What the performance audit measures
A single overall speed score is close to useless for planning work — it tells you there is a problem without telling you which of the eleven things on the page caused it. This audit measures per page, per metric, and attributes each result to the specific resources responsible.
- Largest Contentful Paint (LCP) under 2.5s — When the main content of the page becomes visible on a throttled mobile profile, and which element is the LCP candidate on each page.
- Interaction to Next Paint (INP) under 200ms — How long the page takes to respond to real interaction, and which scripts occupy the main thread when it does not.
- Cumulative Layout Shift (CLS) under 0.1 — How much the layout moves during load, and which unsized images, injected banners or late-loading fonts cause the movement.
- Images sized and optimised — File sizes against rendered dimensions, missing width and height attributes, absent lazy loading, and images still served in formats where a modern one would save materially.
- Third-party scripts justify their cost — Every third-party request inventoried with its transfer size and blocking behaviour, so tag bloat becomes a list of named costs rather than a vague suspicion.
- Viewport meta present and sane —
width=device-width, initial-scale=1, with no fixed width and no restriction on zoom.
How measurements are taken
Pages are loaded in a controlled environment on both a desktop and a throttled mobile profile, several times, and the median run is reported. Repeat runs matter: single measurements of a page with third-party content vary enough to produce contradictory conclusions.
Results are reported per URL, and where several pages share an LCP element the report says so, because one hero definition usually explains all of them. Where a metric fails, the report names the element or request responsible and estimates the improvement available from fixing it, so you can see which two changes recover most of the loss.
These are lab measurements taken by us, not field data from your real users. If you share GA4 or Search Console access we will read your Core Web Vitals field data alongside them, and where the two disagree the report says so rather than picking whichever number looks better.
Findings this audit commonly surfaces
What is out of scope
The audit measures and attributes; it does not tune your infrastructure. We do not configure CDNs, rewrite build pipelines or refactor application code as part of the engagement, though we will happily sit with the engineers who do. Server-side profiling and database work are outside the audit boundary — what we can tell you is when time-to-first-byte, rather than the front end, is the thing holding your pages back.
Related categories
Performance work usually pays off twice. Page speed is a ranking input, so the SEO audit reads these results too, and layout instability is an accessibility problem, which is why CLS findings also appear in the accessibility audit. Scoring bands for every category are documented in the methodology.