Why Google’s CrUX Results Are Not Reproducible With Your Real-User Monitoring
summary
Claims that CrUX and in-house RUM both reflect reality but measure different samples: CrUX reports monthly rolling 28‑day aggregates from opt-in Chrome users (logged in with history sync), so field numbers lag and exclude many sessions. RUM pipeline is implemented with Flink, Elasticsearch and per‑minute pre-aggregations, enables A/B by toggling Speed Kit for ~50% of users, and produces real‑time dashboards. Practical thresholds cited: LCP <2.5s is good and FID <100ms is good; mismatch between CrUX and RUM is expected, not a bug.
AI-written summary · about
chapters (9)
- 0:00Start
- 3:59What Are the Most Important Kpis To Look at
- 6:37Input Delay
- 7:04Cumulative Layout Shift
- 10:16The Chrome User Experience Report
- 17:04The Data Pipeline
- 18:39Low Latency Data
- 27:04Fixed Temporal Granularity
- 31:57The Speed Guide and the Speed Advisor
