From prototype to production: ask new questions.

Preserve the learning

A prototype validates a hypothesis; its code and architecture are not necessarily permanent foundations. Identify what was demonstrated, what remains uncertain and what needs rewriting. Use explicit criteria rather than sunk cost to decide.

Make environments reproducible

Version configuration, automate builds and deployment, and separate environments and secrets. Database schema changes deserve as much attention as application code. Plan compatibility between versions during transitions.

Design for operations

Define critical journeys, service indicators and actionable alerts. Test backups, dependency failures and capacity limits. Decide who receives alerts and what they can do.

• Centralise useful logs without exposing secrets.
• Create representative health checks.
• Document restore, rollback and emergency access.

Release progressively

An initial production rollout can target a limited scope. Define success and stop criteria before switching. Use real observations to prioritise further work; additional architectural complexity should address a measured need.