# Startup

cold-start latency, initialization cost, deferred loading.

## Measure

- startup profiling → process start → ready state
- init timing → per-module/per-component cost
- dependency analysis → large deps, their contribution to init
- metrics → time to first byte, time to interactive, init cost breakdown

## Optimize

- lazy loading → defer until first use
- split init → load only what's needed for initial path
- deferred setup → postpone non-critical initialization
- pre-warming → cache modules/data at startup
- trim deps → audit unused
- optimize critical path → speed the startup sequence

## Pitfalls

- over-lazy → complexity + first-use stalls
- cold ≠ warm start; different problems
- total time alone → break down by component
- caching with startup cost