Choosing a stack for your first product
The stack matters far less than the team that maintains it. Here is how we actually decide.
Three metrics, a handful of fixes, and the ones that are rarely worth chasing.
LCP, CLS and INP describe how a page feels in the first few seconds. Most sites fail them for the same three reasons: an oversized hero image, fonts that swap late, and third-party scripts loaded in the head.
Fixing those three usually moves a site from red to green without touching the framework. Image formats and explicit dimensions solve most layout shift. Deferring analytics and chat widgets solves most interaction delay.
Beyond that, returns drop off quickly. We stop optimising when the numbers are green on real-user data, not lab data, and spend the remaining budget on content instead.
The stack matters far less than the team that maintains it. Here is how we actually decide.
The design fee is the small number. Content, migration and QA are where budgets go.