Performance Health
Performance Health turns the Core Web Vitals data your site already collects into a clear picture: a rating for each metric, a single overall score, a list of anything that has gotten worse recently, and a short written summary.
What this does
Performance Health does not collect any data of its own. It reads the measurements gathered by Real User Monitoring (RUM) and adds a diagnostic layer on top:
- Metric ratings. Each Core Web Vitals metric is classified as good, needs improvement, or poor based on its 75th-percentile value (p75).
- A composite score from 0 to 100, calculated from LCP, INP, and CLS.
- Regression detection. It compares the last 24 hours against the previous 7 days and flags metrics that have gotten noticeably worse.
- An AI narrative. A short, plain-English paragraph explaining the current picture and what changed.
Why use it
Raw performance numbers are hard to read at a glance. Performance Health tells you in one place whether your site’s vitals are healthy, gives you a single score to track over time, and points out exactly which metric slipped and by how much, so you know where to look.
Requirements
- xSpeed Cache Pro must be active.
- Real User Monitoring must be enabled, because Performance Health reads the data RUM collects. RUM is enabled by default. If it is turned off, no data flows in.
- The AI narrative needs an AI provider configured. If one is not available, the metric ratings, score, and anomalies still work; only the written summary is skipped.
The metrics
Performance Health classifies each metric’s p75 value against the standard Core Web Vitals thresholds:
| Metric | Good (up to) | Needs improvement (up to) | Poor (above) |
|---|---|---|---|
| LCP | 2500 ms | 4000 ms | 4000 ms |
| FCP | 1800 ms | 3000 ms | 3000 ms |
| INP | 200 ms | 500 ms | 500 ms |
| FID | 100 ms | 300 ms | 300 ms |
| CLS | 0.1 | 0.25 | 0.25 |
| TTFB | 800 ms | 1800 ms | 1800 ms |
A value at or below the “good” limit is rated good. Above that but at or below the next limit is needs improvement. Anything higher is poor. If there is no data for a metric, it shows as unknown.
The score
The composite score runs from 0 to 100 and is built from the three core metrics: LCP, INP, and CLS. Each one contributes equally:
- A good rating contributes 100.
- A needs improvement rating contributes 50.
- A poor rating contributes 0.
- A metric with no data is left out of the calculation.
The score is the average of the metrics that have data. These three are used because they are the vitals that count as Search ranking signals.
Anomaly detection
Performance Health compares the last 24 hours to the previous 7-day baseline. A metric is flagged as an anomaly when its current p75 is more than 1.5 times its baseline p75, which is a regression of more than 50 percent.
For each flagged metric, it reports the baseline value, the current value, and the percentage increase. Anomalies are sorted worst-first. Improvements are deliberately ignored: the goal is to surface what got worse.
Note: Anomaly detection only fires when both the current period and the 7-day baseline have data for a metric. A brand-new site will not show anomalies until it has enough history.
The AI narrative
On request, or on a daily schedule, Performance Health asks the AI provider to write a one-paragraph summary. The narrative leads with the overall score and then calls out any metric that regressed against the 7-day baseline.
This step runs over aggregated, site-level data, not individual visitor data. If no provider is configured or the request returns nothing, the narrative is skipped and the rest of the panel still works.
How to use it
- Open the xSpeed Cache Pro dashboard and go to the Performance Health panel.
- Review the metric tiles, the score, and any listed anomalies.
- To refresh the written summary, run an analysis. This recomputes the status and asks the AI provider for an updated narrative.
The status also refreshes automatically once a day in the background.
Troubleshooting
The metrics show ”—” or unknown. That metric has no data yet. Confirm that Real User Monitoring is enabled and that your site has received enough traffic for samples to accumulate.
The score is 0. A score of 0 happens when the core metrics (LCP, INP, CLS) are all rated poor, or when none of them have data yet. Check the individual metric tiles to tell which case applies.
No narrative appears. The AI narrative needs an AI provider configured. Without one, the request is skipped. The metric ratings, score, and anomalies are unaffected.
No anomalies are listed. That is normal when nothing has regressed by more than 50 percent against the 7-day baseline, or when there is not yet enough history to compare.
FAQ
Does Performance Health collect data? No. It reads what Real User Monitoring already collects and adds analysis on top.
Which metrics affect the score? Only LCP, INP, and CLS. The other metrics (FCP, FID, TTFB) are still rated and shown, but they do not change the composite score.
How often does it run? It recomputes daily in the background, and you can trigger an analysis manually at any time.
Related
- Real User Monitoring
- AI Privacy and GDPR Mode
- SmartPredict Navigation