Ask “why?” 5 (or more) times to tunnel down into the root cause. The answer to the first why is almost always an obvious symptom. The secret behind the 5-whys technique is to accept the answer, but to then ask why again and again until the root cause is uncovered. Sometimes, the root cause can be found at the fourth or five why. Often, however, you must ask “why?” more than 5-times.
What is-What isn’t Analysis
Often, listing what a problem is and isn’t helps get to the root cause by a matter of elimination. What Is-What’s Isn’t questions include: What happened? & What might you have expected to happen but didn’t? Where did it happen? & Where didn’t it happen? What changed in the process? & What didn’t change in the process? Which supplier was involved? & Which wasn’t?
Data Collection & Data Display
Fact-based problem-solving – that’s what root cause analysis is all about. To get facts, collect data from the process or create data related to the process. To get facts, we collect data from the process or create data related to the process. Once data have been collected, there are a number of simple methods to analyze data using graphical display techniques. Data display tools turn the data into pictures and a picture of what has happened often leads to the root cause.
Techniques for collecting data from failure analysis include reviewing physical evidence (much like crime scene investigation), special testing, accelerated testing, and finite element analysis. You might need special tools or techniques to review the physical evidence (e.g. microscopy to look at a break surface) or you might need to conduct special testing on the product or process itself. Use well-designed and easy to use data collection forms. Good detective skills can turn interviews into effective data collection events. One of the most powerful, but also most under-used, data collection tools is a concentration diagram.
Simulations can be used to collect data using computer modeling software, pilot-plant experimentation, and if need be, experimentation using the actual process itself. With the proper model, a computer could help point the way to the root cause. Or it might be pilot-plant trials or experimentation using the “real” process that generates the data that leads to the root cause. In any case, if you can recreate the problem, you are more apt to find the root cause.
While data display methods are usually easier to use, sometimes a statistical analysis technique is needed to wring the real meaning out of the data. SPC control charts will actively signal a problem with a process. Correlation and regression analysis and multivariate analysis may be needed to make sense of the data.
The “Root Cause” Question
Once you think you are at the root cause, take a step back and ask yourself the root cause question “Does this cause explain all that is known about what the problem is, as well as all that is known about what the problem isn’t?” This is really a two-part question: make sure the root cause found fits both the “is” and the “isn’t” sections of the question. If the cause being tested doesn’t fit both, then it’s probably not the root cause.