Judge load tests against a baseline

Save a run as the baseline, so later runs fail when they get slower.

In short

A baseline is a load test run you trust. Every later run of the same request, environment and load is judged against it, and fails when it is slower, handles less or errs more than you allow. The history draws each endpoint's trend over time.

Before you start

Who: saving a baseline needs Edit on the project. Anyone who can open it sees the verdicts.
Plan: load testing is part of the paid plans.
First: a finished load test of a saved request that you trust.

Steps

Run a load test of a saved request.
Under the verdict, press "Save as baseline".
Run the same request, environment and load again later.

Result: the run passes or fails against the baseline, and the result says which figure went over and by how much.

Good to know

By default a run fails in three cases. Its p95 is more than 15 percent slower. Its throughput is more than 10 percent lower. Or its error rate is more than 1 percentage point higher.
One baseline is active per request, environment and load. Marking another replaces it.
Each run keeps the tolerances it was judged by, so an old result reads the same after the baseline changes.
A baseline taken at 10 callers says nothing about a run at 100: compare like with like.
Every load test is kept below the results, newest first, with who ran it, the load shape, success rate, p95 and requests per second. "Over time" draws p95 and throughput per endpoint, so an endpoint that is slowing stands out before it fails.

If something goes wrong

The run is not compared: it had fewer requests than the baseline needs, or a different load or environment.