Before deployment, establish what changed: files, components, services, configuration, databases, infrastructure, and customer facing behavior. Do not assume diff size equals production impact.

Check dependencies

Identify what the changed code depends on and what depends on it. A small shared component can affect more systems than a large isolated service.

Check testing and verification

Review unit, integration, end to end, staging, manual, static analysis, and security evidence. Also understand what the tests do not cover.

Look at production behavior

Use current errors, latency, traffic, dependency health, anomalies, and component usage where available.

Check history

Previous incidents and recent change activity can reveal risk that is not obvious from the current diff.

Consider deployment and rollback

Ask whether the change can be rolled back, how quickly problems would be detected, whether rollout can be staged, and whether feature flags are available.

A simple framework

Change → Evidence → Uncertainty → Action. Understand the change, gather evidence, identify uncertainty, then choose the control that reduces it.
Tomosu's Production Reliability Index (PRI) brings multiple reliability signals together around a change.
Explore PRI: https://tomosu.ai/start