Response Time Distribution
Response time distribution
The Insights tab of a load test result shows a histogram of every sample in the run, grouped into response-time buckets. Each bar counts how many requests finished within that range; the axis starts just under the fastest response, so the chart uses its full width.
Why an average is not enough
An average of 300 ms can come from two very different runs: one where every request took about 300 ms, and one where half took 100 ms and half took 500 ms. The percentiles on the tiles above narrow it down; the histogram shows the whole shape at once.
How to read the shape
- One narrow peak - consistent behaviour. The average and the median are close and both describe the run.
- A long tail to the right - most requests are fast, a minority is slow. Look at P95 / P99 and at the slowest requests on the Overview to find which ones.
- Two peaks - two populations: cached and uncached responses, two backends behind a balancer, or one slow load-generator location. Check results by location and the per-request Stats table.
- Bars far to the right of the peak - timeouts or retries. Cross-check the errors by type table.
The buckets are computed from the individual samples, so the counts add up to the Samples tile.