Deleting nine apps beat rebuilding the store
A Shopify storefront carrying fourteen apps, losing mobile buyers before the page finished painting.
- Mobile LCP
- 6.4s2.0s-69%
- CLS
- 0.280.03-89%
- Conversion
- 1.0x1.28x+28%
CrUX field data, 28 day rolling window, 60 days after launch. Source: Search Console and Shopify analytics, March 2025
Context
A direct to consumer brand doing most of its volume on mobile, with a theme that had been extended by three agencies over four years.
The constraint
The thing that ruled out the obvious answer.
Peak season was five weeks out. Anything that risked the checkout flow was off the table, and the team had no developer of their own to maintain a custom build afterwards.
Diagnosis
Of fourteen installed apps, nine were injecting render blocking script on every page. Four of those nine were powering features the business had stopped using, and two were duplicates of each other. The theme was not the problem. The accumulated layer on top of it was.
What we rejected, and what we chose instead
Any firm can describe what it built. The useful record is the option that was on the table and did not survive.
Rejected
Headless rebuild on a custom frontend
Twelve weeks minimum, past peak season, and it would have left a business with no developers maintaining a custom codebase. The right answer for a different company.
Chosen
Remove nine apps, defer the rest, keep the theme
Five weeks, no checkout risk, and the client's team can still edit the store themselves. The smallest change that moved the metric.
What was actually implemented
- 01Audited every app against actual usage in the previous 90 days of analytics
- 02Removed nine apps, replaced two with theme level markup
- 03Deferred the remaining third party scripts behind interaction and idle triggers
- 04Responsive image pipeline with AVIF and correct intrinsic sizing to kill layout shift
- 05Lighthouse CI budget on the pull request so a future app install fails the check
Measured, not estimated
Mobile LCP fell from 6.4s to 2.0s and layout shift effectively disappeared. Mobile conversion rose 28% across the following peak, against a comparable prior period.
CrUX field data, 28 day rolling window, 60 days after launch. Source: Search Console and Shopify analytics, March 2025
Performance work produced results faster than a rebuild would have.
More from website engineering
Website engineering
A slow site is a revenue problem with a technical cause. We find the cause, fix it, and leave you with the measurement so it does not come back.