Choosing an MDM platform without falling for the vendor pitch

Every MDM platform demo works. It runs with clean data, small volume, simple rules, and a vendor specialist at the keyboard. Your environment has none of those four conditions.

The choice gets much better when you flip the logic: instead of watching the vendor's script, bring your own problem.

Bring your data to the proof of concept

Ask for the proof of concept with a real extract of your own (anonymized if needed) and with the volume you actually have, not a sample of a thousand rows.

A good part of what differentiates platforms only shows up with truly dirty data: inconsistent accents, fields swapped around, truncated records, mixed encoding. A tool that requires a clean database to work solves a problem you don't have.

The criteria that really separate them

Data model. Can you define your own domains and attributes without depending on the vendor for every change? If every new attribute is a ticket, the real cost is double the contract.

Matching rules. Can you combine deterministic and fuzzy rules, adjust the weight per attribute, and review the result before merging? Automatic merging without review is the fastest way to lose data.

Approval workflow. Is there a real workflow, with roles, approval authority, and a trail? Without it, the tool is a nice-looking database, and the governance problem is still a human one.

Two-way integration. Publishing the master data back to the source systems is where most projects stall. Ask how, with what latency, and what happens in a conflict.

Audit trail. Who changed what, when, and how to roll it back.

The questions that make people uncomfortable

How many customers with my volume and my industry are in production today, and can I talk to two of them without the sales team in the room?

How long did it take, from contract to the first domain in production, for the last customer similar to me?

What, in my case, would require customization, and who maintains that customization in the next upgrade?

Evasive answers to these three say more than any feature matrix.

The cost that isn't in the proposal

Integration with the source systems, almost always the biggest line. Preparing the data before the load. Dedicated internal people during implementation. A license that grows by record volume or by user: read how the bill scales over two years.

The criterion that comes before all the others

If you don't yet have a domain owner, a written standard, and a measured metric, no platform will solve it. It will automate the absence of decisions and hand you back the same problem, faster.

A tool amplifies the governance you already have. It doesn't replace it.