Eh. I did a bunch of statistical modeling for cloud service SLIs/SLOs/SLAs. This approach is good for letting people do black-box modeling. It sounds like complete overkill for recipes.
A recipe is not a black box. Most recipes are short and can be written on a half page of paper. If it calls for two chopped onions and you know that it takes you an “eternity” to chop onions, then add two eternities to the cooking time.
And as long as you are spending two eternities chopping onions, why not spend 90 seconds watching a video on YouTube that explains how to do it correctly? For most people—if you are cooking regularly, and practicing good technique, you will naturally get faster over time. If, like many people, you are cooking for decades, then you will be slow for only a short time.
I forgot an /s. 99.9 percentile would require doing each meal in the cooking book ... 1000 or 10 000 times by different people to get an accurate timing :)
> chopping onions, why not spend 90 seconds watching a video on YouTube that explains how to do it correctly?
That sounds like a good idea, could get back 1 000 x those two minutes, in a lifetime, and more yummy food and can impress others with the onion skills. (No /s here. I'll try to remember tomorrow, youtube.)
(E.g 99.9 being 40 min would mean that 999 out of 1 000 people got it done faster than 40 min)