Planning Retry Rules for Business Automation
Retrying can recover from temporary failures, but it can also repeat an action that already happened. A retry policy needs to distinguish technical interruption from a rejected business record.

Classify failures before repeating work
Separate temporary unavailability, invalid data and uncertain outcomes. Decide which steps are safe to repeat and which need reconciliation first. Limit repeated attempts and move unresolved cases into a visible queue rather than leaving them in an endless loop.
Practical checklist
- Set a bounded retry policy.
- Avoid retrying invalid data unchanged.
- Check for earlier success.
- Escalate unresolved outcomes.
How to review the result
Use a simulated outage and a deliberately invalid record to verify different behaviour. Both should end in a clear state that staff can understand.