Name the record and its owner
Pick one workflow and trace it from start to finish. For each record, identify the system that owns it, who may change it and which other systems need a copy. This makes disagreements about duplicate or outdated information easier to resolve.
- The system that owns it
- Who may change it
- Which systems need a copy
Include the exceptions
A successful request is only part of an integration. Discuss what should happen when a record is incomplete, a payment is reversed or a connected service is unavailable. Decide who investigates an exception and how they will find the information they need.
- A record is incomplete
- A payment is reversed
- A connected service is unavailable
Bring representative examples
Sample records, interface documentation and a description of access restrictions help establish the scope. Use anonymised examples where appropriate. Include typical records and awkward cases rather than only the cleanest examples.
- Sample records
- Interface documentation
- Access restrictions
- Awkward cases, not only clean ones
Plan the transition
Agree how existing data will be checked, which team will sign off and whether old and new processes need to run together for a period. The release plan should include a way to reconcile records and a decision about what happens if a critical check fails.
- Check existing data
- Agree who signs off
- Run old and new together if needed
- Decide what happens if a critical check fails
