Start a project
Menu

Engineering

Core Web Vitals budgets that survive real content

Why speed scores slip after launch, and how page weight budgets, image rules and checks in every build keep a site fast as content grows.

By Aumair Malik7 min read

Published: Last updated:

Most sites launch fast and get slower every month. Not because of one bad change, but because nothing stops a hundred small ones.

Key takeaways

  • A budget only works when a failing build enforces it.
  • Judge speed by field data from real phones, not by a lab score on a laptop.
  • Fix image sizes in the components, not in editors' habits.
  • Load chat, analytics and heavy visuals after the page is usable.

Know the numbers that count

Google's Core Web Vitals (opens in a new tab) measure loading, responsiveness and visual stability. A page passes when 75% of its visits meet each threshold.

Key figures

  • Metrics

    3

    Loading, responsiveness and visual stability

  • Percentile

    75th

    Three in four visits must meet each threshold

  • Window

    28days

    Of real visits behind each field reading

Field data as the Chrome UX Report and PageSpeed Insights report it.

Core Web Vitals thresholds
MetricGoodNeeds improvementPoor
Largest Contentful Paint≤ 2.5 s≤ 4 s> 4 s
Interaction to Next Paint≤ 200 ms≤ 500 ms> 500 ms
Cumulative Layout Shift≤ 0.1≤ 0.25> 0.25

Set budgets the build can enforce

Agree limits for page weight, JavaScript and requests, then fail the build when a change crosses them. A number in a document is a wish; a failing check is a rule.

Set your first budget

5 steps

  1. Measure what visitors get today

    Pull field data for your busiest pages: the 75th percentile, on phones.

  2. Pick the metric that hurts most

    Usually the largest paint on mobile. One metric you enforce beats five you watch.

  3. Set the limit just under today

    A budget that fails from day one gets switched off. Tighten it once the team trusts it.

  4. Fail the build when a page goes over

    Run the check on every pull request, so the change that breaks it is the one that fixes it.

  5. Review it every quarter

    New features earn room; dead code gives it back.

JSON

[
  {
    "path": "/*",
    "resourceSizes": [
      { "resourceType": "script", "budget": 250 },
      { "resourceType": "image", "budget": 300 },
      { "resourceType": "total", "budget": 800 }
    ]
  }
]
CodeA Lighthouse budget file: sizes in kilobytes for every page.

Tip: Start with one page type. Budget your home page and one article first. Templates share code, so the rest of the site usually follows.

Weighing it up

What a build budget buys you

  • Pro: Regressions are caught before they ship
  • Pro: Speed becomes a number the whole team can see
  • Pro: Nothing to remember once it runs

What it costs

  • Con: A limit set too tight blocks honest work
  • Con: Lab numbers never match the field exactly
  • Con: Someone has to own raising it

Treat images as code

Fix aspect ratios, sizes and formats in the components that render images, so an editor uploading a large photo cannot slow the page down.

A 3D render of stacked interface layers on a dark background
Fig. 1Images get their size, format and loading from the component that draws them, never from the upload.

Heads-up: Watch the largest image. The hero image is often the largest paint. Never lazy-load it and never fade it in: both delay LCP.

Load the extras last

  • Chat widgets on the first interaction
  • Analytics after consent
  • Heavy visuals once they are near the screen

A number in a document is a wish. A failing check is a rule.

Setting a performance budget, step by step4:12Video coming soon
VideoPlays from YouTube once you press play. YouTube may then set cookies.

Measure on real phones

Lab scores on a fast laptop flatter everyone. Field data from real visitors on mid-range phones shows what customers actually feel.

Where to find field data

  1. Open the Chrome UX Report (opens in a new tab) for your domain
  2. Filter to phones and the 75th percentile
  3. Compare it with your lab budget each month

Note: Field data needs traffic. A new or quiet page shows no Chrome UX Report data until enough visitors have loaded it.

Why not test in the lab alone?

Lab tests run on one device and one network, so they are repeatable and good at catching regressions. Real visitors bring slower phones, busy networks and full caches, which no lab run reproduces.


Questions

How strict should the first budget be?

Just under what the page weighs today. A budget that fails from day one gets switched off; tighten it once the team trusts the check.

Do budgets slow development down?

Rarely. Most changes pass without anyone noticing, and the few that fail are the ones worth a second look.

What about pages editors build in a CMS?

Put the rules in the components editors use: image sizes, embeds that load on click and limits on what each block can hold.

Performance budget checklist

The metrics, limits and checks from this article on one page.

PDF · 1.2 MB

Coming soon

Want budgets your build enforces?

Tell us about your site. A senior engineer replies within one business day.

Talk to an engineer

Sources & official references

Thresholds and tools change. Check the official pages above before you set your budgets.

Start a project

Got an idea worth building?

Share a few lines about what you have in mind and we'll reply with next steps.