Web performance guides

Compare Two Saved Lighthouse Runs

·

Open saved Lighthouse JSON before comparing a claimed performance improvement. Keep the device and throttling information alongside each score so a change in the test profile is visible.

Page Speed Checker

Compare two mobile runs of the same page

Load example and select Review page speed. Both saved runs use Lighthouse 13, a mobile device and simulated throttling with the same recorded timings. The Performance score goes from 68 to 91. The comparison reports a 23-point gain, LCP decreases by 1,300 ms and TBT decreases by 190 ms.

Download page-speed-report.json and check labComparison. The scoreDelta value is 23 and metrics includes LCP with value −1300 and unit ms. The primary and comparison sections still contain each measurement. A recorded version or device mismatch leaves the individual results visible and removes the deltas.

Chrome Lighthouse result formatChrome Lighthouse performance scoring

Choose one savings estimate to investigate

In the primary example, the render-blocking insight estimates 620 ms for LCP and 460 ms for FCP. The unused JavaScript opportunity has a separate 250 ms estimate and 42,000 bytes. Start with the metric you want to improve, inspect the underlying request, and save another run after changing it.

The reports can count overlapping work. Keep the measured LCP change in your result and use the savings list to decide what to test next.

Chrome Lighthouse result formatGoogle PageSpeed Insights response fields

Save and check a report from your own page

Export JSON from the Lighthouse report, or save a PageSpeed Insights API response from an audit you already ran. Open it as the primary report. When comparing a later run, keep the URL, device, throttling and Lighthouse version comparable. Repeat measurements and account for lab variability before you draw a conclusion.

Chrome Lighthouse result formatChrome Lighthouse performance scoringGoogle PageSpeed Insights response fields

Common questions

What should I do when the comparison reports different settings?

Review both reports independently, then repeat the test using the same Lighthouse version, device and throttling profile.

Does a better lab score establish that visitors had a faster experience?

A lab score describes the saved synthetic run. Use a field record, such as available CrUX readings for the relevant page and device, to review visitor experience.

Sources

Put it into practice.

Related guides