Results

History and reference runs

Every performance test and compare is kept on your computer. Find old runs, add notes, and pick a reference run to see whether later runs got faster or slower.

Every run is kept

Each finished performance test and compare is saved automatically — on your computer, as a plain file. Open History in the sidebar to see them, grouped by day (Today, Yesterday, then the date). The line under the title counts them and shows how much space they take — for example 12 performance tests · 3 compares · 4.2 MB on this computer.

  • Search by name, URL, or note.
  • Show: All, Performance tests, Compares, or Reference runs.
  • Each performance test shows its result, address, p95, error rate, and time. A ◆ marks a reference run.

Click a record to open it on the right:

  • Rename it by clicking its name; add a note — what changed, which build, why you ran it. Both save as soon as you click away or press Enter.
  • Show file opens the folder with that record's file selected. The file's path is shown under the name.
  • Delete removes that record.
  • A performance test opens its full report, exactly as it was.
  • A compare shows each API with its result, p95, error rate, and request count. Open overlay opens it on Compare with the full results.

Show folder at the top opens the folder where all the files live.

Select several

Tick the box beside records to select them. A bar shows how many are selected:

  • Overlay — see the ticked performance tests on the same charts on Compare. It needs 2 to 6 performance tests and no compares ticked.
  • Delete — removes every ticked record. You are asked first: This cannot be undone — export first if you want a copy.
  • × clears the selection.

Before / after

Every finished report of a request you've run before has a Before / after section: the answer to “did my release make it slower?”

  • It starts with one plain sentence — “After is 18% slower: 95% of requests within 142 ms → 168 ms. That is more than normal run-to-run variation.” — or “About the same.”
  • Then each measure side by side — typical, p95 and p99 response time, requests per second, failed requests, checks — marked better, worse, or about the same.
  • Then both runs on the same charts.

What counts as a change. Two runs of an unchanged API never match exactly, so differences under 5% (or under 10 ms) count as about the same. More failed requests or fewer passing checks always count as worse, whatever the speed.

Fair comparisons only. If the two runs tested different requests, or used different load (20 users vs 50, simulated users vs a request rate), the section says “Not a fair comparison” and shows the numbers without calling a winner. Runs where most responses were errors don't get a speed verdict.

Which run is “before”? Your reference run of the request if you've marked one, otherwise the previous run (or, for the oldest run, the next one). Pick any other run of the same request with Compare with. The two runs are always shown in time order — before is the earlier one — and the open report is marked this run.

Do all settings need to match? Two things must: the request (method and address — the query string may differ) and the load (the same kind — simulated users or requests per second — and the same peak). Length may differ; if it differs by more than a quarter, the result says so. Headers, body, checks, goals, think time, and the ramp shape aren't checked — for a meaningful answer, keep everything the same except the thing you're testing.

Comparing different APIs, or more than two runs? Use Add to compare on each report — see Compare APIs.

Reference runs

A reference run is a good run you compare later runs against — for example the last run before a release.

  1. Open a run you trust and choose Use as reference run (in the report or in History).
  2. Run the same request again later. Its report compares with the reference automatically, and the HTML export includes the same Before / after section.

The trend panel

When History holds two or more runs of the same request, opening one of them shows p95 trend for this endpoint above its report:

  • a chart of p95 for every run of that request, oldest to newest, with your reference run as a dashed line;
  • a table of the 8 most recent runs — p95, throughput, and errors — with the change against the reference run (▲ worse or ▼ better, or ≈ same).

Click a point or a row to open that run. Without a reference run, the panel says pick a reference run to see change. It makes a slow drift over weeks visible. In a pipeline, the same comparison can fail the build — see Command line and CI.

Overlay runs

Tick 2 to 6 performance tests in History and choose Overlay to see them on the same charts on Compare: a Comparison tab and one tab per run. They can be runs of different requests. To pick runs from different pages instead, use Add to compare — see Compare APIs.

Opening a report file

Open report in the sidebar opens a JSON export — a single run or a compare, yours or one a colleague sent. Drop the file on the page, or click to choose it.

  • A run opens with its full report and charts.
  • A compare opens with its overlaid charts. To run those APIs again, open Compare.
  • Keep in History copies the file into your History; the button then says Saved to History.
  • Replace JSON opens a different file.

HTML, Excel, and Zip exports are for sharing and can't be reopened here — keep the JSON export if you may want the report back.

No app at hand, or sending a report to someone who doesn't have it? The report viewer does the same in any browser, and also opens the Zip: History, the trend, reference runs and compares, on the files you drop. Nothing is uploaded. Reports saved by older versions are opened with their numbers corrected where newer versions measure better.

Where the files are

History files live in your app-data folder — see Privacy and your data. They are ordinary JSON files: you can back them up, copy them to another computer, or delete them by hand. If a file can't be read, History lists it and offers to delete it.

Something unclear or missing? Write to support@apitestcenter.com — the guide improves with every question.