Lighthouse 96, not vanity: how we keep clients fast in year two
Most agency-built websites are fast on launch day. The score is in the pitch deck. Then the client adds a cookie banner plugin, a chat widget, three analytics tags, and a hero video. By year two the Lighthouse score is 54 and the agency is long gone.
We keep clients fast because we treat performance as an ongoing commitment, not a launch metric.
What we enforce in CI
Every client site has a Lighthouse budget in the deployment pipeline. If a change drops the performance score below the agreed threshold, the deployment fails. Not a warning — a failure. This has stopped four sites from going live with an accidental 40-point regression.
The budgets we typically enforce: performance score 90 minimum; total page weight under 500kb on the homepage; third-party scripts reviewed and approved before they go in; images in WebP or AVIF, compressed, lazy-loaded below the fold.
The third-party script problem
Third-party scripts are the single biggest cause of performance regression on client sites. Every marketing tool, chat widget, and heatmap tool ships JavaScript that runs on your page. Our standard: every third-party script needs a business case and a performance review before we add it. We have turned down three chat widgets in the last year because they added 400ms to Time to Interactive on mobile.
What this costs in real terms
Slower sites lose conversions. Google puts it at roughly 7% conversion loss per one-second delay on mobile. For a client doing £30,000 a month through their website, a three-second slowdown is a £63,000 annual problem.
The performance budget is not pedantry. It is protection.
If your site was fast when it launched and is slow now, talk to us. We will audit it and tell you exactly what is dragging it down.
Leave a comment