Statistical process control after W. Edwards Deming

> there isn't a hard theorem why they should sit at +/-3σ

There is Vysochanskij-Petunin inequality [1]. 95% of observations should fall within +/-3σ away from the mean, assuming unimodal distribution and finite variance.

[1] https://en.wikipedia.org/wiki/Vysochanskij%E2%80%93Petunin_i...

Thanks, I wasn't aware of this and it seems very useful! Good to know a tighter bound than Chabyshev's inequality (assuming unimodality).

It's also where Six Sigma (3 sigma on each side) comes from. It implies "Total Control".

Online resources on applying Deming's ideas https://blog.deming.org/online-deming-resources/

If I remember correctly, Deming was a key inspiration for Charles Koch's (of Koch industry fame/notoriety) managerial strategy.

Philip B. Crosby and Joseph M. Duran

It's actually Joseph M. Juran, sounds the same if you're listening to an audiobook.

Yes, you’re absolutely right. Iphone spellcheck likes Duranduran. My department worshipped Juran when we were manufacturing things.