A release ticket says the portal must support 120 users. If those users are submitting forms, the first thing I’d ask is when the submissions arrive.

A quiet average can hide a busy window

Take a fictional portal. Over 15 minutes, 120 people each submit three requests. That’s 360 submissions in 900 seconds, or 0.4 submissions per second on average.

Now put 90 of those same submissions in the first 30 seconds. You still have 360 over 15 minutes, but the first window averages 3 submissions per second. If a registration window opens at 9:00, those first 30 seconds could matter more for the release decision than the 15-minute average.

Here, one business transaction means a submission that reaches confirmation. A submission can involve several HTTP requests, and the figures above describe rates of submissions. Concurrent sessions and HTTP request rates are separate inputs to the test.

What I’d clarify before scoping the test

  • What counts as success? I’d measure from Submit to confirmation if the user needs proof their request was received. A fast form load tells me about a different part of the journey.
  • What concentrates demand? An opening time, a reminder email or a deadline can put a large share of submissions into a short window. If there’s traffic history, use it to check the assumption.
  • What would a user notice when it goes wrong? Watch for delayed confirmations, failed submissions and retries. Agree with the team on which outcome would change the release decision.

The figures are illustrative. I’ve shared the original workload calculation and its assumptions in Platform Performance Brief on GitHub.