Your MDM Suite Lives Outside Your Data

Most MDM suites were designed when the warehouse was a reporting copy and operational systems were the center of gravity. They run on their own infrastructure, so every record has to be extracted, loaded into the suite, matched and published back.

Matching usually depends on deterministic and weighted rules written by hand. They work for the data they were tuned on and degrade as new sources arrive. Changing them is risky, because nobody can see in advance which records a rule change will move.

Meanwhile analytics, data science and AI moved to the lakehouse and the cloud warehouse. They need resolved entities there, current and at full scale, not a nightly feed from a separate system with its own license and its own copy of sensitive data.

Move matching into the data platform. Resolve entities with models trained on examples from your own data, combined with the exact rules you still need, and write golden records straight back to tables in the same account.

Migrate a domain at a time. Keep the identifiers downstream systems already use, compare the new results with the old ones before switching over, and retire the suite when the last domain has moved.