Web performance is a business lever, not a vanity score

Engineering teams talk about performance in scores. The business experiences it as conversion, bounce, and search visibility. A web leader's job is to hold both frames at once, and to know when the work has reached the point of diminishing returns.

Translate every performance number into funnel language. Faster pages convert better, lose fewer visitors, and rank higher in search. A performance report that stops at the score is unfinished; the version that earns investment ends in business terms, because that is the language in which the work gets funded.

Chase impact, not a perfect score. Taking a slow site to healthy is a business project with measurable returns. Taking a healthy site to perfect is usually an engineering hobby. Know where the diminishing-returns line sits for your funnel, stop there deliberately, and say in public that you stopped on purpose. Restraint is a strategy signal, not a compromise.

Build performance into the components, not into the cleanup sprint. When performance budgets live inside the component library, every page composed from it inherits them, and speed stops depending on anyone remembering to check. The alternative is the periodic heroic cleanup, which is the most expensive way to buy back the same speed repeatedly.

Aim the work at the pages that earn revenue. Performance effort is not spread evenly; it is aimed. The pages where buyers decide deserve the budget first, because a regression there costs real conversions while the same regression on a low-stakes page costs almost nothing.

Hold the gain like a budget, not a milestone. A performance win announced once and never watched again will quietly erode. Regressions are cheapest the week they appear, so the win is only durable if something owns it: a budget, a check, a person.