Getting the detail right.
WHAT TO DEFINE FOR EACH INTEGRATION
What data moves In which direction How often What triggers it What happens on failure
WHAT MAPPING ESTABLISHES
Which field corresponds to which.
WHAT MAKES IT DIFFICULT
Systems structuring the same information differently.
WHAT COMMON MISMATCHES ARE
Different code sets for the same concept Different levels of detail Fields present in one and absent in the other Different validation rules
WHAT TO DO ABOUT CODE DIFFERENCES
Maintain a mapping, deliberately, with an owner.
WHAT IDEMPOTENCY MEANS
The same message applied twice producing the same result.
WHY IT MATTERS
Delivery may be retried, and duplicates must not create duplicate records.
WHAT TO USE
A unique identifier per message, checked before processing.
WHAT ERROR HANDLING MUST DECIDE
Whether to retry How many times What happens when retries are exhausted Who is told
WHAT TO NEVER DO
Discard failed messages silently.
WHAT TO PROVIDE
A queue of failures, reviewable and reprocessable.
WHAT TO LOG
Every transfer, with enough detail to investigate.
WHAT TO MONITOR
Volume against expectation, and failures.
WHY VOLUME SPECIFICALLY
An integration that stops silently looks identical to one with no data.