Noise Canceling RUM
summary
Real User Monitoring contains measurable noise: on aggregated data and a 6M‑monthly‑visit site HVN reduces about 22% total noise down to 78% human visible navigations, cutting bot spikes (≈1%) and back‑forward noise (10–23%) that skew percentiles. Built rumarchive.com and R Insights to inspect real examples, used multiple percentiles and visibility/navigation filters, and see fewer fluctuations, clearer correlation with business metrics, and targeted wins where backend or cache changes actually moved user experience.
AI-written summary · about
chapters (17)
- 0:00Introduction
- 1:06Real user monitoring matters
- 3:13Single percentile analysis pitfalls
- 5:03CDN caching and percentiles
- 8:16Traffic patterns mistakes
- 11:12Filtering RUM data
- 14:00Human navigations explained
- 20:21Noise across page types
- 23:19Hidden page navigations impact
- 26:32Hidden pages skew metrics
- 30:07Visibility state noise
- 32:15Excluding bots from RUM
- 34:46Bot traffic patterns
- 36:51Combining noise filters
- 37:59Back-forward cache noise
- 41:56Summary
- 42:36Q&A
related talks
49mModern Metrics
53mReal User Performance Monitoring at Netflix scale
05mHow real is real-user measurement?
16mHome Brewing R.U.M - Analyzing application performance with real user monitoring
34mWhy Google’s CrUX Results Are Not Reproducible With Your Real-User Monitoring
02mMobile Real User Monitoring
19mSensory Friendly Monitoring: Keeping the Noise Down
40mTraffic Steering using Rum DNS @ LinkedIn