What

is

master

data

management

(MDM)?

What is master data?

Master data is the data that describes who you do business with and what you do business with. A product with its code, description and unit is master data. The order in which that product is sold is transaction data.

The difference is lifespan. Master data changes little and is reused constantly; transactions are created every day and refer to that master data. In an ERP this is, for example, the item master and the customer master. An error in the master data carries through to every transaction that refers to it.

What does master data management do?

Master data management makes sure there is one reliable version per account, product or location, and that every system uses that version. It works in four steps: match data from different sources, merge duplicate records, enrich missing data and deliver the result back to the systems that need it.

The result per object is called a golden record: the combined, verified version that counts as the truth. The agreements around it matter at least as much as the technology. Who may create a new supplier, who changes a product description, and in which system does that happen?

Which domains fall under master data?

Customers and accounts.

Names, addresses, registration numbers, contacts and payment terms. This domain appears in the CRM, the ERP and often in a portal or web store too, which makes it the most prone to duplicates. Where personal data is involved, the GDPR also requires it to be accurate and, where necessary, kept up to date (Article 5(1)(d)).

Suppliers.

Supplier data resembles account data but has its own use: purchasing terms, bank details and certifications. An error in a bank account number hits payments directly.

Products and items.

Item codes, descriptions, units, weights and specifications. The ERP often holds the logistics side, while texts and images for sales channels are managed in a PIM system. MDM decides which system determines which attribute.

Locations and assets.

Sites, warehouse locations, machines and installations. For organisations with many physical objects, such as in real estate or energy, this is often the domain most systems depend on.

What is the difference between MDM, data governance and data integration?

These three terms are often used interchangeably, but each answers a different question. The next section shows, per term, what it is, what it covers, who does it and what it delivers.

Which MDM architectures are there?

There are four common set-ups, which differ in how much control the central MDM facility has.

In a registry the data stays in the source systems and the central facility only tracks which records belong together. That is the lightest form. With consolidation data is merged centrally into golden records for reporting, but the source systems stay leading for data entry.

With coexistence the verified data also flows back to the source systems, so they stay aligned. In a centralised or transactional hub master data is only created and changed centrally, and all systems read from it.

The heavier the set-up, the more control, but also the bigger the impact on existing processes. Most projects start light and grow from there.

Do you need an MDM package, or agreements?

An MDM package does not solve an agreement that isn't there. As long as nobody knows who may change a product description, a package just moves the problem to a new system.

So start with the agreements: an owner per domain, a leading system per data type, and rules for what a valid value is. A dedicated package becomes relevant when there are many source systems, several domains at once or strict traceability requirements. It is a software category with heavy implementations. Weigh that investment against what the agreements alone deliver.

Master data management without a dedicated package.

Many organisations can set up MDM with the systems they already have. The ERP is often the logical source for customers and items, because that is where financial and logistics processing happens. Other systems then read from that source instead of keeping their own copy: the principle of a single source of truth.

What that takes: validation at data entry, checks for duplicates and integrations that pass changes on. That makes MDM part of how you organise data management, and it stops data silos from coming back. Whether it works shows in the data quality of your master data.

In short.

Master data management is managing master data, such as accounts, suppliers, products and locations, in one reliable version that every system uses. It differs from data governance, which sets the agreements, and from data integration, which moves the data. MDM starts with agreements on ownership and a leading system per data type. A dedicated package is only needed once those agreements exist and the landscape is too large to manage them with existing systems.