Real user monitoring

Real User Monitoring (RUM) collects performance and error data from actual sessions—devices, networks, geographies, releases—and reports distributions (p75/p95), not single lab scores.

| Signal | Why RUM matters | |--------|-----------------| | LCP/INP/CLS field | What users truly experience | | Route / view | Which pages regress after deploy | | Release version | Tie metrics to build SHA | | Device class | Mobile vs desktop gap |

					// Concept: beacon on navigation
navigator.sendBeacon('/rum', JSON.stringify({ route, vitals, release }));
				

Use CrUX for public origin data; add first-party or vendor RUM (Sentry, Datadog RUM, web-vitals library) for granular routes. Sample high-traffic sites to control cost.

On interviews: RUM vs synthetic; privacy and PII in beacons; sampling strategies; how to debug a p75 LCP regression with cohort filters.

Common pitfalls: averaging global metrics; ignoring long-tail countries; no release dimension; RUM script hurting INP.

The trade-off is fidelity to real users versus instrumentation cost and privacy constraints.

Checklist:

• Track vitals by route and release.
• Report percentiles, not only means.
• Redact PII in client beacons.
• Compare cohorts after deploy.