What does p95 response time mean?

P95 is the response time at or below which 95 percent of successful checks completed. It excludes only the slowest 5 percent; it does not remove the fastest results.

How does SiteInformant compare deployment performance?

When you record a deployment marker, SiteInformant saves a baseline from up to 30 successful checks before the release and measures up to 30 successful checks after it, then compares the two p95 values.

Is SiteInformant an application performance monitoring agent?

No. SiteInformant measures external HTTP and HTTPS endpoint response time. It does not collect application traces or automatically identify the line of code behind a regression.

Endpoint performance

P95 Response Time Monitoring

See the slower edge of real endpoint performance, retain it across long-term history, and compare releases with an explicit deployment marker.

Monitor an Endpoint Explore Uptime Monitoring

A useful percentile

What p95 tells you

Average response time can look healthy while a meaningful share of requests are slow. P95 is the value at or below which 95% of successful checks completed. Only the slowest 5% fall above it; fast results are not discarded.

Successful checks: Failed checks remain availability signals and are not mixed into the latency percentile.
Retained history: Bounded daily histograms preserve p95 when old raw checks are compacted.
Visible context: Card and Table views show p95 alongside uptime and the latest response time.

Release comparison

Catch performance regressions after deployment

Record a release label and optional version on an endpoint. SiteInformant compares two concrete windows instead of guessing when a deployment happened.

Capture the baseline. Use up to the 30 successful checks immediately before the marker.
Measure the release. Collect up to the next 30 successful checks after deployment.
Compare p95. Keep the before value, after value, percentage change, sample counts, and a clear impact classification.

Built for honest interpretation

A comparison stays in a measuring state until 30 post-deployment successes arrive or the 45-minute analysis window closes. Sample counts remain visible, so a partial window cannot masquerade as a full one.

Know what it is not

This is external endpoint monitoring, not application tracing. It can reveal that latency changed after a release; your logs, traces, and profiler explain why.

Make deployment performance visible

Start with one endpoint, record the next release, and let two real check windows tell the story.

Try SiteInformant

Next: route and escalate the alerts that matter