Case study

A multi-region DTC fashion brand

Shopify Plus · US / UK / EU storefronts · client for 6+ years · anonymized

This is the store behind the numbers on the homepage. A multi-region fashion brand selling direct in the US, UK and EU on Shopify Plus. They haven't agreed to be named, so I won't name them; the numbers below are theirs, and every caveat that belongs next to a number is next to it. Two pieces of work, measured two different ways.

Story one: the email popup that weighed 418 KB

Their email-signup popup came from a third-party onsite app, and that app shipped 417.9 KB of code to every visitor, whether or not the popup ever appeared. I rebuilt it as a 2.8 KB native component: same signup flow, same list, same behaviour for the shopper.

418 KB 2.8 KB

About 418 KB and 70 to 80 ms of main-thread work removed from every mobile pageview.

104 82 ms

Field INP, about 21% better, with no LCP or CLS regression.

This one was a controlled before/after: the only change between the two measurements was the popup, so the improvement is attributable to it. That's also why it's the result I lead with. Real-user responsiveness (INP) moved from 104 ms to 82 ms, and nothing else got worse.

Story two: putting the hero image on the critical path

Their old theme lazy-loaded the hero image, the single biggest thing on the page, so the browser discovered it late on every load. As part of a theme rebuild I put it on the critical path. Mobile LCP improved 28 to 33% across home, collection and product pages.

PageLCP beforeLCP after
Home1,525 ms1,065 ms−30%
Collection1,831 ms1,311 ms−28%
Product1,720 ms1,147 ms−33%

Median of 5 throttled runs per page: 4x CPU slowdown, Slow 4G. CLS was 0.00 on every run, before and after.

The asterisks

You should have these before you weigh the numbers. The LCP figures are lab measurements on a throttled profile (a mid-range phone on a weak connection), not field data. The popup result is the controlled A/B; the theme numbers are not. They compare two theme versions of near-identical stores, roughly 100% the same code and about 95% the same apps, so read them as a same-store upgrade comparison rather than a controlled experiment. I'd rather you trust the numbers than be wowed by them.

One more number that isn't a benchmark: this client has kept me around for more than six years. Speed work is easy to sell once. The retention is the part I'm proudest of.

If you want to know what the same tests say about your store, the teardown is free: I measure your live storefront the way Google does and send you a short, honest report. If there's nothing worth doing, I'll say so.

Get a free teardown →

Alex · alex@storespeedup.com · storespeedup.com