Analysis of Lean Six Sigma

This is a post in my organizing “between and beyond” series. Other posts are here. The purpose of this post is to provide a high level analysis of Lean Six Sigma.

Background
I first encountered Lean Six Sigma (L6S) two years ago (2017). This post is based on my L6S Yellow Belt and L6S Green Belt trainings. I am now L6S Yellow Belt certified, on my way to become L6S Green Belt certified. This is required by the company where I’m currently working.

The decision-making in L6S is data-driven. This is part of the long-term, cultural transformation which L6S strives for. L6S is about data! The L6S problem solving is done in five clearly defined phases (Define, Measure, Analyze, Improve, Control). The goal is to identify the most significant variables affecting the output of the process. A rule of thumb is that 85% of the variation in any process is controlled by two to five critical process input variables.

Objectives
The objectives of L6S are to:
– Identify the biggest problems
– Assign the best people to fix these problems
– Provide them with all the tools, resources and support needed
– Guranteeing them uninterrupted time to focus on permanently eliminate the problems

Assumptions
The assumptions in L6S are that:
1. A system can be put in place to ensure that improvements are maintained for the long term when these critical process input variables have been identified.
2. The need to inspect and measure the process outputs can be eliminated, if the process inputs can be controlled. The inputs can be considered the causes, while the outputs are the effects.
3. The process inputs that need to be controlled can be found by a structured, problem-solving methodology.

Incompatibilities
L6S acknowledges that the equation that relates process output to process inputs doesn’t have to be simple. Process outputs can, for example, be a function of many process input factors. Some of these factors may affect the output in a non-linear fashion. The factors can also be co-dependent. Interactions might, in other words, be going on between the variables. My question, however, is how often an equation can be determined that relates process outputs to process inputs in a business environment?

Conclusions
I used to be an authorized Personal Software Process instructor and certified Team Software Process coach, and have personal experience of using the kind of statistical process control in L6S on myself, in my projects. My conclusion is that it is only in special cases that it is possible to determine an equation that relates process outputs to process inputs. Cognitively heavy processes (that is, processes where you think a lot, like in software development) are difficult to put under statistical process control. I have tried! I don’t think my personal software process ever was under statistical control.

I suspect that it is not possible to determine an equation that relates process outputs to process inputs for the great majority of processes. Work that requires creative thinking is by its very nature unpredictable. There is a claim in L6S that most processes are only 3% to 5% value add, and that the majority of processes are non-value add. Maybe, just maybe, only 3% to 5% of the processes are under statistical process control?

Related posts:
Analysis of the CMM, PSP, and TSP