Tag Archive | Root cause analysis

Root cause analysis Rule #2

Correlation

Another classic mistake in Root Cause Analysis is to believe that correlation implies causation. It does not. In simple words, the observation of 2 events happening almost at the same time or 2 variables going up and down together is not enough to prove that one is the cause of the other. This is a bit tricky because some statistical tests (e.g. regression analysis) used in Lean or 6 sigma look for correlation and it’s easy to go too far and assume causation, but this is wrong.

This is an example. The daily average temperature (ºC, ºF) is correlated with the number of people (#) suffering a heart attack that day (and, probably, there is also a cause-effect relationship). The daily average temperature (ºC, ºF) is correlated with the amount of ice-cream (Kg) consumed that day (and, again, there is probably also a cause-effect relationship). However, the amount of ice-cream (Kg) consumed one day is correlated with the number of people (#) suffering a heart attack that day, but the probability of cause-effect relationship is low:

IMG_4463

Find more examples here: http://tylervigen.com/spurious-correlations

Some say that real causation can never be proved and therefore we only see correlation and imagine causation (more on this here). We can intuit that a cause-effect relationship is stronger than simple correlation. Anyway this belongs to the field of philosophy and in most cases common sense and observation can help us take an educated guess. The use of Ishikawa diagrams, 5 whys, DOE and PDCA experiments, besides the implication of the true process experts (those who work everyday there where the problem happened) can make a difference.

Picture from: http://jwilson.coe.uga.edu/EMAT6680Fa11/Shannon/STATS/Comics.html

Advertisements

Root cause analysis Rule #1

IMG_3909

Rule #1 seems evident. This is probably because it IS evident. However it’s easy to find RCAs (root cause analysis) that violate this rule. Here is an example I’ve recently seen:

Our machine was not clean before we started the new batch. The operator that verified it did not see that one piece of the machine was dirty. He was new to the job and failed to follow the standard checklist properly. The root cause of our defect is “defective cleaning verification due to incomplete training”.

Yes, you got it. The “defective cleaning verification” cannot be the root cause because it happened AFTER the problem appeared. The machine was already dirty at that moment, which is the real problem. The adherence to the standards and the correct training of all operators are, of course, fundamental but it is not creating this issue.

Our quality systems are based in 3 types of actions: those that eliminate the problem, those that mitigate the effect of the problem and those make the problem visible. All of them are important, but only the first solve the root cause forever and improve the quality of our work. Without proper RCA, we will have the same issues again and again. This is why RCA is so important.