A performance budget is a project decision about the resources or measurements a team wants to keep within agreed limits. The right budget depends on the page, its audience, supported devices, network conditions, and the features it needs. Treat the budget as a tool for noticing changes, not as a universal definition of a fast website.
Start with a measured baseline
Choose representative pages and measure them with the same tools and conditions before making changes. Record the page, device or test environment, network assumptions, and date alongside the result. A lab run can help compare builds under controlled conditions; field data shows how real visitors experience the site across different devices and networks.
Separate local budgets from Core Web Vitals
Teams can set their own limits for items such as transferred image bytes, total page weight, or JavaScript. These are local engineering choices: explain why each limit exists and revise it when the page or audience changes.
Google's current Core Web Vitals guidance describes three user-experience metrics: Largest Contentful Paint (LCP) for loading, Interaction to Next Paint (INP) for interactivity, and Cumulative Layout Shift (CLS) for visual stability. The guidance recommends evaluating the 75th percentile separately for mobile and desktop. Its current “good” thresholds are LCP at or below 2.5 seconds, INP at or below 200 milliseconds, and CLS at or below 0.1. These are recommended field thresholds, not guaranteed outcomes for an individual visit.
Turn a budget into a review habit
For each chosen limit, name the page or asset it applies to, how you will measure it, and who will review a regression. Keep the first limits few enough to check during normal work. When a build crosses a limit, compare the affected page and recent changes before deciding whether to optimize, adjust the budget, or accept a documented trade-off.
Repeat measurements after material layout, content, hosting, or media changes. A passing lab check alone does not establish how visitors experience the released site; use field measurements when available and interpret them with their sample and time period in mind.