Calibre gives you Synthetic testing, CrUX and Real User Monitoring together. This guide helps you pick the right one for the question in front of you. None of them is "more correct" than the others: each is best at a different job.
Start with the question you are answering#
- Do you need to catch a regression before it ships? Reach for Synthetic.
- Do you need to know how Google sees your site for search? Reach for CrUX.
- Do you need to know how your real visitors are experiencing the site right now? Reach for RUM.
At a glance#
| Synthetic | CrUX | RUM | |
|---|---|---|---|
| What it is | Controlled lab test | Google's field dataset | Your own field data |
| Best for | Catching and diagnosing regressions | The search ranking signal | How real visitors experience your site |
| Latency | Immediate, on every run | Updates slowly | Near real time |
| Control | Full: device, network, location | None: Google's sample | Sampling and segmentation |
| Diagnostics | Full Lighthouse audits | Metric values only | Subparts and attribution |
| Needs traffic | No | Yes, eligibility-gated | Yes, plus the snippet |
Use Synthetic testing when#
- You want to catch a regression before it reaches users, with performance budgets and pull request reviews.
- You need to diagnose why a page is slow, using the Snapshot Overview, Long Task Timeline and third-party analysis.
- You need to test a page that has no traffic yet, such as a staging environment or a logged-in page.
Use CrUX when#
- You want to track the Core Web Vitals that feed Google's search ranking assessment.
- You want real-user data without installing anything, on any site eligible in the dataset.
- You want to benchmark competitors: CrUX is public for any eligible origin, so you can track sites you don't own, not just your own.
- You are comparing your origin or key pages against the field on the CrUX Dashboard.
Use Real User Monitoring when#
- You need a deeper, faster data set than CrUX: every qualifying session across all browsers, in near real time.
- You want to find the biggest, most widely-seen issues by sorting by traffic, then segment by page, device, country or browser to find who is affected.
- You want element attribution for Largest Contentful Paint, Cumulative Layout Shift and Interaction to Next Paint, plus metric subparts, from real sessions.
Use them together#
The three sources are strongest in combination. A typical flow:
- Catch a regression in Synthetic before it ships, and confirm the fix in a pull request review.
- Watch the change reach real users in RUM within hours.
- Confirm the longer-term trend in CrUX, where the search ranking signal lives.
Common scenarios#
Catching a regression before release#
Use Synthetic with performance budgets and pull request reviews so a regression is flagged before it merges.
Debugging a specific slow page#
Use a Synthetic Snapshot Overview for the full picture, the Long Task Timeline for main-thread blocking, and third-party analysis for external scripts.
Tracking Core Web Vitals for search#
Use the CrUX Dashboard for the 75th-percentile field data Google assesses, and the free Core Web Vitals test for a quick check.
Understanding a real incident or spike#
Use RUM to see the impact on real visitors as it happens, and segment to find which pages, devices or countries are affected.