Performance budgets that survive the roadmap
Small measurable budgets work when they belong to product decisions and are enforced close to everyday changes.

A performance budget is a promise about the experience, not a score kept by the engineering team. Budgets fail when they live in an audit document, measure only a perfect lab run, or have no owner when a release crosses the line.
Budget what people feel
Choose a few measures tied to a key path. Include page weight because it is easy to enforce before deployment, and field metrics because real devices and networks reveal costs the lab cannot.
| Measure | Starter budget | Where to enforce |
|---|---|---|
| Initial JavaScript | 80 KB compressed | Build report |
| Largest Contentful Paint | 2.5 seconds p75 | Field monitoring |
| Interaction to Next Paint | 200 ms p75 | Field monitoring |
| Layout shift | 0.1 p75 | CI and field data |
Put the check near the change
Static assets can be checked during a build. Pull requests can show a size delta. Image dimensions and formats can be validated in the content pipeline. The closer feedback is to the authoring moment, the less often a team needs an emergency optimization sprint.
Define the exception
Occasionally a larger asset or dependency earns its cost. An exception should name the user benefit, the measured impact, and an expiry or follow-up. This turns “we will fix it later” into a visible decision that can be reviewed.
The goal is not a permanently tiny website. It is a team that notices the cost of change and chooses it deliberately.
A budget survives when it belongs to a product decision, not an audit document.
A useful note should change the next decision.