Make data sync errors visible and recoverable
Design an error queue, safe retry and regular reconciliation between connected systems.

In short
- Distinguish an unavailable service, invalid input and an ownership conflict.
- Retain the original input.
- Use a window that allows for records still in transit and different time zones.
- Next steps
On this page
Classify errors by cause
Distinguish an unavailable service, invalid input and an ownership conflict. An outage can be retried. A missing customer ID needs correction. Show the reference and what has already been updated.
Check before retrying
Retain the original input. Before sending it again, check whether the action succeeded without an acknowledgement. Reuse its unique key and log manual corrections. A retry should not produce another booking or order.
Compare the systems regularly
Use a window that allows for records still in transit and different time zones. Track the number of open errors, the oldest error and recovery time. Decide who receives an alert and what they can do next.
Next steps
Creagrid / actie
Practical checklist
Describe the problem and intended outcome.
Involve the people who use the process every day.
Test the riskiest assumption with a small prototype.
Define tests, ownership and maintenance up front.

