Backoff

What it is

Backoff means waiting before trying the same operation again.

It is not just “retry with a pause.”

It is a way to slow the retries down so the failure does not get worse.

When it matters

Backoff matters anywhere retry already makes sense:

• service-to-service calls
• async workers
• webhooks
• third-party integrations

Without it, many failing attempts usually hit the same unhealthy dependency again at the same time.

Common mistake

The classic mistake is immediate retry or the same fixed wait for everyone.

That synchronizes the attempts and creates a new spike against a system that was already struggling.

Short example

A worker tries to send an event to an external partner.

The first attempt times out.

Instead of retrying immediately, it waits 1 second, then 2, then 4.

If it still fails, it stops and moves to another path.

Why it helps

Backoff gives the system room to recover.

It does not fix the failure on its own, but it keeps retry from becoming an incident multiplier.

Retry without backoff is blind persistence. Retry with backoff is persistence with brakes.