Translating between models.
WHY IT IS THE HARD PART
Two systems structure the same information differently.
WHAT DIFFERS
Field names Levels of detail Code sets for the same concept Required and optional fields Validation rules
WHAT A MAPPING IS
A documented correspondence between them.
WHAT TO RECORD FOR EACH FIELD
The source The destination Any transformation What happens when it is absent
WHY THAT LAST COLUMN MATTERS
It is where most integration defects live.
WHAT TO DO ABOUT CODE DIFFERENCES
Maintain a lookup, deliberately, with an owner.
WHAT TO DO ABOUT VALUES WITH NO EQUIVALENT
Decide explicitly: reject, default, or pass through.
WHAT TO NEVER DO
Guess silently.
WHAT TO VALIDATE
That the mapped result is valid in the destination, before sending.
WHY
Rejections at the destination are harder to diagnose than local failures.
WHAT TO PRESERVE
The original message, as received.
WHY
It is the only way to reprocess after fixing a mapping fault.
WHAT TO BE CAREFUL WITH
Dates and time zones Decimal precision Character encoding Names and addresses with unusual characters
WHAT TO TEST WITH
Real data, including the awkward records.
