15
2 Comments

How Lighthouse Performance Scores Are Recorded and Calculated

What a Lighthouse Performance score actually is: HTTP Archive curves, Lighthouse 10 weights, and why lab greens are not CrUX field status.

You paste a client URL into PageSpeed Insights and the green circle pops up. The account manager forwards it as "Google speed." That number is a lab Performance score from one controlled run. It is not a ranking factor by itself, and it is not a Chrome User Experience Report percentile.

Lighthouse records metrics under a throttled profile (mobile or desktop), maps each raw value through an HTTP Archive scoring curve, then blends those metric scores with documented weights. Under Lighthouse 10 the blend is Total Blocking Time 30%, Largest Contentful Paint and Cumulative Layout Shift 25% each, First Contentful Paint and Speed Index 10% each. Opportunities and Diagnostics do not add points directly; they explain what might move the underlying metrics.

What agencies keep mixing up on the same PSI screen:

  • Lab Performance is synthetic, versioned, and fast after a deploy

  • Field Core Web Vitals come from CrUX over a rolling window, with INP instead of TBT

  • Colour bands (0–49 / 50–89 / 90–100) describe that lab run, not every visitor on every network

  • A five-point swing across a Lighthouse major can be scoring math, not a regression

How we report the number without score-chasing:

  • Name the artefact: Lighthouse / PSI lab Performance, plus device and Lighthouse major when known

  • Show the drivers: LCP, TBT, and CLS usually explain most of the blend

  • Keep CrUX / Search Console on a separate slide

  • Prefer three or more scheduled runs over one heroic paste after a tag change

  • Open the Lighthouse scoring calculator before promising "ten points is one easy fix"

Apogee Watcher stores those lab scores on a schedule across listed URLs so you compare like with like. Pair the trend with metric budgets on money pages. Layer field status separately when samples exist.

Read more: how Lighthouse performance scores are recorded and calculated

posted toAvatar for product Apogee Watcher
Apogee Watcher
  1. 2
    For an authenticated SaaS app, the public landing page is easy to put through PSI, but most of the actual work happens after sign-in. Do you run scheduled Lighthouse checks against signed-in routes too, or rely on RUM there? I'd be wary of reporting a healthy public-page score as if it covered the product people use every day.
    1. 1
      This is a valid concern. A healthy public-page Performance score does not cover the signed-in product. We schedule Lighthouse / PageSpeed Insights on landing, pricing, and signup URLs, plus any app routes that load without a session. RUM is still needed for other use cases. See https://apogeewatcher.com/blog/when-to-use-synthetic-vs-real-user-monitoring-performance