Should budgets target Lighthouse scores or Core Web Vitals?

Target Core Web Vitals (LCP, INP, CLS) at p75/p95 in production RUM. Use Lighthouse/TBT as lab proxies and CI guardrails, not the source of truth.

How strict should budgets be on low-end devices?

Set budgets per device class. Don’t force low-end targets to match iPhone 15 Pro. Define acceptable ranges (e.g., p75 LCP ≤ 3.0s on low-end Android) and improve through JS dieting and image strategy.

What about Single Page Apps and hydration?

Budget INP aggressively. Defer non-critical hydration, split routes/components, and move work server-side (RSC/SSR). Track long tasks and keep TBT ≤ 200ms in lab. Prioritize the LCP element and avoid lazy-loading it.

How do we handle third-party scripts politically?

Make them a first-class budget line item with an owner and an expiry date. Load after interaction or asynchronously, isolate via web workers if possible, and remove if they don’t demonstrably lift a KPI.

Performance-optimization · Oct 13, 2025 · 9 minute read

The Quiet Outage: How Performance Budgets Keep Your UX (and Revenue) From Flapping

Most teams don’t notice the “brownouts” that kill conversions. Performance budgets make UX reliable, measurable, and enforceable—without cargo-culting Lighthouse scores.

Alex R. McKinley

Principal Engineer, GitPlumbers

20 years scaling web platforms at Shopify, the BBC, and two unicorns you’ve definitely rage-refreshed. I turn flaky UX into boringly fast systems.

Performance isn’t a feature, it’s availability for humans. Budgets make it measurable and enforceable.

Back to all posts

The Quiet Outage: How Performance Budgets Keep Your UX (and Revenue) From Flapping

Key takeaways

Implementation checklist