Lean Six Sigma in a Snapshot
The objective of this document is not to be an elaborate book, but for the CSSGB Trainees to
understand Lean Six Sigma in an easier manner. During the training program, these topics would be
covered in greater detail.
Why Six Sigma?
Helps companies improving their bottom-lines, with quality and customer satisfaction as by-
Helps companies eliminate waste, and improve their process efficiencies by respecting people.
How Lean and Six Sigma come together?
Lean should be applied before applying Six Sigma. Example, if you want to measure a boy’s running
capabilities for a 100 meter race, you have to only do that if he is Lean. If he is fat, he will show
different readings. So, first reduce the fat (Waste) and then improve his efficiency (Six Sigma).
Top management is the most important buy in for any Lean Six Sigma project
Many organizational roadblocks need to be overcome before someone decides to
implement Six Sigma.
Professionals shouldn’t hurry in starting with Lean Six Sigma. First, they should work on all
the pre-requisites and qualifications of a Lean Six Sigma project before they even get
Lean and Six Sigma projects are not done alone.
Lean Six Sigma projects are Continuous Improvement projects, so the objective is to make
sure the organization continuously improves from where it is today.
Statistical definition of Six Sigma --- When the difference between Process Mean and nearest
spec limit is 6 times the standard deviation, the process works at Six Sigma levels. Sigma
means Standard Deviation.
The Define Phase starts by the team understanding what the customer wants. In other
words, they listen to VOC, with the help of a tool called Focus Groups.
VOC is subjective and is wordy. This needs to be converted to something measurable. To do
so, we use QFD Matrix.
QFD is a tool one would use to translate customer needs to functional requirements.
The QFD will tell you which are the important Ys that need to be improved. Improving the Y
is going to bring in customer satisfaction and increased profits.
Another tool that can be used is CTQ Tree, where the Customer needs (VOC) is broken down
into key product characteristics and each characteristic is further quantified.
Assume you have an existing process, and the customer has already rejected a lot of your
products. Suppose most of them have said one specific quality is bad, that becomes your
Critical Y for the project.
You should also draw SIPOC map showing the overall process/macro process for the Y.
Finally, you should update the project charter with your problem statement and any
boundaries you may think of.
A cup of Dal in a hotel is the product that you as a customer want. You tell the hotel waiter the dal
needs to be in a certain manner. In other words, you are telling him your VOC. He takes the VOC and
works on what needs to be done in the kitchen with the chef. Basically, he develops the CTQ. He
says, 100 ml of water, 50 grams of Turmeric Powder, 2 teaspoons of salt, 2 tea spoons of Chilli
Powder, heating for 20 minutes in Medium flame and so on.
All these are CTQs.
For example, all this while, the customers have said, the Dal is too spicy. It means consistently Chilli
Powder has been added differently. That becomes your output (Y) for the project. In other words,
the project should be done on Controlling the spice content of the Dal.
Objective of measure phase is to validate measurement system, collect baseline data and
understand the inputs that we may be measuring.
1. Understand what type of data you are measuring for Y – Continuous, Discrete or Attribute.
2. Do a Measurement System Analysis GAGE RR for Continuous Data to measure the validity of
your measurement system.
Repeatability is a measure of how reliable your measurement system is
Reproducibility is a measure of how precise the operators are
GAGE RR also measures part variation, showing interaction between operator and
When you use the ANOVA tool with GAGE RR, the Error term that you see in the
ANOVA Sheet represents repeatability.
Always fix EV first, then move to AV.
If GRR on GRR Toolsheet is less than 30%, GAGE is acceptable.
Gage acceptable means you can live with the measurement system.
3. After doing GRR, collect data using sampling.
4. As per CLT (central limit theorem), sample mean or mean of sample means is equal to
population mean when sample size is > 30.
5. By sampling you collect a sample statistic, based on which you pass an assessment about the
population parameter. Basically, sample statistic is the mean and standard deviation (SD).
Population parameter is also mean and standard deviation (SD).
6. Once you collect the data, check the stability of a process. If the process is not stable, you
cannot do anything with the process.
A stable process will only have CCV (Common Cause of Variation). All processes will
A process goes unstable due to SCV. SCV means the process behaviour has changed.
Some times this change can be for good also.
Identify SCV with the help of Run and Control Charts Find if they are desirable for
the process If they are, let them be If they are not, knock them off or eliminate
7. Process Capability or Cp is how good the process is in delivering what the customer wants.
Process Capability Index or Cpk is Cp with consistency.
Cpk will be less than Cp if the process mean is not centered
Cpk will be equal to Cp if process mean is centered
Cpk calculations can be done only on continuous data. On attribute data, the
measure of capability is how much good or bad products produced out of total
Cpk value is more important than Cp because it helps you know if mean is centered
8. After doing your Cp and Cpk calculation, you would know three things --- 1) Is my mean
centered, 2) What kind of variations exist in my system, and 3) What should I analyse first
Mean Shift or Variations?
9. Check if your data is normal with the help of Normality test. Normal processes are used in
most production processes and are considered the easiest to work with.
1. Use Scatter and SLR to validate the impact of X on Y. You want to know if input identified by
you with the CE Matrix in Measure Phase is accurate or not.
2. Once you have identified X as critical, you need to know which factors impact X.
3. Brainstorm with team to find out reasons.
4. Draw a Pareto to prioritize
5. Draw a Multivari Analysis if you wish to study all observations and trends
6. Draw a Fishbone if the problem is complex
7. Conduct Hypothesis Tests to know what type of variation exists
8. All these will help you doing a Root Cause Analysis. At the same time, you don’t have to
conduct all the tests mentioned from 4-7.
1. Put the Root Causes from the Analyze phase on the table.
2. Deliberate on it using brainstorming and multivoting.
3. Once a solution is selected, re-check with your FMEA to see if you are on the right track.
4. Test the solution. Give it a test run or run it on a small team to see its effectiveness.
5. If it is effective, run a Paired T test to statistically validate the effectiveness.
6. If none of your solutions turn out to be effective, ask your BB to do a Designed Experiment
setup for you.
7. Designed Experiments is a series of sequential experiments done to test various factors at
different levels that gives you the desired or the best model.
8. Once your pilot tests are successful, put them enterprise wide.
9. Re-do a paired t test on the enterprise and see if the improvements are controlled.
10. Document your key improvement measures that are successful (Did you Pokayoke anything,
Did you change anything)
If all has gone well thus far, in the Control Phase, select a few samples of data and put them
in your Control Charts.
Check if your process is still in control or not.
Validate your Measurement System once more, to measure the reduced variability.
If the process is in control, you can re-calculate Cp and Cpk and other results.
Update the FMEA with revised RPN.
After doing all these, it is time to celebrate….