Page speed: the ten cheap wins before a re-platform
You can usually halve LCP without touching the framework.

Re-platforming for performance is the most expensive way to fix a speed problem. Start cheap. The ten interventions below routinely deliver 40–60% LCP improvement in a single sprint — often enough to defer a full re-platform by twelve months and put the budget where it actually matters.
The ten cheap wins
- Image CDN with on-the-fly resizing — Cloudinary, Imgix or a Cloudflare Worker.
- Preconnect to font origins and self-host critical fonts as woff2.
- Defer all third-party tags below the fold, especially chat and review widgets.
- Add fetchpriority='high' to the LCP hero image.
- Enable server compression — gzip minimum, brotli ideally.
- Audit and remove redirect chains; aim for zero redirects to the homepage.
- Route-level code splitting — ship one chunk per page, not one bundle for all.
- Convert hero images to WebP or AVIF with a JPEG fallback.
- Lazy-load the chat widget — most chats can wait until the user scrolls.
- DNS prefetch for any third-party domain referenced after the fold.
The win
Typically 40–60% LCP improvement in a single sprint. Often enough to defer a full re-platform by twelve months. None of these require a framework change; most can be shipped in a single PR.
How to prioritise
Run a CrUX-based audit first to see which intervention will move the needle on field data, not lab data. Lab improvements that don't show up in CrUX after 28 days were probably solving the wrong problem.
