How to Measure API Response Time Properly
By SyncBuilder
Response time is one of the most important qualities of any API you build an integration on. When an API is slow, every workflow that depends on it slows down too, so knowing exactly how fast it responds is not optional. The tricky part is that a single measurement rarely tells you the truth.
Why one measurement is not enough
Latency is never a fixed number. It shifts with network conditions, server load, cold starts, and rate limiting, so the same request can be fast one moment and slow the next. To get a reliable picture you need to send several requests and look at how the results are spread out, not just at one lucky or unlucky call.
Look at the distribution, not the average
An average smooths everything into a single number and quietly hides the worst cases. That is why you should read the distribution instead: the median (p50) tells you the typical experience, while p95 exposes the tail, meaning 95% of requests are faster than that value. A large gap between the median and p95 is a warning that some requests are far slower than they look on average.
- Measure latency across multiple runs rather than trusting a single call.
- Compare the median (p50) with the p95 to see how wide the tail is.
- Watch whether a cold start skews the very first request upward.
- Check whether rate limits start throttling once you send more requests.
A single fast response can hide a slow tail. Always look at p95, not just the fastest or the average time, before you trust an API in production.
Our free API Speed Test does exactly this for you: it sends several requests and reports the min, average, median, p95 and max. It is free and runs entirely through a stateless proxy in your browser, so you can measure any endpoint in seconds without setting up anything.
Try the free tool