Cross Validated
2026-11-02 11:12 UTC
By Erik Rasmussen
AI-113-20261102-social-media-d45398e3
Calculating Safety Stock based on percentiles of real observations
I am trying to determine the optimal amount of safety stock to hold for a given product, but my demand history is not normally distributed. I have a large number of days with 0 sales, and a long tail of some days with very high sales. One way I've thought of doing this is instead of using the standard formula of Z factor * standard deviation, to just calculate the percentile of my data based on the service level I want to achieve, i.e. for a 95% service level calculate the 95th percentile of by demand history and work with that. Would this approach cause any statistical issues?
I am trying to determine the optimal amount of safety stock to hold for a given product, but my demand history is not normally distributed. I have a large number of days with 0 sales, and a long tail of some days with very high sales. One way I've thought of doing this is instead of using the standard formula of Z factor * standard deviation, to just calculate the percentile of my data based on the service level I want to achieve, i.e. for a 95% service level calculate the 95th percentile of by demand history and work with that. Would this approach cause any statistical issues?
Full article content could not be extracted automatically. Read the original below.
Source:
Cross Validated
· stats.stackexchange.com