Lighthouse, Rendering and Coverage: Reading Audits Correctly
In Lighthouse 13 the audits merged with Performance Insights, but the score is calculated only from metrics: fixing cache-insight left the score unchanged, and three runs on the same page varied between 89 and 91. Simulated vs observed LCP, lab vs field data, navigation/timespan/snapshot modes, Rendering emulations, and using a source map to show that only 2.6 KB of the 42 percent in Coverage is our own code.
8 Nov 2026
·
9 min
·
5
