Percentage of o/p meeting customer req = 99.73%
307
Properties of Normal Distribution
• In this case , we can fit 6 std dev between mean and secs, so sigma level = 6 sigma.
• Percentage of o/p meeting customer req = 99.99967%
LSL USL
• 99.99967%
• 100 – 99.99967
• 0.00034%
• DPMO = 0.00034*10^6/100 = 3.4
• 6sigma = 3.4 DPMO
-6 -5 - 4 -3 -2 -1 1 2 3 4 5 6
MEAN
308
Fundamentals of Process Improvement
The Anderson-Darling test's p-value at 0.892 indicates that the data follows a normal distribution.
311
Determine
Process
Capability
Discrete Data
312
Defect
Defect
A defect refers to a specific quality characteristic for an
item, such as a tear, scratch, or discoloration. A defect
indicates that the product result is not entirely as
intended.
For example, analysts at a textile company inspect towels for tears,
pulls, or improper stitching and record the number of defects. Each
towel can have 1 or more than 1 defect, such as tear; pull; stitching
error etc.
For example, a car is one finished unit in a process. A car also
contains many different areas that are assembled to create a finished
vehicle. Any of these areas – the seats, the dashboard, the engine,
the exhaust system, etc. – could have defects.
313
Defective
Defective:
A defective refers to whether the overall status for an
entire item is acceptable or not. Therefore, the data is
often of the form yes/no, pass/fail, or defective/non-
defective. Because an item may have many quality
characteristics, it may have many defects, but the item
itself is either defective or not defective.
For example, an analyst inspects a sample of bulbs from a supplier
and counts the number of broken bulbs in each sample.
Again, using the car scenario, this means that 10 cars can have a
maximum 10 defective units, because each car represents one unit.
314
Attribute Data - DPMO
315
Example
• Department: Purchase
• Defect: Incorrect entry
• Total Defects: 56
• Unit: Each MIS report (monthly)
• No of units inspected: 5,000
• DPU (defects per unit) = Total Number of defects / Total number of units inspected
• DPU = 56 / 5000 = 0.0112
• PPM = DPU * 10^6 = 11,200
317
Attribute Data – PPM and DPMO
Opportunities
2 3 2
per Pad
Opportunities
Per 100 Pads
Number of Defects
7 6 9
(in 100 Pads)
Defects Per
Opportunity (DPO)
Two ways to save an organization money are cost avoidance and cost savings. Although these two concepts are similar,
they also include important differences that can impact your budget.
Refers to strategies that prevent a business or Cost savings are expense mitigation strategies that
organization from spending unnecessary money in the lower existing spending or debt quantities. Cost
future. Since the expenses saved using cost savings are usually tangible and measurable. Cost
avoidance are usually hypothetical, they don't typically savings measures will appear in documents such as
appear in documents such as a budget or financial your budget and financial statements since they affect
statement. If a cost avoidance measure doesn't work, existing line items, when comparing one accounting
it might appear on a financial statement as the period to the next.
additional expense that you were unable to avoid.
By preemptively solving the need for future spending,
you can reduce future financial pitfalls and challenges.
324
Example
• Some data sets are not normally distributed. For example, some measures have zero or some other constant as a
physical block.
• Examples: Time, concentricity, warpage, flatness
327
Non-Normal Process
Sometimes data sets are not normal because the process is not yet in statistical control.
Average Chart
UCL
LCL
Range Chart
UCL
When a data set is naturally non-normal, we can still make predictions about the population.
In such a case, we must either…
Transform the data (for example, take the log of every data value) to achieve a Normal Distribution. Use some other type of
distribution that approximates the actual shape of the data.
329
What is data transformation?
The formula for Box-Cox transformation is Common lambdas ( 𝛌𝛌) for data transformation:
332
Johnson Transformation
Processing Time
7.84 1.85 10.04 14.15
• If P < 0.05, try to transform the data 42.4 26.3 4.88 1.53
Processing Time
• The P value is < than 0.05, the data does not fall 7.84 1.85 10.04 14.15
60
50 19.59 1.37 13.32 7.83
40
30 6.36 10.71 5.29 11.97
20
8.7 7 2.46 6.62
10
5 6.06 5.39 5.16 22.44
Processing Time
Stat > Quality Tools > Johnson Transformation 7.84 1.85 10.04 14.15
The P value for the transformed data is > than 19.84 5.38 18.58 4.54
Processing Time
Stat > Control Charts > Box Cox Transformation 7.84 1.85 10.04 14.15
Minitab calculates the optimum value wherein the 19.84 5.38 18.58 4.54
Once you have the data in the new column, run the Test of Normality to check for the distribution
Stat > Basic Statistics > Normality Test
99.9
Mean 1.901
99 StDev 0.835
4
95 N 100
90 AD 0.345
80 P- 0.479 The P value for the transformed data is greater
Value
70
60 than 0.05, the data now falls a normal distribution.
50
40
Percent
30
20
10
5
0.1
-1 0 1 2 3 4 5
Processing TimeNew
339
Non-Normal data with Weibull transformation
• If we are unable to find a transformation that successfully converts non-normal data into a normal distribution, then
can use the Weibull distribution, directly, to estimate the percentage of product beyond specifications.
• The Weibull is a family of distributions
• The Weibull has two parameters
‒ 𝛃𝛃 or the “shape” parameter
or the “scale” parameter
• Minitab estimates these two parameters from the data set that is being analyzed.
• When 𝛃𝛃 1, the Weibull distribution is exponential. When 𝛃𝛃 3.5 the Weibull distribution approximates the
Normal distribution.
340
Using Non-Normal data
Is there a No
Is P-value special cause Transform the data. Is
> 0.05? No p-value > 0.05?
variation
Yes
Yes
Eliminate special cause Is data
(5%) with risk mitigation normal?
plan
• The team has defined the operational definition of the CTQ and is
understood by all involved in data collection
• The team has agreed upon the key measurements and come up
with a baseline measurement of the process performance.
• The team has analyzed measurement systems and identified any
issues that might contribute to the analysis errors
• Where possible, the team has corrected the measurement system
to remove the error risks
• The team has measured the process data in terms of Central
tendency and Dispersion
• The team has validated the data distribution and calculated the
process Sigma level.
• The sponsor or champion has reviewed and signed off on all
elements of the measure phase.
342
Measure Phase Recap
The hierarchy of Control Limits Vs Customer Specifications for a process that is both stable and capable
should be?
a. Mean; UCL; USL; LCL; LSL c. USL; UCL; Mean; LCL; LSL
b. USL; LSL; Mean; UCL; LSL d. UCL; USL; Mean; LSL; LCL
Suppose the test scores of 600 students are normally distributed with a mean of 76 and standard deviation of
6. What is the approx. number of students scoring between 70 and 82?
a. 272 c. 260
b. 408 d. 136
What is the defect?
A. A feeling that the process is causing a problem
B. Something we should ignore if possible
C. Something that customer is paying for.
D. A measurable characteristic not conforming to a customer requirement
344
Analyze Phase
345
Analyze Phase Roadmap
01 02 03 04 05
Review the Brainstorm and Apply Lean Validate the Validate the
process map RCA Principles assumptions with Parameters using
Hypothesis Correlation
346
Learning outcomes
Process
Outputs
(Ys)
Input
Variables
(Xs)
What Are The Vital Few Process Inputs And Variables (X’s)
That Affect CTQ Performance Or Output Measures (Y’s) ?
348
Root Cause Analysis
The diagram below shows an example value stream map. The example is taken from the investment bank and shows the
flow of a trade through IB operations where there is a requirement for manual intervention to settle.
C/T = 28 mins
Value Stream
Worker / Operator
Processing Box
Process Box Customer
• C/T: 12s
• C/O: 15 min
• 1 Worker
Withdrawal
Supplier
Problem Idea
358
VSM Step by Step Guide
Information Flow
Material Flow
Lead
Time Ladder
359
VSM Step by Step Guide – Step 1
Requirements of the
customer
360
VSM Step by Step Guide – Step 2
Requirements of
the customer
Requirements of
the customer
1
Step 2 Step 3 Step 4
Step 1
2
Dept. 1
• Activity 1
• Activity 2
3
• Activity 3
System
5
4
List the Inputs and the wait time between the processes.
Map the time taken for each of the process steps.
Customer
Requirements of
the customer
Dept. 1
• Activity 1
• Activity 2
• Activity 3
System
Requirements of
the customer
• Activity 1
• Activity 2
• Activity 3
System
Supplier Customer
7 pointed star in
red or blue
Requests
Broker 1 product every
Applications
The customer The customer
30 seconds
Delivery Times
1 batch
Delivery Times
Excessive with 12 items
transport 100% defects
per batch
1 2 3 3 12 90
1 2 3 3 2 3
365
Calculate metrics
Timeline
• A timeline shown below the process map helps to illustrate value added versus non value added time.
Total value
added time =
Value added
Lead time =
• Dividing the value added time by the total time gives an indication of waste levels (Efficiency of
process/Value added ratio – VAR).
• A similar process can be used to capture other information e.g. inventory levels and distance travelled.
366
Waste analysis
Waste (Non Value Added) Value Added Necessary Non Value Add
All other meaningless, Any activity that physically changes Any work carried out, which is
non essential activities i.e. eight the material or information being necessary under current conditions
waste categories. Aim is to worked on and increases its value. but does not increase value
Eliminate. Aim is to Maximise. e.g. Compliance. Aim is to
Minimise.
367
Terms you should know
Inventory
02 Build up of work between steps in the process that prevents seamless flow to the customer.
Lead time
03 The time taken for the development and delivery of a product or service to a client from the time
the request was made.
Process Map
06 A term used to describe a portrayal of the operation, systems, people and process with
supporting documentation accompanying each process step.
Pull
07 A flow mechanism where products or services are pulled through the value stream once
they are required to be worked on next.
Push
08 A flow mechanism where products or services are pushed through the value stream
once complete.
Queue Time
09 The waiting time or the idle time between steps, its also known as Lag time.
Bharat Paper Co. Ltd is into manufacturing and distribution of newsprint. Customer generates PO at a rate of 1 per 30
seconds and first note arrives at 0900AM. The delivery operator who receives the order adds time of collection, serial
number at 09:03AM (90 seconds) and passed on to the incoming analyst who receives at 09:08AM who adds the
locator code (70 seconds) from the code catalogue which takes up to 09:10AM.
Incoming analyst pass the same to delivery operator for verifying locator code (150 seconds), where it remains up to
09:15AM minutes. The filled up delivery note from delivery operator goes to incoming analyst in a different department
requiring 2 minutes to reach and there adds distance code based on locator code and at 09:18AM (60 seconds) it
passes to delivery operator where it remains till 09:20AM from where distance fee analyst takes possession of the
same on which he creates a distance fee request slip based on distance code (180 seconds) which takes up to
09:25AM and returns to delivery operator for verification to identify issues in locator code, where it gets holed up till
09:30 AM and then passes on to customer accounts manager for verification that takes till 09:38AM (60 seconds).
Customer accounts manager calculates the total in invoice log and adds distance fee to invoice before both papers
passed on to delivery operator (60 seconds). Delivery operator affixes validation stamp on invoice and without delay
transmits the invoice to outward analyst who validates the invoice (45 seconds) by 09:45AM and sends to customer
where customer logs the time of receipt at 09:50AM.
372
Lean Principles
373
Principles of Lean Thinking
1. Identify
Value
4. Pull 3. Flow
Establish Pull Create flow
374
Principle 1 – Identify Value
What is value?
Who defines value?
According to Womack and Jones – Value is expressed in terms of a specific product or service (or both) which meets
customers’ need at a specific price at a specific time.
The key question to be asked here is – What is the timeline? What is the price point? What are the other requirements
that must be met?
Providing value to the customers is the only reason for the existence of our business.
375
Principle 2 – Map the value stream
According to Womack and Jones – Mapping the value stream is a step in taking a specific product or service (from
scratch) to its final recipient i.e. the end customer.
Start Finish
376
Principle 3 – Create Flow
What is a Flow?
It is a movement of the product or service from the supplier to the end customer
Womack and Jones advises – “make the value creating steps occur in tight sequence, such that the product or
service flow smoothly towards the end customer.”
• The rationale behind flow is that the product or service flow smoothly in the value chain without interruptions
• Waiting time and hand-offs are eliminated from the process flow
Pull is the movement of product or service from the supplier to the customer
only when the customer needs it.
• Making or processing the product only when there is a customer demand
• Create a “Just in Time” manufacturing or delivery
• Reduced storage, stockpile, inventory
Customer Pull’s from the
value stream
Customer places an
Order Processing Customer receipt of order
order online
378
Principle 5 – Seek Perfection
Alex Osborn
The word brainstorming was originally introduced by Alex F. Osborn in 1953 through his book Applied
Imagination: Principles and Procedures of Creative Thinking.
Osborn suggested that the participants in the group should have varying amounts of experience in the task
that was at hand, however, he discouraged mixing participants from various levels in the company’s
hierarchy within a brainstorming group. Other specifics that Osborn suggested were that participants had to
be adequately informed in advance about the problem that would be tackled during the brainstorming session
so that the creative focus would be on the specific problem.
381
Cause and Effect diagram
A visual tool used by an improvement team to brainstorm and logically organize possible causes for a
specific problem or effect.
Potential High Y
(Effect)
Level Causes (Xs)
Brainstorm the “major” cause categories and connect to the centerline of the Cause & Effect diagram.
Measurement Methods
Why is there a
delay in
processing?
People Machinery
383
Example of Fish Bone Diagram
These are causes that the team These are causes that are actually These causes are the causes that
unanimously conclude are beyond solutions that can be implemented have passed the above two filters
the control of the present process directly and need no further and need further analysis.
boundaries or outside the physical analysis. They are usually stated
location of the process execution. as lack of resources, equipment,
tools or training.
385
Taking to next level
Brainstorming + Affinity
• The immediate next step after the segregation is to attack the likely and controllable causes and ask at least 3 – 5
why’s?….. for each cause. This is called root cause drill down.
• Only after we have asked why 3 - 5 times to each of the likely causes, we will be able to arrive at the possible root
cause, also known as KPIV.
387
A Monumental Mystery
1. Why?:
Because of usage of Harsh Chemicals to clean the monument.
2. Why?
Harsh chemicals were used to clean bird droppings
3. Why?
Birds Feasted on spiders on the monument
Problem: 4. Why?
The Jefferson Monument in Spiders Feasted on Midgets/ Gnats on monuments
Washington DC was falling
apart/ deteriorating. 5. Why?
Midgets/ Gnats were attracted to the lights at dusk.
Root Cause: Timing of
the dusk lights.
388
Pareto Analysis
Minitab path:
Refer the data: “Clothing Defect”.
• Click on Stat → Quality Tools → Pareto
Chart.
• In Defects or attribute data in,
enter Defect.
• In Frequencies in, enter Count.
• Select Combine remaining defects into
one category after this percent, and
enter 95.
• Click OK.
390
Pareto Analysis
In the above graph, 45.2% of the defects are missing buttons and 23.3% are stitching errors. The cumulative
percentage for missing buttons and stitching errors is 68.5%. Thus, the largest improvement to the entire clothing
process might be achieved by solving the missing button and stitching problems.
391
Tackling the other causes
• Null is the hypothesis which is tested. • There is a significant difference between the two or
• There is NO significant difference between the two more groups that were selected for the test.
or more groups that were selected for the study • The term “Alternative hypothesis" refers to a claim
• Null Hypothesis is considered as the Status Quo. that completely contradicts a null hypothesis by
Meaning, there is no change between sample asserting that the true value of a population
statistics and population parameter. parameter is greater, less, or not equal to the value
given in the null hypothesis.
Example:
• Alternate hypothesis is the purpose of the study.
• The call resolution time for Team A is same as What we are trying to prove.
Team B for the same process.
Example:
• Which can be represented as H0: µ1 = µ2
• The call resolution time for Team A is not equal or
more than or less than Team B for the same
process.
• Which can be represented as H1: µ1 ≠ µ2 or µ1 >
µ2 or µ1 < µ2
397
Significance Level (“α”)
Now as we know that there are some %age of risks (Ex: 10%, 5%, 1%) involved while taking the decision so let
us know about what are those risks.
Example:
Let's consider a hypothetical example, I am having fever and I am assuming that I am malaria positive.
So here,
Ho = I am not Malaria positive, and
Ha = I am Malaria positive
Example:
Let's consider another example,
• An automotive company had set a limit of 2% defective for the auto components it receives from one of its vendor.
• A sample of 1000 parts taken for inspection out of 5000 lot. Out of which, 30 parts found defective means 3%
defective which is greater than the acceptable limit of 2%.
• The shipment was rejected by the customer.
• A 100% inspection was done by vendor, and they found that only 55 parts were defective in that lot. Which means
only 1.1% defective.
• In this scenario, 1.1 percent of the parts were defective as compared to an acceptable level of 2%, hence rejecting
the shipment was an error.
• If we convert this scenario into hypothesis language, we can say that the null hypothesis is rejected but, the lot was
well within acceptable limit.
• In this case we have committed Type I error or “α” error. This is also known as producer’s risk because the
manufacturer or the producer is affected
403
Type II Error (“β” Error)
Example:
In the same example,
• A sample of 1000 parts taken for inspection out of 5000 lot. Out of which, only 11 parts found defective means 1.1%
defective which is less than the acceptable limit of 2%.
• The shipment was accepted.
• But while using the components it was found that there were 130 defective parts in the lot. Which means 2.6%
defective.
• In this scenario, 2.6% of the parts were defective as compared to an acceptable level of 2%, hence accepting the
shipment was an error.
• If we convert this scenario into hypothesis language, we can say that we failed to reject the null hypothesis while,
the lot was not within acceptable limit.
• In this case we have committed Type II error or “β” error. This is also known as consumer’s risk. Because here the
consumer is affected
404
How to select hypothesis!!!
With 1 set of data to a specific With 1 set of data with With More than 2 or
Comparing
standard (1:Standard) another set of data (1:1) multiple sets of data
For Variance
• 1 variance test • 2 variance • Equal variance
(Continuous Data)
For Median
(Non-Parametric
• 1 sample sign test • Mann Whitney • Mood’s Median
test)
(Continuous data)
405
Path for Hypothesis tests in Minitab (Reference)
1-Sample Z
1-Sample t
2-Sample t
Paired t
1 Proportion
2 Proportions
408
Path for Hypothesis tests in Minitab
One-Way ANOVA
409
Path for Hypothesis tests in Minitab
Test Result
Inference
• Sample Mean = 32.152
• The 95% upper bound for mean is 32.471 means, in the
current context, the maximum mean value for the population
would be 32.471.
• As the hypothesized mean is more than the 95% upper
bound, we can say that the test is significant, and we can
reject the null hypothesis.
• P-Value is 0.000 < 0.05. Reject Null Hypothesis.
• The claim is valid
Note:
• Easy Remember Technique:
• “P” Low Null should go. (Reject Null)
413
1 Sample T test
Test Result
Inference
• Sample Mean = 5.0833
• The 95% lower bound for mean is 4.9212 means, with the
current sample, the minimum mean value for the population
would be 4.9212.
• As the hypothesized mean is less than the 95% upper
bound, we can say that the test is significant, and we can
reject the null hypothesis.
• P-Value is 0.000 < 0.05. Reject Null Hypothesis.
• The claim of your logistic team is valid
416
2 Sample T test
Test Result
Inference
• The 95% upper & lower bound for the difference in
mean for Sample 1 and sample 2 is (-0.224 to 0.949)
• As the hypothesized difference is falling in the 95%
upper & lower bound difference range, we can say that
the test is not significant, and we can not reject the null
hypothesis.
• P-Value is 0.220 > 0.05. Fail to reject Null Hypothesis.
• We can say that there is no difference between Team A
and Team B
419
Paired T test
Test Result
Inference
• The 95% upper bound for the difference in mean for
Sample 1 and sample 2 is -0.351.
• As the hypothesized difference is more than the 95%
upper bound difference, we can say that the test is
significant, and we can reject the null hypothesis.
• P-Value is 0.001 < 0.05. Reject Null Hypothesis.
• We can say that the mileage of cars without additive is
less than that of with additive.
422
One-Way ANOVA
Test Result
Inference
• P-Value is 0.263 > 0.05. Fail to reject null.
• In the Interval plot, overlap of means can be observed in the 95% CI.
• We can conclude that there is no significant difference in the means.
425
1 Proportion Test
• Step 1: Select the path (Stat > Basic Statistics > 1 Proportion)
• Step 2: Select the summarized data from drop down
• Step 3: Enter the data in the given area
• Step 4: Select the perform hypothesis test and enter
hypothesized proportion.
• Step 5: Go to options & select Confidence level & Alternative
hypothesis
427
1 Proportion Test
Test Result
Inference
• Sample proportion = 0.53
• The 95% lower bound for p is 0.510217 means, with the
current sample, the minimum proportion for the
population would be 0.51 or 51%.
• As the 95% lower bound proportion is less than
hypothesized proportion, we can say that the test is not
significant, we failed to reject the null hypothesis.
• P-Value is 0.944 > 0.05. Fail to reject Null Hypothesis.
• Proportion of people supporting the candidate is equals
to 55% or less
428
2 Proportion Test
• Step 1: Select the path (Stat > Basic Statistics > 2 Proportion)
• Step 2: Select the summarized data from drop down
• Step 3: Enter the data in the given area (Sample 1 & Sample 2)
• Step 4: Select the perform hypothesis test and enter
hypothesized proportion.
• Step 5: Go to options & select Confidence level & Alternative
hypothesis
430
2 Proportion Test
Test Result
Inference
• Normal approximation P-Value is 0.410 > 0.05. Fail to
reject Null Hypothesis.
• There is no difference in the defect percentage.
431
Chi-Square Test
Test Result
Inference
• Pearson P-Value is 0.000 < 0.05. Reject Null Hypothesis.
• There is association between region and claim denial exists.
434
1 sample sign test
Since the P-value is greater than level of significance (0.05), we fail to reject
the null hypothesis.
Therefore we fail to conclude that the price index for homes is greater than
115
437
Mann Whitney Test
Since the P-value is greater than level of significance (0.05), we fail to reject
thenull hypothesis.
Therefore we conclude that there is no difference between population medians
440
Mood’s Median Test
Since the P-value is less than level of significance (0.05), we reject the null hypothesis.
Here, the participant scores are classified as being below or above the overall median. The p-value < than 0.05
indicates that the medians are not equal.
443
Probability Value or “P” Value
Definition:
• There is a probability is that the samples were taken from the same population that was being analyzed.
Explanation:
• If a sample is taken from a population, the population's characteristics and the sample's statistics won't differ much,
and a high probability value (P-value) would suggest the same.
• A high P-value, like 0.55, indicates that there is 55% chance that the sample represents the population.
• When the P-value is low, for an example 0.01 means there is only a 1% likelihood that the sample represents the
population.
Sample 1 Sample 2
High probability Low probability sample is
sample is drawn drawn from the
Population from the population population
444
Probability Value or “P” Value
Definition:
• The probability that the value which is being tested will fall within the specified confidence interval at a specified level
of confidence.
Explanation:
• For reference, have a look at the Average Handling Time summary in the graph below.
P value
Phase Purpose Intended outcome Preferred type of test
expected
• 2 sample t
To validate Root Causes stated as • ANOVA
To prove that the observed difference between the
theories and to take decision on • 2 Proportion
Analyze original data and the trial data is real and change will P < 0.05
whether or not to implement the • Chi SQ
significantly benefit the outcome of the process.
change • Mann Whitney
• Moods Median
To compare data on the "Y" before and after
• 2 sample t
To validate that improvement claimed improvement and prove that after improvement data
• 2 Proportion P < 0.05
is real on "Y" is statistically and significantly different from
• Mann Whitney
Measure Phase data on "Y"
Improve To compare the post improvement data on the "Y" with
the Set Target and determine whether it has been
To validate whether the improvement • 1 sample t for Normal
achieved. Even if the 2-sample test validates that the P > 0.05
target has been achieved or not • 1 sample sign for Skewed
process has improved, we need to know whether we
have actually achieved the target.
Any situation when it is required to To compare mean of current performace data against
Other than • 1 sample t for Normal
understand whether current a Target Performance and determine whetehre we are P > 0.05
Six Sigma • 1 sample sign for skewed
performance is meeting set target meeting the target or not
446
Correlation &
Regression
Scatterplot
447
Correlation Analysis
Correlation
449
Correlation Analysis
• Now the engineer, being a logical man, refused to believe that this man’s car was allergic to vanilla ice cream. He
arranged, therefore, to continue his visits for as long as it took to solve the problem. And toward this end he began
to take notes: He jotted down all sorts of data: time of day, type of gas uses, time to drive back and forth etc.
• In a short time, he had a clue: the man took less time to buy vanilla than any other flavor. Why? The answer was in
the layout of the store. Vanilla, being the most popular flavor, was in a separate case at the front of the store for
quick pickup. All the other flavors were kept in the back of the store at a different counter where it took considerably
longer to check out the flavor.
• Now, the question for the Engineer was why the car wouldn’t start when it took less time. Eureka – Time was now
the problem – not the vanilla ice cream!!!! The engineer quickly came up with the answer: “vapor lock”. It was
happening every night; but the extra time taken to get the other flavors allowed the engine to cool down sufficiently
to start. When the man got vanilla, the engine was still too hot for the vapor lock to dissipate!!!
453
Correlation Analysis
Equation:
• Regression Equation is Y = MX+C or Y = β0+ β1X
Were,
• Y = Dependent Variable
• X = Independent Variable
• β0 / C = Y-interceptor or constant (Where X = 0)
• β1 / M = Slope
456
Path for Regression Analysis in Minitab
Test Result
Inference
• Coefficient of determination R-Sq = 95.71% and p value = 0.000
• R-Sq = 95.71% Means, 95.71% of variability in revenue depends on
the amount spent on advertisements.
459
Correlation &
Regression
Multiple Linear Regression
460
Multiple Regression
• Multiple regression is to compare relationship • When process input variables and the output is
between a continuous dependent variable (Y) and continuous, multiple regression can be used to
several continuous independent variables (X1,X2…) investigate the relationship between the inputs and
the outputs
461
Multiple Regression
The researchers were interested in determining if a relationship exists between blood pressure and age, weight, body
surface area, duration, pulse rate and/or stress level.
462
Multiple Regression
Pulse -
0.0516 -1.64 0.126 4.41
0.0845
The R-sq value indicates that the predictors explain
99.62% of the variation in BP.
Stress 0.00557 0.00341 1.63 0.126 1.83
Thus proving that the model fits the data well.
465
Multiple Regression
BP = -12.87 + 0.7033 Age + 0.9699 Weight + 3.78 BSA + 0.0684 Dur - 0.0845 Pulse
+ 0.00557 Stress
R Large residual
466
Multi-Collinearity
• Multi-Collinearity is a condition
wherein high correlation exists
between two independent
variables
1
VIF =
1 −𝑅𝑅2
Where;
𝑅𝑅2 = Coefficient of determination
• High values of pairwise correlation (generally > 0.8) provide warnings of potential
multicollinearity problems
• We see that Weight and Body Surface Area are most highly related to Blood Pressure.
470
Checking Multi-Collinearity- Relook at the output
Multi-collinearity exist between Weight and Body Surface Area (BSA). The correlation coefficient between
the two variables can be estimated through correlation
474
Checking Multi-Collinearity with correlation
We see that Weight and Body Surface Area are most highly related to Blood Pressure.
475
Correlation &
Regression
Logistics Regression
476
Logistics Regression
• Logistic Regression also creates a prediction equation, but for a discrete dependent variable (Y)
• Logistic Regression can be applied to output data that are Binary (yes or no), or Ordinal (high, medium,
low), or Nominal (North, South, East, West)
• As part of this course, we only focus on Binary Logistic Regression
• While the result of logistic regression is similar to least squares regression (a prediction equation), the math
to get there is much different as it is based on natural logarithm and odds ratio
Usually a transactional projects often utilize non-continuous data like:
• Was the service provided in an appropriate manner?
• Was the order filled?
• Was the product available for shipping?
• Degree of agreement scales, satisfaction ratings
• Defect severity ratings (A is better than B is better than C)
Response Information
Variable Value Count
RestingPulse Low 70 (Event)
High 22
Total 92
Deviance Table
Source Adj Dev Adj Mean Chi-Square P-Value
Regression D2 7.574 3.787 7.57 0.023 Since the P-value for Weight and Smokes is lesser
F Weight 1 4.629 4.629 4.63 0.031 than the level of significance, there is enough
Smokes 1 4.737 4.737 4.74 0.030 evidence to indicate that both has a significant
impact on Resting Pulse
Error 89 93.640 1.052
Total 91 101.214
481
Logistics Regression
Frequenc
20
Percen
50
t
10 10
y
1
0. 0
1 -2 0 2 4 -2.0 -1.5 -1.0 -0.5 0.0 0.5 1.0
1.5
Residua
l Residual
Versus
Order
1
0
Residu
-1
al
-2
1 10 20 30 40 50 60 70 80 90
Observation Order
The Normality Plot does not follow a line & Histogram shows a bimodal pattern which means the data is
not enough for normal approximation theory to apply.
484
Multivariate
Analysis
485
Multivariate Analysis
• MANOVA is used to determine whether the means between two or more groups differ when you have multiple
continuous response variables i.e., multiple Y’s dependent variable, Multiple X’s that are categorical, and optional
covariates
• The response variable should be correlated. It is multivariate version of the ANOVA version.
• For example, an analyst at a manufacturer of shoes products wants to study the affects of different leathers on the
strength and flexibility of the products. Since strength and flexibility are correlated response variables, the analyst
performs a general MANOVA to analyze the data.
488
MANOVA- Multivariate Analysis of Variance
Exercise
A manager of a car auto parts supplier wants to assesses the flexibility and the quality ( Both response “Y”) of the
clutch system that they provide. The clutch system are manufactured using two different methods at three plants.
The manager wants to determine whether the method and the plant affect the final product. To assess how method
and plant affect both response variables at the same time, the manager collects data and does a general MANOVA.
A significance level of 0.10 was used to perform the analysis.
Exercise
By looking at the height data it becomes easier of us to distinguish the persons. But suppose they almost have the
same height, then perhaps it becomes difficult for us to differentiate because the variance is lower.
Person Height
Alex 145
Ben 160
Chris 185
In the second scenario we have another data set for weight. But since the variance in weight is very small, we still
have to rely on height . But suppose if the height and weight have got the same variance, then it would become difficult
for us to choose the variable to delete.
Person Height weight
Alex 145 68
Ben 160 67
Chris 185 69
PCA combines both height and weight to create two brand new variables. It could be 30% height and 70% weight, or
87.2% height and 13.8% weight, or any other combinations depending on the data that we have.
These two new variables are called the first principal component (PC1) and the second principal component (PC2).
492
Principle Component Analysis- Exercise
A bank requires information such as - income, education, age, current residence, tenure with current employer,
savings, debt, and number of credit cards - of the applicants for processing the loan. The Bank manager wants to
reduce the number of variables, so he collects 30 loan applicants data to analyze. The manager wants enough
components to explain 90% of the variation in the data.
The Factor Analysis is used to Factor analysis summarizes data We can model each original
assess the structure of data by into a few dimensions by variable as a linear function of
evaluating the correlations condensing large number of these underlying factors
between variables. variables into a smaller set of
latent factors that we do not
directly measure or observe, but
which may be easier to interpret
496
Factor Analysis- Exercise
The Human Resource department measures the job applicant on various variables such as - academic records,
appreance, communication skills, company fit, experience, job fit, letter, likeability, organization, potential , resume
and self confidence . They rate each job applicant on these 12 variables using a 1 (low) through 10 (high) scale. The
HR manager wants to identify the underlying factors that explain these variables and for that he collects the data for
50 job applicants. Previous analysis determined that 4 factors account for most of the total variability in the data.
Forecasting Techniques are used in business to estimate the future value based on current
and past data. For example,
• What will be the demand of a particular item A in the next three months based on last twelve months data.
• How many products to order in next three month based on last twelve-month data.
There are three categories
• Qualitative technique - Based on Human judgement where we don't have historical data to forecast demand for a
newly launched product. Use of expert opinion for a short period of time. E.g. Delphi Method
• Causal Technique - Regression Model.
• Time series Model - Use statistical technique when sufficiently large amount of past data are available.
The quality of a forecasting depends on how well the data fits the model. These models assume that future demand
will follow the historical pattern that contain both
• This technique is used to provide short term forecast and the data shouldn't have a trend or seasonal component.
• Check for any trends before using moving average method
Exercise:
An analyst wants to predict the sales of an electronic items by using moving average method. He collects sales data
and the number of commercials that were broadcast during the previous 60 weeks.
504
Moving Average-Exercise
Interpretation of Results
From the plot it can be observed that
the data fits the model