Validate identity and matching
Integration failures can involve record matching as well as transport: the wrong record, a duplicate, or a record that could not be found. Define how entities are matched across systems and what happens when the match is uncertain.
Treat mapping as a product
Field mappings, code translations and default values should be documented, versioned and reviewed, not buried in code. When a mapping changes, the people who depend on the data should know.
Design exception handling up front
Decide in advance what happens to messages that fail: retry, queue for review, or reject with a reason. Assign a person or team to every exception queue. An exception without an owner can leave data missing or unprocessed.
Reconcile end to end
Compare what was sent with what arrived and what changed downstream. A simple daily reconciliation report often finds problems weeks before users do.
Use synthetic data until access is approved
Build and test with realistic synthetic examples. Move to production data only after the necessary agreements, access controls and approvals are in place.