Lean Six Sigma

Module 1- Introduction-

50Q in 50min

Business why? To provide value

5 stakeholders- customers, suppliers/partners, govts shareholders,

society, employees/people

Quality- How well the services are performing with respect to the
expectations of the customer

Quality= Performance/Expectation

P/E<1 customer attrition

P/E=1 customer satisfaction

P/E>1 customer delight

Delight -> customer loyalty -> repeat business

LSS helps identify factors that cause defects or delays in the process

Y=f(x) – focus on x

Inspection- It is not what drives the success of quality, don’t rely on it;/,
some may go unnoticed

Remember the number of ‘f’

Bill Smith formed Six Sigma (Motorola – Japan)

Origin story

Lean focuses mainly on the reduction of wastage in our process

1. Value added – the customer is willing to pay for this

2. Value enablers – we cannot operate without these
3. Non-value added – remaining activities
 8 types of wastages- TIMWOODS (Transportation, Inventory, Motion,
Waiting time, Overproduction, Overprocessing, Defects, Skills)

Kaizen (kai – change, zen - for good), kanban (nothing is produced

unless we get the green light from up), poka yoke, 5 S

Six Sigma is a fact-based, data-driven approach that focuses mainly on

the reduction of variation in our process

Accuracy – close towards the target (99%)

vs precision – close towards each other( 15.232422%)

3 m’s of 6 sigma

1. Metric—it is a measurement system. When we say that the process

accuracy is 99.99967 % or 3.4 demo, we say that the process is
operating at the Lean Six Sigma level. Depending on the criticality,
whether it is defective or not. We need to measure!

Defects mean non-conformities to a set of parameters

Defective means having 1 or more defects

Define Measure(y) Analyse(x1,x2,x3) Improve Control

Customer Define the baseline/ Validate root Documentation
Collect the data causes
Voice of DCP(data collection Brainstorming Control charts
Customer plan)
Segregation Data can be Segregate
done with CC – Continuous according to the
tool- Data fishbone diagram
AFFINITY ATT – Discrete
diagram data
Prioritization Measurement of Causes can be
is done with data(central non-controllable,
the KANO tendency and direct
model dispersion) improvement,
controllable
Find the CTQ Graphical Controllable causes
– critical to representation can be event-
quality using Histogram, based, theory-
boxplot based, parameter-
based
Develop Calculate a. Event based-
Project Cp,Cpk(Continuous) pareto chart,
charter – DPU,DPMO(Discrete b. theory based
outcome data) – hypothesis
testing
c. Parameter
based –
corelation
and
regression

Process Define baseline , Validate the causes

Mapping - MSA
outcome
2. Methodology – It works on the framework known as DMAIC
(Define, Measure, Analyze, improve, control)
 3. Management philosophy – commitment by people/managers

Sigma levels and parts per million ( like in a million there are 3.4 defects
is 6 sigma) as defects increases sigma level decreases

Walter Shewhart-

PDCA( plan do check act)

Edward Deming – PDSA( study)

Joseph juran – Pareto chart

Kaoru Ishikawa – fish bone

Phillip Crosby- 4 absolutes of quality mgt

6 sigma team-

- Apex Council

- Champion/Sponsor

- Process Owner

- Master Black Belt

- Black Belt

- Green Belt

- Team Members

DMADV(Design Verify also known as DFSS- Design for Six Sigma – build
a new system) vs DMAIC (improve something)
Module 2 – Define Phase

Road Map-

- Identify the customer

- Capturing the VOC(Voice of Customer)
- Categorize the VOC
- Prioritize the VOC
- CTQ (Critical to quality)
- Develop the Project Charter
- Map the process

Someone who consumes the outcome of my process is my customer

Types of customer-

1. Internal customer
2. External customer

Capture the VOC (feedback for external – meetings for internal)

Segragation of VOC – AFFINITY

1,2,14,16 3,21,31 4,5,11,13,17 ……

1
2
3
…
80

Prioritization – KANO Model

Now we need to divide the segregated from the affinity diagram and
prioritize them .
Delighter (not focused), more the better (focused), must be (should be
empty)

CTQ- critical to quality

A specific measurable attribute of the output of the process
(similar to KPIs or CSF)

Example: VOC -> CTQ

Deliver me product faster -> Delivery time between
Deliver me pizza hot -> Temperature of Pizza
Loan app form submitted by officer has too many errors -> Number of
defects

Project Charter- MOU ( memorandum of understanding ) with higher

management.
7 elements of Project Charter

- Project Title
- Business Cases (Elaboration of problem statement and should end
with sense of urgency)
- Problem statement (maximum 2 to 3 lines and it should
specifically define what the problem is, where the problem is, since
when it is occurring and what the impact is)
- Goal statement (SMART – Specific, Measurable, Achievable,
Relevant, Timebound)
- Project scope – (boundary of project)
- Project timelines / Milestones (6-7 month is usual but can differ,
eg. D – 1mth, M- 1mth, A- 2.5mth, I-1mth, C-1mth)
- Project team members (Champion, Business Head, …)

Project charter Note-

(Don’t put any causes or any solutions in a project charter)
1. When we know the problem, but we don’t know the causes and
the solutions we to do the project
2. When the problem is chronic(happens frequently) we need project
and not for sporadic(happens rarely)
3. For one CTQ we have 1 project charter
4. LSS is done for complex cross-sectional process issue

Process Mapping-

The process takes some inputs and provides some output….

1. Always map the “as is” process and “not to” be process
2. Stay on the process, be on the process and map it the way it is –
Actions are where the process is
3. Don’t take any reference from any existing process map
Macro level – SIPOC
Suppliers Inputs Process Output Customers

Eg- car servicing

Supplier Input Process Output Customer
Customer CAr Car coming Job order Station team
to station
Station Job order Issue Update the Servicing team
team identification job order
Servicing Update the Car servicing Update the Washing team
team job order, job order
grease
Washing Update the Car Washing Update the …
team job order, job order
washing
… …
… Owner/customer

1. This process helps us to identify who is involved in the process.

2. It helps us identify the scope of the process
3.

Micro level – Flowchart, Swimlane(Cross-funtional flowchart) and VSM

arrow
Document
Activity box Decision
Terminator box

Swimlane has the departments mentioned along with flowchart

Module 3- Measure Phase
The objective of the measure phase is to define the baseline (y).
(what is my current sigma level?)
Roadmap –
1. Collecting the data
2. Measuring the data
3. Graphically presentation the data – Data Distribution
4. MSA- Validating the measurement system
5.

Data-

Continuous / Variable Attribute / Discrete

data data
• Anything that can be • Counting
measured on a scale • How many
• How much • They dont have units
• They got SI units • Binary
• Nominal
• Ordinal

Confusing Eg:-
Percent cream content in milk bottle (continuous/continuous i.e
continuous…. As ml/ml)
Percent defective parts in hourly production (discrete by discrete i.e
discrete…. As 3/10 i.e 0.3)

Eg : flights are getting delayed

Participant 1-
Maintenance
Security
Technical issue
Number of runways

Participant 2-
Refueling time
Weather
Air traffic
Crew size

To avoid ambiguity, we need a plan which is known as a Data collection

plan

DCP – (not restricted but should have these)

What is to be collected- Operation definition
Where to collect -source
When to collect- Frequency
How to collect- sampling technique: different types of sampling technique
Who is responsible-

Sampling :–
1. Random (anything can be part of sample, equal probability of
taking up)
2. Stratified sampling – grouping (we make groups and group se we
select some samples from each group)
3. Systematic sampling (when the interval is constant of each group
like every 10th house, every 10th minute, or something like that)
Descriptive
Inferential
Sample size(min 30 but higher is usually better):-
Level of confidence n= [Zc σ/D]^2
Precision
Standard deviation

Data Measurement
Central tendency-
Mean, Median, Mode
Dispersion-
Range, Variance, Standard deviation
Why standard deviation is important?
1. Unlike Range standard deviation, it takes into account all the data
point
2. Unlike the variance, standard deviation has the same unit as that
of the original data

Types of Data Distribution:

Histogram helps us to identify the types of data distribution

Properties of normal distribution
1. It is a bell-shaped curve, It has a mirror image at the center and
doesn’t touch the X-axis
2. Mean, median, and mode are the same
3. The total area under the curve is 100%
4. 68% will fall under +- σ, 95% under +-2 σ, 99.73 under +-3 σ

To check whether the graph is normal or not we use proof of

normality:
i.e Anderson-Darling test (p-Value >0.05 then the data falls a normal
distribution)

Boxplot(We use boxplot usually when the data is not normal and we use
a control chart for normal) –
It helps us to identify the outliers
It takes into consideration the median performance
It helps us to compare different sets of data
Minitab:-

C2-T (text)
D (Date)
Heading wale maid ala toh it doesn’t change

Step 1:-
Copy paste from [Link]
Click on Stats-> Basic Stats->Display Descriptive Statistics

Double click C1 Diameter

Click on none then select the relevant:-
2nd Graph for histogram-
Graph -> Histogram
Ok for simple histogram

Double click Diameter then ok

It seems to follow normal use boxplot to check
Now perform Anderson darling test
Stats-> Basic stats -> Normality
As the p-value is greater than 0.05 so it follows a normal distribution

Interpretation is important!!

Validate Measurement System:-

Process variation or human measurement variation

There are 2 types of variation here:-

Repeatability- The same part, the same gage, the same operator,
multiple times the variation that is occurring is known as repeatability
(Siddhesh says 10,9.3,10.7 on different days - training)

Reproducibility- The same part, same gage, multiple operators, the

variations that is occurring is known as reproducibility
(Ravi says 9.7, Siddhesh says 10.2, Nikhil says 10.5 - SOP)

Measurement system analysis-

Done on Human Interventions

Discrete data,
Continuous data.
Attribute Agreement
Gauge R&R
Analysis

Min 30
Min 10
Operator -2
Operator -2
Replicate 2
Replicate 2
Master Calibration

% contribution
Kappa Value
% study variation
Stats-> Quality tools -> Gage study -> Gage R&R Crossed
Fabric AAA-
Process Stability/Control –
When all the outputs of my process are within +-3 std deviation, then
we say that my process is a stable process.

If any data point is outside the control limits/natural boundaries then we

say that the process is not stable

These data points that are outside the control limits are spatial cause
variation (outliers)
And within the limits are common causes of variation

Using right(CCV) and left hand(SCV)- there is variation….

Quality
Quality
Quality
Quality
Quality

Control charts are used to check whether our process is stable or not….

VOP- Voice of Process (distance between the control limits)

Process capability- when all the output of my stable process is within the
specification limits set by the customer I say that my process is capable.

USL LSL (Upper specification limit), the difference is CTQ/VOC

1) Stability – Control Chart

2) Normal - Anderson Darling Test
Capability index Cp = VOC/VOP= (USL-LSL)/(UCL-LCL)=(USL-LSL)/6sd
CP > 1…. Capable
Cp = 1…. Just Capable
Cp < 1…. Not Capable

Eg-
Sample 100
Mean 6.4
SD 0.25
Customer 6.4 +-0.6

VOC=0.6+0.6=1.2
VOP=0.25*3+0.25*3=1.5
… so Cp=1.2/1.5=0.8 so not capable

So,
Mean=10
Sd=1
Customer 12+-4

VOC=4+4=8
VOP=3+3=6

Cp=8/6=1.33

But the mean is 10, and customer ka mean is 12, so ye capable nhi hai,
so this is caused by shifting of mean, as capability doesn’t take care of
this don’t use Cp

Use Cpk(i.e cut karke half karke analyse karo)

Cpk= (USL-mean)/(3*s.d) --- z-upper
(mean-LSL)/(3*s.d) --- z-lower
Dono ka minimum is Cpk value(i.e., the mean is shifted in that direction),
should not be negative….

Cpk * 3 = current sigma level

Q.
Mean =178.6
Sd=3.6
Target=171
USL=182
LSL=160

Cpk = 182-178.6/3*3.6=0.314
178.6-160/3*3.6=1.722

So min is 0.314
0.314*3=0.942 so it is 1 sigma level

What is sigma?
If I can incorporate 1 sigma lvl on 1 side and 1 on other side than 68%
falls under the customer specification limit
99.99967%

Now

 We first check normality

 Then whether within control and
 finally we check the capability
Now for discrete consider the example of pen

Number of
opportunities(5)
Length
Diameter
Label
Cover
Grip

1. Defect:-
DPO(Defects per Opportunity) = Total defects /(number of opportunity *
Total number of samples inspected) = 0.0048

DPMO(Defects per million Opportunity) = DPO * 10^6 =4800

Current sigma level(Use Sigma & DPMO table) = 4.09

DPU(Defects per unit) = Total defects/number of samples inspected =

0.024

Proportion = 24000

Current sigma = 3.48 (Use any one of the methods)

2. Defective:-
Proportion = Total defective / number of samples inspected

In eg, we don’t have exact but we can calculate the min and max range
i.e 3 is min(5,5,2) and max 12(1,1,1,1,….,1 12 times)

Note-
- If something is repairable go with the concept of defect
- If it isn’t repairable go with the concept of Defective
- Within defect if we can calculate/count the number of
opportunities go with DPO/DPMO or else go with DPU/Proportion

Module 4- Analyze Phase

The objective of Analyse phase is validated root causes
There are 2 doors-
Process Door -> Lean Principle
Data Door

Roadmap-
- Perform critical analysis of a process through lean thinking
- Identify the different types of wastages.(TIMHWOOD) and Describe
Value stream mapping.
- Identify and prioritize critical X's
- Use data to validate critical X’s (causes based on events/categories)
- Perform the assumption-based causes analysis and validate it
statistically by applying the tools and
- technique taught during the session.
- Use correlation between variables X’s and established relation
between them and develop model for
- future prediction
- Understanding of basic concepts of AI & ML

Data Door
1. Non controllable
2. Direct Improvements
3. Controllable
a. Event – Pareto chart
b. Theory – Hypothesis testing
c. Parameter – Correlation

Process Analysis-

1. Examine and challenge the Decision box

a. Include only when something is processed and the original state
cannot be retrieved
 b. Before sending it to the client we need a decision box
2. Checking mechanism
a. Check whether there is SLA and proper agreement between the
departments before moving the data/things…
3. Repair/Rework
a. Replace is better than repair if the repair cost is 70% of the
replacement then do replace
4. Document (Blackbelt)
a. There should be an independent audit system to update and
upgrade the central database
5. Try to convert any sequential activities into parallel activities

Lean – Developed in Toyota

5 Principals of lean

Principles
1. Identifying the value (TIMWOODS)

- Transportation
- Inventory (Raw material , finished goods, in progress(most dangerous
as no one wants it, it is a waste))
- Motion
- Waiting time
- Overproduction
- Overprocessing
- Defects
- Skills
2. Mapping the Value Stream (VSM)

Process cycle efficiency(PCE)= VA/Leadtime where LT= VA+NVA+VE

So, we need to reduce the NVA and Optimize the VE

Takt Time(depending on customer demands) is the minimum time taken

to get a unit of product to meet customer demand and lead time should
be less than takt time
3. Create Flow
4. Establish Pull
Kanban is triggering the flow/pull…
Pull examples of luxury cars/airplanes/ship
5. Seek perfection (Follow it up…. – Kaizen, Kanban, Poka yoka, SS)

Data-driven---
1. Brainstorming (send agenda 1-2 days before, maximum 8-10
people, only 1 with a token will speak, 3-4 mins speak, stick it
to the board, maximum 1 hr and boards should be filled) this is
idea generation, iske baad others can speak like idea harvesting
and we can speak
2. Fishbone Diagram
Cause and Effect diagram/fishbone/Ishikawa diagram
- 6 M’s (Man, Machine, Materials, Methods, Measurement, Mother
Nature)
After identifying and segregation the causes, we segregate as per the
causes
1. Non controllable
2. Lack of Solutions or Direct Improvements
3. Likely and Controllable Causes (3 Event-Pareto, Theories-Hypo,
Parameter-Regression)

5 Why Analysis

Pareto chart-
Control impact matrix is used to prioritize the causes

Hypothesis Testing-
The hypothesis testing is done in order to validate whether the observed
or the obvious difference is actual or is due to chance.
(is there an issue or is it due to chance)
Steps-
- Draft the null and alternate hypothesis
- Set the level of significance
- Collect data
- Select the hypo test
- Inference

a. Set Null and alternate hypo

Null- no correlation, equality, or no change
Alternate – it is the challenging statement, it means there is change and
equality will never be there

Eg-
1. Null H0- The average TAT for both the shift is the same
TAT1=TAT2
Alternate Ha -TAT1 != TAT2

2. H0: Dia = 5mm

Ha: Dia !=5mm

3. H0: Time <=10

Ha: Time >10

4. The project leader believe that bolt with thread a has a stronger
torque on average than bolt with thread B and wants to perform
the hypo
H0: Toq a= Toq B
Ha: Toq A> Toq B

b. Set the level of significance

 Level of confidence- 95% (if nothing is mentioned)
So level of significance = 1-0.95) = 0.05

(if level of confidence is 100% then what is the use of hypothesis testing??
When sample =population then it can be 100%)

c. Collect data
d. Select type of hypo test

Types of hypothesis testing

1: standard 1:1 More than 2

Continuous data 1 sample t test( sample 2 sample t test( One way ANOVA
Test of mean size<30) independent – TAT of
1 sample z test( machine A and machine
ss>30) B)
Paired t-
test(dependent- Weight
of participant before and
after training, subject
remains the same)
Discrete Data 1 proportion 2 proportion Chi-square test
Test of
Proportion

Eg-
1. You are working on a project to reduce the time it takes to deliver
pizza. One of the causes that came up during the brainstorming is
the equality of dough. You believe that the new quality of dough
takes less time to bake pizza as compared to existing dough which
 takes on an average of 13 mins. To bake the pizza. you have to
take 18 samples baked with new quality and note the time.

H0: Average Time taken to bake pizza with new quality of dough =13
mins
Ha: Average Time taken to bake pizza with new quality of dough <13
mins
1 sample t test

2. A company has stated that their straw machine makes straws that
are 5 mm in diameter. An employee believes that the machine no
longer makes straws of this size and samples of 100 straws were
taken to perform the hypothesis with a 95% confidence level.

H0: diameter =5mm

Ha: diameter != 5mm
1 sample z test

3. AHT/Average handling time 2 shifts which handles the customer

service call. Yoh have an assumption that the night shift has a
better performance as compared to day shift.

H0: AHTnight = AHTday

Ha: AHTnight < AHTday
2 sample t test

4. A drug manufacturer believes that its new drug will help reduce
the glucose level for patients suffering from diabetes. Glucose level
of 14 patient were noted and drug was provided to 14 patients…

H0: Glucosebefore=Glocoseafter
Ha:Glucosebefore>Glucoseafter
2 sample paired t test

5. 1 way ANOVA
6. 2 proportion test
….

p-value is the probability of null being true….

If value of p<0.05 … we reject the null hypo

p>0.05 …. We fail to reject the null hypo

Mantra:
if p is low null will go
if p is high null will fly
Similarly for 1 sample t test, z test, 2 sample t test , 2 sample
paired t test, 1 way anova

Now for 1 proportion

For 2 proportions always check the normal approximation value
Use p-value (pearson wala)

Scatter plot-
It is used to check whether the relationship exists between x and y
or not….

R= correlation coefficient (check the linear relationship of variables)

-1<r<1

0<r<0.5 weak
0.5<r<0.7 Moderate
0.7<r<1 strong

Regression- Predictive model

Y= mx +c
--- equation of best-fit line

1. -1<r<-0.7 or 0.7<r<1
2. R2 >50%
 3. P<0.05
Only then we can put the value of X and get Y
Module 5- Improve phase

Brainstorming principles to generate solutions-

1. Channeling
2. Anti solution
3. Analogy
4. Brainwriting

After finding the solutions we need to prioritize them…

For eg-
a. Payoff matrix
Sol 1,2 – High benefit, low effort -- yes
Sol 3,6 -

b. NGT – nominal group techniques

c. N/3 Voting
 d. Criteria-based

matrix (use mode to get the value)

FMEA – Failure Mode Effect Analysis

Iska main USP hai to find failure before it occurs…
(Control is to prevent it from happening, monitoring is after the
failure has occurred..)

RPN (Risk priority number) = S* O * D (Severity , occurrence ,

detection)

- 5 S- Improve the productivity

- Kanban –
Card system or flag system- the main objective is to reduce work-
in-progress
- SMED(Single Minute Exchange of dies)
- Spaghetti diagram
- Kaizen (continuous change)
- Poka Yoke (Error proofing )
 Module 6 – Control Phase

There are 2 things in control phase –

1. Project documentation
2. Control charts

Here a

Control charts are used to find out whether this process is stable or not

If any data point lies outside the control limits then it is a stable
process…
But there are also other special cause variations within the control limits
Not within control…