**Parent Topic**: [[Software/README]] ## By Time: Easy but Flawed The simple way to link site health to business outcomes is to plot them on one graph by time — if delay spikes and conversion drops together, investigate. But three flaws limit it: it relies on **averages** (no single-visitor-to-single-outcome link); it's **not well controlled** (a marketing campaign can drive traffic up, slow the site, *and* raise conversions, faking an inverse correlation); and the **performance measurements often aren't from the users actually on the site** (a synthetic "slow from California" result is irrelevant if no Californians are browsing). ## By Visit: Harder but Precise The better way is to record each visitor's session performance with a RUM tool and get latency into the analytics package — JavaScript marking a visit "fast" or "slow," or merging server-side RUM logs with analytics logs via cookies/usernames. "This is a lot of work" (often needing data-warehouse investment). ## The Pragmatic Bridge Because RUM data (segmented by city/carrier/URL) and analytics data (funnels/sessions) store differently, the best lightweight method is to **mark visits by Apdex performance level** (satisfied/tolerating/frustrated) and segment goal funnels along that dimension — "creating a new analytics segment devoted to performance." This is what turns [[Faster Sites Make More Money]] from a claim into a measured number; it uses the [[Apdex Score]]. --- *Source: [[Web Operations]] (Allspaw & Robbins, O'Reilly 2010) — Ch 11 — How Your Visitors Feel: User-Facing Metrics*