Skip to content

Replay Is Much Slower Than Recording

Replay completes successfully, but the wall-clock time is several times the recording. The recording took a minute, the replay takes ten. Almost always, this is think time doing exactly what it's supposed to do, plus a slower test environment than production. Both are normal. Neither is a problem you need to solve unless the slowness is actually blocking your work.

Root Causes

  • Think time delays: The default Play mode replays at "limited think time," which includes pauses between pages so the replay looks like a real session. Recording captures the actual user pauses, which can be long. Replay applies them.
  • Server is slow: Test environments routinely have weaker hardware than production. A test server with one CPU and 4 GB of RAM will not respond as fast as a production cluster.
  • Network latency: If the replay runs from a different network location than the recording (different region, slower link), every request adds latency.

Solutions

Solution 1: Use Play Fast Mode (Remove Think Time)

To eliminate think-time delays for the cases where you don't need them:

  1. Click the dropdown arrow next to the Play button in the toolbar
  2. Select Play Fast (zero think time)
  3. Replay runs immediately without pauses between pages

For deeper coverage of all five replay modes, see Replay Modes.

Use Play Fast for:

  • Debugging configuration issues quickly
  • Verifying the test case works without waiting
  • Single-user functional testing

Don't use Play Fast for:

  • Realistic load testing (think time matters for accurate load simulation)
  • Performance benchmarking (no real user clicks instantly, so the timing data is misleading)

Solution 2: Verify Server Performance

If the server itself is slow:

  1. Check server resources (CPU, memory, database performance) using Server Monitoring
  2. Coordinate with operations/development to investigate slow responses
  3. Compare response times between recording and replay in Headers View

Test environments often have slower hardware than production. This is expected. As long as replay succeeds (green transactions), slower performance is acceptable for configuration verification. Save the actual performance measurement for load tests run against a representative environment.