67% of the stores we tested fail LCP.
We ran the first stores through the speed test and read Lighthouse’s third-party summary on every one. Third-party scripts account for 61% of median render delay — and the worst offenders are the ones merchants pay the most for. We are not comparing themes yet; that needs a detector we have not built.
What we measured
Swedish storefronts, three PageSpeed Insights runs each on an emulated Moto G Power over a throttled 4G connection, cold cache, median taken. Home page and one product page. No store was contacted and none of this comes from a client engagement — everything here is public data anyone can re-run.
Sixteen of the twenty-four fail the Largest Contentful Paint threshold on mobile. The median store ships 600 kb of JavaScript and loads fourteen third parties. The lightest ships 218 kb and loads five.
Nobody near the top of the board got there with thirteen third parties on the page.
Where the delay actually comes from
Lighthouse names every third party it finds and attributes main-thread time to each one, so this is not inference. Summed across the board, third-party scripts account for 61% of median render delay — more than images, theme CSS and server response time combined.
| What costs the time | Median | Owner |
|---|---|---|
| Third-party app scripts | 1.9 s | Apps |
| Unoptimised hero imagery | 0.8 s | Build |
| Render-blocking theme CSS | 0.4 s | Build |
| Server response time | 0.2 s | Platform |
The expensive ones are the paid ones
The pattern that surprised us: cost and weight track together. The heaviest third party on the median storefront is a review widget, and it is also usually the most expensive line on the app bill. Free scripts — a pixel, a consent banner — are cheap in both senses.
- 01ReviewsHeaviest single third party on more than half the board. Almost always loaded store-wide for a widget that appears on one template.
- 02Personalisation and recommendationsSecond heaviest, and the hardest to attribute to revenue because it rewrites the page it is measured on.
- 03ChatSmall on paper, expensive in practice: it usually loads a font and an iframe of its own.
What we are not claiming
We are not comparing themes. Telling Dawn from a fork from a page-builder theme reliably needs a detector Lighthouse does not give you, and guessing from markup would produce a ranking that looks authoritative and is not. That is the next thing we build, and the benchmark will say so when it lands.