Percentile 90 or P90 is the value below which 90% of the data fall. For example, let's say you have 1000 User clicks, and a P90 value of 8 seconds, then you would know that 900 of those clicks would be shorter than 8 seconds and the remaining 100 (10%) would be longer than 8 seconds.

Same logic applies to p95, p99, p999

In Germain, we need to be able to provide percentile values without having access to the raw values, we use an approximation algorithm called t-digest which is described in greater details here: https://www.sciencedirect.com/science/article/pii/S2665963820300403#:~:text=The%20t%2Ddigest%20is%20an,of%20data%20with%20arbitrary%20distribution.