Load Testing Insights

What the Insights tab shows

The Insights tab of a load test result (cloud, k6 or JMeter) summarizes the whole run in a few numbers: how many requests ran, how fast they were, how many failed and which response codes came back. Open it first after a test, then move to the other tabs for detail. It fills in once the test has finished; until then it shows "Results will be visible once the test is completed."

Key metrics explained

Samples Total number of requests executed during the test.
Avg. Response Time (mean) Mean time in ms for all requests.
Min (ms) - Max (ms) Fastest and slowest request durations.
Hits / sec Average number of requests per second the application received.
Total Errors and Errors % Failed requests (4xx and 5xx responses or other connection errors) and their share of all samples.
Percentile 50% (median) and Percentile 90% The response time that 50% and 90% of requests stayed under. They show typical and upper-bound delays.
Percentile 95% and Percentile 99% High-percentile latency, useful for checking an SLA.
Sent KBytes / s and Received KBytes / s Average upload and download data rate generated by your test.
Standard Deviation How spread out the response times are around the average. A high value means inconsistent performance.
Response codes Counts of 2xx, 3xx, 4xx, 5xx and Other responses, with a bar chart. Hover over a segment for the exact count and percentage.
Response time distribution and Apdex A histogram of every sample by response time, with an Apdex score (T = 500 ms). See Response Time Distribution.

How to use Insights in LoadFocus

Run your test. Start a cloud URL test, a k6 script or a JMeter .jmx as usual.
Open the Insights tab. After the test completes, click Insights in the tabs above the results.
Pick a request. Use the Request dropdown below the tabs to see the numbers for one request instead of the whole test.

Interpreting your Insights

High average, low median If the average is much higher than the 50th percentile, a subset of requests is very slow. Look for outliers in the Timeline or Errors tabs.
Throughput plateau If Hits / sec flattens or drops while virtual users keep increasing, resources may be saturated, on your servers or on a load generator. The Overview shows a saturation warning when a load generator peaks at or above 80% CPU (see Load Generator Saturation). JMeter runs also have the Engine Health tab with the full CPU and memory series (see Engine Health Monitoring).
High standard deviation Large swings often point to external dependencies (third-party APIs, database locks) or garbage collection pauses.
No errors but high latency Zero errors does not mean all is well. Long response times still hurt users; check server-side logs or profiling data.
Errors above zero Open the Errors tab for the failing requests and their response codes.

Best practices

Baseline with a light load. Run a small number of users first, then compare your full-load Insights with that baseline.
Drill down by request. Use the Request dropdown to isolate slow or error-prone requests.
Correlate with other tabs. Read Insights together with Timeline and Errors, and with Engine Health on JMeter runs.
Set pass/fail thresholds. Limits such as a maximum P95 response time or error rate mark every run pass or fail automatically. See Pass/Fail Thresholds.