What’s the difference between a performance budget and a performance SLO?

A budget is a hard limit you enforce in CI (bytes, LCP/INP thresholds in a lab). An SLO is a production commitment measured with RUM (p75 Web Vitals per route/device). Budgets prevent obvious regressions before deploy; SLOs validate real-world experience and govern release velocity via error budgets.

Why p75 instead of p95?

p75 balances experience and practicality. It reflects what most users feel without letting tail events dominate. Track p95 to understand tail pain, but budget and gate releases on p75 so you can actually ship. If you’re a bank or trading platform, you may choose p90/p95 for critical flows—just be ready for slower velocity.

Can I trust Lighthouse in CI to represent real users?

Use Lighthouse CI for guardrails, not truth. It’s deterministic and good at catching obvious regressions. Pair it with production RUM SLOs and canary gating. When they disagree, production wins—always.

How do I handle third-party scripts within budgets?

Give each third party a line item (bytes, CPU cost) and load strategy (defer, async, after interaction). Lazy-load personalization and chat. If a vendor can’t meet the budget, negotiate, sandbox (iframe/worker), or remove it during peak using feature flags.

Performance-optimization · Oct 8, 2025 · 10 minute read

Stop Chasing 100 Lighthouse: Design Performance Budgets That Keep UX Consistent

Speed isn't a trophy—it's a contract with your users. Here's how to set, enforce, and hit performance budgets that protect customer experience and revenue.

Back to all posts

Stop Chasing 100 Lighthouse: Design Performance Budgets That Keep UX Consistent

Key takeaways

Implementation checklist