Search across all documentation pages
Bundle size, time-to-interactive (TTI), and frame-rate thresholds with named owners for Expo SDK 57 apps on React Native 0.86. Performance budgets turn subjective "feels slow" into merge and release gates. Assign owners - anonymous budgets are ignored.
| Metric | Threshold (starter) | Owner role |
|---|---|---|
| Main JS bundle (Hermes) | ≤ 4 MB gzipped | Mobile platform lead |
| Total download (OTA update) | ≤ 8 MB | Release engineer |
| Cold start TTI (primary tab) | ≤ 2.5 s on reference Android | Feature squad lead |
| Scroll FPS (primary list) | ≥ 55 fps sustained | UI performance champ |
| Memory after 10 min session | ≤ 350 MB on reference Android | Mobile on-call |
Adjust numbers per app category - document your thresholds in docs/performance-budgets.md and link from README.
Publish written budgets before feature freeze: Bundle MB, TTI seconds, FPS floors, and memory ceilings - starter table above is a template, not a universal law.
Pick reference devices and OS versions: One mid-tier Android and one older iOS device - all audits use the same hardware pool.
Define TTI measurement: Cold start → first interactive frame on primary revenue tab (e.g., home feed submit button enabled).
performance.mark in root layout and screen onLayout, or React Native Performance API.Define bundle scope: Main Hermes bundle + lazy chunks loaded before TTI - document which async splits count toward first-open budget.
npx react-native bundle with --verbose or Expo export analysis.Link regressions to rollback policy: Bundle or TTI regression beyond X% triggers Release & OTA Rules rollback consideration.
ADR for permanent budget exceptions: Heavy offline maps or video apps may need higher ceilings - document user value and mitigation (lazy load, on-demand download).
Hermes bundle size check in CI: Script compares exported bundle to threshold - fail PR when over budget without perf-budget label approval.
scripts/check-bundle-size.sh.Dependency size review on native and JS adds: npx expo install new package → check import cost and native binary size impact.
date-fns per-function imports, lodash-es subsets.No new synchronous require on startup path: Defer heavy modules with import() in non-critical tabs - grep require( in app/_layout.tsx and root providers.
Image and font budgets: WebP/AVIF where supported; subset fonts via expo-font - max two font families on critical path.
List performance lint in review: FlatList / FlashList requires keyExtractor, stable renderItem component, and memoized rows - no hooks inside renderItem.
renderItem hook violations.Reanimated and gesture worklets stay off JS thread: Animations dropping frames trigger investigation before merge - profile with react-native-performance.
LayoutAnimation on lists with 500+ rows.Cold start TTI on reference Android every RC: Record median of 5 cold starts - compare to last RC and budget.
Scroll FPS on primary list screen: Flip test through 50 items - sustained fps floor (55+) on reference device.
FlashList, smaller row components, image size fixes.Memory snapshot after 10-minute scripted session: Navigate tabs, open detail, return - heap stable without climb indicating leaks.
SharedObject handles from Expo modules on unmount.OTA update download size before production publish: eas update artifact size - warn when over OTA budget.
expo-asset pipeline; dedupe images.Network waterfall on primary screen: Parallelize independent requests; avoid serial await chains blocking paint.
Battery and thermal spot check: 15-minute navigation session - device should not throttle severely on reference Android.
Performance PR label: Changes touching lists, images, startup, or deps > 50 KB require perf-review from budget owner.
Weekly dashboard: TTI, crash-free, fps sampling (if RUM available), bundle size trend - visible to product and engineering.
New Architecture verification on upgrade: RN 0.86 New Architecture may shift perf characteristics - re-baseline after SDK bumps.
Feature flags for perf experiments: Roll out infinite scroll or new image pipeline to 5% before full OTA - watch RUM metrics.
No perf debt tickets without metric: "Make app faster" is not a ticket - "Reduce home TTI from 3.1s to 2.5s on Pixel 6a" is.
Store submission perf appendix: Release ticket includes TTI, bundle size, FPS audit results, and waiver ADR links - QA sign-off references numbers.
Many consumer apps target 3–5 MB gzipped Hermes main bundle at TTI. Heavy apps (maps, video) need higher ceilings with ADR justification. Measure your baseline before copying a number.
Low-end Android exposes jank, memory pressure, and slow storage that flagship iPhones hide. If it passes on mid-tier Android, iOS is usually fine; the reverse is false.
Mark app launch in root _layout.tsx, mark interactive when primary screen finishes layout and data gate opens. Use Performance API or custom analytics events - exclude dev client builds.
Yes - each update has a download size budget. Large OTA bundles slow updates and increase failure rates. Check artifact size before production eas update.
Default to FlashList for long heterogeneous lists on SDK 57. FlatList is fine for short static lists. Budget owner reviews list changes on primary feeds.
Rarely reliably - FPS audit stays manual or RUM-based. CI excels at bundle size and dependency checks; device lab covers frames.
Team-defined - often TTI regression > 15%, crash increase, or bundle overrun linked to startup crashes. Document triggers in Release & OTA Rules table.
Yes if bundled at build time. Prefer appropriately sized assets, modern formats, and remote CDN for rarely used media.
Fabric can improve list scroll perf but may shift memory - re-baseline after enabling on RN 0.86. Do not assume identical numbers to Old Architecture.
The named owner for that metric plus mobile lead. Waiver ADR includes expiry release version and mitigation plan.
Yes with sampling RUM (Sentry performance, custom analytics) - respect privacy, no PII in perf events. Staging mirrors prod architecture for deep dives.
Expo Go includes extra overhead - never use for budget sign-off. Use preview or production-profile EAS builds on reference devices.
Measure per app export from apps/mobile. Shared packages count toward each app that imports them - enforce package boundaries to avoid accidental bloat imports.
Stack versions: This page was written for React 19.2.3, React Native 0.86.0, and Expo SDK 57 (
expo~57.0.4).
Reviewed by Chris St. John·Last updated Jul 9, 2026