I use Chrome to benchmark my local version of a website. The use case is very limited, because the results fluctuate a lot depending on the overall load on my machine (including the dev server), so I usually do three passes on the master branch of my project, get scores like 53, 56, 51, then switch to my development branch, three passes again, if I see e.g. 58, 56, 61 then it's fine, if the results are noticeably lower than master then time to investigate.

In any case we have a dedicated environment with lighthouse-ci and this is where the real pre-release benchmarks are executed.