What actually breaks first — and how to prepare

Scaling problems rarely look the way teams expect. Here are five patterns we've seen repeatedly in teams that grew from a handful of users to a million, and what helped them get there smoothly.

1. Cache aggressively, invalidate carefully

Most early performance wins come from caching reads that don't change often — and the discipline to invalidate them correctly when they do.

2. Move heavy work off the request path

Anything that doesn't need to happen before you respond to a user — emails, exports, syncs — belongs in a background job, not the request handler.

3. Index for your actual queries

Add indexes based on what your application actually queries, not what seems important in theory.

Scaling is rarely one big rewrite — it's a series of small, well-timed changes.

4. Watch your API's shape, not just its speed

Endpoints that return unbounded lists or deeply nested data will eventually become a bottleneck, even if they're fast today.

5. Automate what used to be manual

As traffic grows, anything a human does repeatedly should become an automated workflow — from onboarding emails to data cleanup jobs.