By The Skyline Project ·
Write a one-page integration brief
Scrollable table: swipe horizontally to see every column. You can also focus the table and use the arrow keys.
| Decision | What to write down |
|---|---|
| Process | The action that starts the flow and the result people need |
| Data | Record ID, key fields, system that owns changes and permissions |
| Failure | What can be retried, what needs human review and how alerts reach the owner |
| Acceptance | Sample records, expected results, comparison window and sign-off owner |
| Rollback | How to stop new changes, return to the manual procedure and reconcile records |
Illustrative example. A team wants every web order to appear in an operations queue. The web store owns the original order ID. The queue uses that ID to update the existing record if the same event arrives again. If the source records cancellations, a cancellation should change the queue state and alert staff when work is already in progress. Test a new order, duplicate notification, cancellation and a temporary outage. Count missing and duplicate records against the source before widening the rollout.
Set a pilot exit rule
Limit the first test to one workflow or a small group of items. Log exceptions and who resolved them. Approve expansion after the agreed tests pass and the owner can run the fallback. A repeated action may create duplicate records unless implementation accounts for it, so check which writes are safe to retry. Microsoft Azure Architecture Center, Retry pattern.
For a specific inventory example, use our inventory reconciliation worksheet. Explore technology integration or describe the tools you need to connect.
