One record. And a reason to believe it.
Master data management is mostly the unglamorous business of deciding which of five conflicting records is right, and making that decision the same way every time — then being able to explain it afterwards.
The technology is the easy half.
The same customer appears in your CRM, your billing system, your service platform and two spreadsheets, under four spellings and three ABNs. Every report built on top of that disagrees with every other report, and the executive response is usually to commission a fifth system.
The matching and merging is solvable — that part is well-understood engineering. The hard half is organisational: getting four teams to agree on what a customer is, who gets to decide when the rules disagree, and what happens to the record when the system that created it is retired.
How we run it
One domain at a time, in six-month blocks. We model the entity with the people who use it, agree the match rules against your real data rather than against theory, and build survivorship rules that a business person can read. Then we put a stewardship workflow around the exceptions, because there will always be exceptions and a queue nobody works is just a slower kind of failure.
Every merged value carries its lineage: which source contributed it, which rule selected it, and when. When someone senior asks why their number changed, the answer is on screen rather than in a consultant's memory.
What we won't do
We won't model all your domains at once. We won't build a match rule set so elaborate that nobody can maintain it after we leave. And we won't automate a merge decision your business hasn't agreed to — a confidently wrong golden record does more damage than five obviously conflicting ones, because people stop checking.
- Cadence
- 6-month blocks
- Scope
- One domain at a time
- Typical domains
- Customer, product, asset
- Checkpoint
- Month 3
- Platform
- Yours or recommended
- Handover
- Runbooks + training
If your assessment shows a single domain in poor shape and the rest sound, say so early — a targeted three-month MDM engagement is often the right answer and we'd rather scope that than pad it.
What gets built.
- Entity model per domainAgreed definitions and attributes, signed off by the people who have to live with them.
- Match & merge rulesTuned against your real data, with precision and recall measured rather than assumed.
- Survivorship rulesWhich source wins for which attribute, and why — written so a business reader can audit it.
- Stewardship workflowException queues, review screens and escalation paths that a real steward can work in the time they actually have.
- Lineage & explainabilityEvery value on a golden record traceable to its source and the rule that selected it.
- Downstream cutoverConsuming systems and reports moved onto the golden record, with reconciliation before and after.
- Quality measurementCompleteness, duplication and conformity tracked over time — so the improvement is a number, not a claim.
- Operating documentationRunbooks, rule catalogue and training for the team taking it over.
Worth knowing before you buy.
Do we need to buy an MDM platform?+
How long until we see a golden record?+
What about records that can't be matched automatically?+
Start with the assessment.
Two to three weeks. You get a map of your systems, named owners, a ranked list of gaps, and a costed 6-month plan. No obligation to run it with us.