Core Web Vitals are an architecture problem
You cannot fix LCP with a plugin. Rendering strategy, not micro-optimisation, is what moves the numbers.
Most Core Web Vitals work we inherit is micro-optimisation: defer this, lazy-load that, compress the other. It plateaus quickly.
The wins that hold come from rendering strategy. Which routes are static, which stream, where the data boundary sits, and what genuinely needs to be client-side at all.
On a recent retail engagement we moved the product grid from client-fetched to streamed server components. LCP went from 4.1s to 1.2s and stayed there through three subsequent feature releases, because the improvement was structural rather than a set of patches waiting to regress.
Measure against revenue, not lab scores. A green Lighthouse run on a page nobody buys from is a vanity metric.
An engineer replies within one working day — not a sales sequence.
