React Performance Optimization: From Slow to Lightning Fast

React is fast by default — until it isn't. As applications grow, rendering bottlenecks creep in. This guide covers the techniques for diagnosing and fixing performance issues in production React applications.

Rule Zero: Measure First

Use React DevTools Profiler to find the actual bottleneck. Open the Profiler tab, record an interaction, and examine which components re-rendered and why.

Preventing Unnecessary Re-renders

React.memo

const ExpensiveChart = memo(function ExpensiveChart({
  data, width, height
}: ChartProps) {
  // Only re-renders when props actually change
  return <canvas />
})

useMemo for Expensive Computations

const summary = useMemo(() => ({
  total: transactions.reduce((sum, t) => sum + t.amount, 0),
  byCategory: groupBy(transactions, "category"),
}), [transactions])

React Compiler (React 19)

React 19 introduces automatic memoization at build time, making manual memo/useMemo/useCallback largely unnecessary for new code. Enable with experimental: { reactCompiler: true }.

Code Splitting

const DataViz = lazy(() => import("./DataVisualization"))

function Dashboard() {
  return (
    <Suspense fallback={<Skeleton />}>
      <DataViz />
    </Suspense>
  )
}

Virtualization for Long Lists

Render only visible items. Use @tanstack/react-virtual for lists over 100 items. The overhead of the virtualizer is negligible compared to rendering thousands of DOM nodes.

Colocate State

The most impactful technique: put state as close as possible to where it's used. Global state that changes frequently causes the entire tree to re-render.

Performance Checklist

1. Profile first — don't guess
2. Colocate state — move it down
3. Virtualize long lists (100+ items)
4. Code split heavy routes
5. Optimize images (next/image or lazy loading)
6. Defer non-urgent updates (useDeferredValue)
7. Enable React Compiler in React 19

Wrapping Up

Focus on architecture (where state lives, how data flows) rather than micro-optimizations. A well-structured app is fast by default.