Skip to content
Tutorials Center
My learning

Topics

Lesson 28 of 36 ยท Lighthouse and Audits

Lighthouse Scores and Metrics

Understand the Lighthouse performance score: the weighted metrics (FCP, LCP, TBT, CLS, Speed Index), how raw values become 0-100, and which metric to attack first.

The number in the Lighthouse circle looks like a grade, but it is a calculation with clear parts. If you understand them, you know where to spend your effort: a score is the sum of several metric scores, each with a weight, and the weights are far from equal.

This lesson explains the metrics, the weights, how a measurement turns into a score, and why lab metrics differ from the field metrics you saw in the Core Web Vitals lesson.

The five performance metrics

In current Lighthouse versions (12 and later) the performance score combines five metrics:

Metric What it measures Weight
First Contentful Paint (FCP) When the first text or image appears 10%
Speed Index (SI) How quickly the visible part of the page fills in 10%
Largest Contentful Paint (LCP) When the largest element is painted 25%
Total Blocking Time (TBT) Sum of main-thread blocking beyond 50 ms between FCP and interactivity 30%
Cumulative Layout Shift (CLS) Visual stability 25%

Weights have changed between releases (older versions included Time to Interactive and used other weights), so always check the version in the report footer before comparing numbers with an old article. The important lesson: TBT, LCP and CLS together make up 80% of the score.

Lighthouse is a lab tool: it runs on one simulated device and cannot click, so it has no INP. It uses TBT as a stand-in for responsiveness, because pages with a low TBT tend to have a good INP.

From a measurement to a metric score

Each raw value, such as an LCP of 3,100 ms, is mapped to a score from 0 to 1 using a log-normal curve fitted to real-world data from HTTP Archive. The curve has two control points: the value that earns a score of 0.9 (green) and the median that earns 0.5. Being far from the target hurts a lot; improving from "already good" to "slightly better" adds little, and you can never get more than 1.

The stand-in below uses a simplified piecewise-linear version of this mapping, with real-looking control points for mobile, to show how a score becomes the weighted total. The numbers are illustrative, not Lighthouse's exact curve.

JS Stand-in: weighted performance score (simplified)

Press Run (or Ctrl + Enter) to see the console output here.

No output.

Run it and note that TBT (450 ms) scores only 0.375 but carries 30% weight, so it removes about 19 points all by itself. That is why fixing long tasks often beats shaving 100 ms off FCP.

What-if analysis: which metric is worth fixing?

The next stand-in reuses the same idea to answer the practical question "if I improved this one metric to green, how many points would I gain?" Because each weighted contribution is limited to its weight, the best target is the metric with the largest lost points.

JS Stand-in: rank metrics by points you could win

Press Run (or Ctrl + Enter) to see the console output here.

No output.

The table shows TBT first (19 points available), then LCP (15). CLS is already perfect, so working on it would be wasted effort for the score, although in reality you would still keep it from regressing.

Reading the audits behind the numbers

Under the metrics, each audit relates to a metric through small metric filter chips (FCP, LCP, TBT, CLS). Click a chip to show only the audits that affect that metric. This turns the report into a to-do list per metric:

  • LCP: "Preload Largest Contentful Paint image", "Serve images in next-gen formats", "Reduce server response times (TTFB)", "Eliminate render-blocking resources".
  • TBT: "Reduce JavaScript execution time", "Minimize main-thread work", "Reduce unused JavaScript", "Avoid long main-thread tasks".
  • CLS: "Image elements do not have explicit width and height", "Avoid large layout shifts".
  • FCP: "Eliminate render-blocking resources", "Enable text compression", "Preconnect to required origins".

Estimated savings are independent guesses and do not add up: fixing two opportunities that overlap will not save the sum of both numbers.

Lab versus field

A Lighthouse score describes one simulated load. It does not say how real visitors on many devices and networks experience the site. For that you need field data: the Chrome UX Report, your own real-user monitoring (RUM) with the web-vitals library, or the field section at the top of PageSpeed Insights. Use Lighthouse to find and fix causes, then confirm the result with field data after a few weeks, because field data is a 28-day rolling window.

Also keep in mind that the score is not the goal. A site can reach 100 in the lab by hiding content from the test, and still feel slow. Optimize for users, use the score as a regression alarm.

Check your understanding

Quick check ยท 4 questions

of 4 answered You scored / 4 ยท perfect!

  1. 1. Which metric has the highest weight in the Lighthouse performance score (current versions)?

    TBT is weighted at 30%. INP is not measured by Lighthouse's navigation mode at all.
  2. 2. Why does Lighthouse use Total Blocking Time instead of INP?

    A lab load has no real clicks; TBT is a proxy for the main-thread blocking that causes poor INP.
  3. 3. A metric with a low score and a large weight is:

    Points lost are roughly (1 - score) x weight, so low score plus high weight means the biggest potential gain.
  4. 4. Estimated savings of two opportunities are 400 ms and 300 ms. Total gain?

    Each estimate is computed on its own; overlapping fixes do not add up linearly.

Summary

  • The performance score is a weighted mix of FCP, Speed Index, LCP, TBT and CLS; TBT, LCP and CLS dominate.
  • Raw values map to 0-1 scores through log-normal curves; far-from-target values cost the most.
  • Filter audits by metric chip to build a to-do list, but do not add up estimated savings.
  • Lighthouse is lab data; verify improvements with field data (CrUX, RUM).
  • The score is a tool, not the goal.

Next: Accessibility, SEO and Best Practices Audits