Response Time
Learn how response time measurement is collected and how to read trends.
Measurement Method
The frontend probe measures the time in milliseconds between the start of the request and the completion of the response using the SSRF secure HTTP client. The measurement represents the total request time of the client; TTFB, TLS handshake or transfer time are not reported as separate fields.
The /metrics endpoint returns the average and p95 summary for the selected period.
Percentile Values
Percentile allows you to look at the overall distribution, not the worst individual jumps. p50 is useful for understanding the median experience, and p95 is useful for understanding the upper limit that the vast majority of users see.
There is no p50 or p99 in the API output; average_response_ms and p95_response_ms are returned. This chart is an example visual to illustrate the concept of percentile.
p95 is one of the most useful alarm indicators.
Even if the median value looks good, p95 shows more clearly the sudden queue backlogs and delays users experience at the edge.
Time Windows
- 24 hours: ideal for deployment, cache cleanup or to examine momentary performance degradation.
- 7 days: suitable for seeing the weekly traffic pattern and recurring peak hours.
- 30 days: A more balanced window for SLA and overall service quality assessment.
- 90 days: useful for reading the long-term performance trend in the dashboard. Old data is kept for a certain period for continuity; the /metrics endpoint accepts 24h, 7d and 30d period values.
Interpreting Values
| ms interval | Evaluation |
|---|---|
< 200 |
Perfect |
200–500 |
Good. |
500–1000 |
Average |
> 1000 |
Improvement needed |
These are not in-product fixed thresholds.
The values in the table are a guide for general web performance practice. Anchor Uptime provides raw ms values and a summary of p95_response_ms.