Part II. Methodical mistakes

What mistakes we may make in Six Sigma projects, and how to avoid them

In the previous Part, I collected several possible management mistakes in Six Sigma projects. Along these mistakes may arise methodical mistakes when we do not use the method or the tools properly.

1. Project charter mistakes: problem description is hazy, the problem is not clear: what is the product/service, what is the symptom (defect), physical location, time frame, extent and the effect of the problem. We must set a target, where we want to get (quality goal) with what timeframe, with what resources (people, production time etc.) and what savings we expect. Use SMART (Specific, Measurable, Achievable, Realistic, Time-framed) definition and do NOT forget to get management approval

2. Using a discrete variable as output: it is not necessarily a direct mistake but it complicates our life: it requires 1 or 2 magnitude more samples (if you want to detect a 1% difference in defect rate you may end up close to 10000 parts) for analysis. Try to translate your output variable into a continuous variable.

3. Measurement System not acceptable: most of the time we trust our measurement results. What if they are not precise or adding unnecessary variation? We draw not the right conclusion and make decisions based on that. Gage R&R for discrete variables also result in problems. Use Gage R&R studies not only for the output variables.

4. Sample size: If we start our measurement without knowing what difference we want to detect, we end up redoing the measurement. It ruins time spirit and credibility to management. Be focused, learn methods, define goals and use a sample size calculator.

5. Data Collection Plan: a proper data collection plan is a guarantee for success. However if we have not identified the possible input variables, not defined a proper output variable and not quantified the difference we want to detect, our statistical analyses do not function well and we are unable to identify root causes. Additionally, we must know statistical methods to be used. Create a Fishbone diagram, identify the possible vital few input variables, define the difference to detect, calculate sample size, consider time-dependent variables, determine time-frame, finally add the method how it will be evaluated.

6. Non-normal distributions: as we analyse data, we look for a pattern of what could be the root causes. If they are not normal is a sign of special causes, but we must be careful what statistical tools can be used. Most of the statistical tools were developed to analyse normal distributions, if we apply them to non-normal distributions, we draw wrong conclusions. The mean and standard deviation has no further information anymore, we must switch to non-parametric tests. Perform normality test before any tests are carried out.

7. Stability in time: non-normality is not the only sign for special causes, before any analysis we must look for how it behaves in time. Any pattern help to identify underlying causes. Use Time Series Plot.

8. False correlation: if you find a correlation between two variables that do not prove the logical connection. If you want to have some fun please search on the internet for examples, like As ice cream sales increase, the rate of drowning deaths increases sharply. Therefore, ice cream consumption causes drowning. 🙂 Understand physics and logic, challenge your assumption, use team.

9. Non-verified solution: if we find the solution looks the best, we must first perform a pilot solution, and based on those findings, we have to adjust the process, settings and training requirements. Do not forget to carry out a proper communication plan.

10. No or not properly supervised Control Plan: if we are not able to keep the new process in control, it will converge back to the old one. We ruin not only the achieved results but motivation and support will vanish. Put extra effort in Control Phase with proper communication, training, checkpoints.

(For the graphic special thanks to Sara Gránásy.)

by Péter Gránásy, Six Sigma Experter