Web Performance1 min read
Measuring INP With Real User Monitoring — Beyond Lab Lighthouse Runs
By Priyatham Rama Sai
Lab INP stabilizes poorly when third-party scripts differ by geography. Sampling interaction latency in RUM exposed an analytics chunk blocking main thread only on Android mid-tier devices.
Field truth
Aggregate INP by device class and geography. Marketing pages often regress from tag managers added without performance review.
Attribution
Split long tasks by script URL where possible so owners exist. Without attribution, frontend teams argue while ads teams ship another pixel.
Budgets
Pair INP budgets with ownership in CI — bundle analyzers catch regressions early; RUM validates whether users feel the fix.