One agreed version of your customers, suppliers and materials — and the ownership that keeps it that way.
Master data management is the discipline of maintaining a single, governed version of the core entities a business shares across systems: customers, suppliers, materials, employees and the chart of accounts. It covers definition, ownership, creation standards, deduplication and distribution.
Bad master data is quiet and expensive. The same customer under three records makes credit limits meaningless and spend analysis wrong. Duplicate material codes generate excess stock of an item you already hold. None of this shows up as an error; it shows up as decisions made on numbers that were never right.
What a customer is, which fields are mandatory, and who owns the record. Ownership is the part most often skipped and most needed.
Naming conventions, mandatory fields and validation at entry. Preventing a bad record is far cheaper than merging it later.
Fuzzy matching at creation against name, registration and address. Duplicates created today are merge projects next year.
How the governed record reaches the systems that consume it, and which system is the acknowledged source of truth.
Named data stewards per domain who resolve conflicts and approve changes. Without them governance is policy nobody applies.
A business runs a duplicate analysis on its customer master and finds that one large account exists as five records: two created by different sales staff with name variations, one from an acquired company's data load, one created for a different delivery address, and one with a trailing space in the name.
The consequences are not merely untidy. Credit limits are set against individual records rather than the group, so total exposure to that customer exceeds the limit anyone approved. Sales reporting understates the relationship, so it is not prioritised as it should be. Statements go out separately and get queried. And a pricing agreement applies to one record, so the customer is charged differently depending on who takes the order.
Deduplication fixes the history. What prevents recurrence is governance: a defined matching rule applied at creation, a single owner for the customer domain, and a creation process that searches before it creates. Without those, a cleansed master file degrades back to the same state within two years, which is why cleansing projects that do not establish governance get repeated.
| Term | What it means |
|---|---|
| Golden record | The single agreed version of an entity, reconciled from all sources. |
| Data steward | The person accountable for the quality of a data domain. |
| Survivorship rule | The logic deciding which value wins when sources disagree. |
| Match and merge | Identifying records that represent the same entity and combining them. |
| Data domain | A category of master data — customer, supplier, item, employee — each with its own owner. |
A cleanup project without creation controls simply regenerates the duplicates over the following year.
Two systems both authoritative for customers means neither is, and reconciliation never ends.
Records created in CRM or a webshop bypass it. Validate wherever creation happens.
If nobody owns a domain, nobody resolves its conflicts and the standards decay.
| Metric | What it tells you | What distorts it |
|---|---|---|
| Duplicate rate by domain | The headline quality measure, and the one most likely to surprise. | |
| Records failing validation rules | Incomplete or inconsistent data, by domain and by field. | |
| Time to create a master record | Governance friction. Slow creation drives workarounds and shadow records. | |
| Downstream errors traced to master data | The business impact, which is what justifies the governance effort. |
We build and support the systems this page describes. Start with a free working proof of concept — no cost, no commitment.