As your article also suggests, customer feedback is about so much more than bits of paper and any business that takes feedback seriously needs to acknowledge that.

Thanks for the query and for allowing me to go into a little more depth than I could have done in the article.

There were two reasons I chose 1 standard deviation in the examples above. Firstly, with limited data (there are only about 12 months), the degree of variance was limited and, so, for illustrative purposes, it was better to show something breaching 1 standard deviation than, say 3. Given data on a daily or weekly basis instead of monthly, I suspect there would have been scope for more than one standard deviation.

Secondly, there is the issue with the upper limit. As we are close enough to 100%, 3 standard deviations would probably be larger than the range of responses ever could be which I though, again, would detract from the legitimacy of the principle I was trying to illustrate.

I’m aware that 3 standard deviations is the norm if you are following the Demming/Six Sigma approach and would say that, if you were using the approach to monitor, say, visitor numbers, that would be the approach I would take.

So, I would agree that, in the longer term, 1 standard deviation is too low and that if you have data that stretched over a much longer period, then you should at least have 2 standard deviations at a lower control limit level. I would see 1 standard deviation as a kind of calibration exercise.

Is a 1 in 3 chance that an out of control point is actually part of the expected variation not a little high?