Median calculation

Eng source doc: +PA Brain Dump

How we calculate time in stage, is the sum of all the time in each individual stage, for all stages.

If you have a pipeline with stages ABCD, time in stage between stage A and stage D is sum(TIS A-B, TIS B-C, TIS C-D). It is NOT TIS(A-D).

Avg and median calculations also build off this assumption. Avg = sum(avg TIS A-B, avg TIS B-C, avg TIS C-D). Median = sum(median TIS A-B, med TIS B-C, med TIS C-D).

Why?

• Pros
• You have more data points – e.g. earlier stages will have lots more candidates, so your statistic will be more representative
• you’ll include candidates who you didn’t hire, which could be an advantage – imagine a company where every candidate who takes a long time in the hiring process just drops out. That wouldn’t show up if you’re only looking at people you hired
• Cons
• it’s not the statistic people are expecting us to construct, which is confusing and may harm trust
• with median in particular, it’s probably not a very statistically sound thing to do

This results in funky numbers for median (Cedar example) https://app.asana.com/0/1205263594820271/1206683001203532/f.

Task for calculating median using total time, instead of sum of individual stages: https://app.asana.com/0/1199546868692983/1204631605724350/f.

• Median calculation