The goal to ensuring quality was to see your business as a repeatable process. That way you could apply statistical methods, measure and optimize. The flaw wit

The goal to ensuring quality was to see your business as a repeatable process. That way you could apply statistical methods, measure and optimize. The flaw with six sigma was that it is applicable to only very repeatable manufacturing processes where the defects can stem virtually only from random variation in environment. Eg. room's temperature changes, worker sneezing, etc. In this case you are expecting many small probability defects compounded together, yielding a normal distributuon by the central limit theorem. That way you can apply all the statistical methods around the gaussian distribution and try to eliminate 'six sigma' of defects. The flaw is that people tried to apply this to highly explorative disciplines like software engineering, where the process is very far from being normally distributed. As Nassim Taleb likes to explain, most of the human processes follow an exponential distribution. In the case of software defects, you tend to have many small ones that nobody even logs anywhere and just fixes them (thus being invisible to anybody (PMs) trying to find a pattern). And the other category which are defects where you go down a deep rabbit hole to understand and the fix might require huge rearchitecting of the app. There is no normal distribution, averages on exponential distribution don't make any sense and delivery estimates don't either. Part of the reason is why lean and agile are the only successful methodologies in software is because they treat it as a discovery process which eventually can be turned into highly repeatable process - by that point operated 90% by a well designed machine. Many parts od the industry still got stuck with the flawed analogy between software development and construction business. They posit that both should have an architect to design the system and then an army of low paid workers to execute on the plan. The flaw is in thinking that the software developer is the 'worker' and not the architect. He is actually the designer and the worker in the analogy is the compiler, which performs its job so quickly, that project managers don't even see it to include it in their gantt charts. It's also sad to observe how the certification business is turning the original lean and agile ideas into another black belt type cargo scrum cults. It helps going back to the roots and listen to Deming, Juran, the guys around agile manifesto and especially system thinkers like Ackoff.