What is master data, and why does your reporting depend on it?

Master data is the data that does not belong to a single transaction but shows up everywhere: your customers, items, locations, suppliers and employees. An order changes daily; an item number should mean the same thing for years.

When that fixed data is shaky, everything resting on it is shaky. Two sites that know the same customer under two numbers do not produce two small errors, they produce one report the board does not dare to act on.

Where master data sits in the stack

Under every steering number sit transactions, and under those transactions sits the fixed data.

This is exactly why a data platform alone is not enough. You can join transactions flawlessly and still get a wrong total, because the same customer exists twice.

How to recognise it

The symptoms are recognisable without a study:

• Two departments report a different customer count, and both numbers are defensible.
• A spreadsheet exists that translates codes from one system to another, maintained by one person.
• Lead time per site cannot be calculated, because sites are not named the same everywhere.
• New items are created without required fields, because otherwise the order cannot ship.

That last point matters most. Poor master data almost always appears because careful entry slows the work down right now, and the cost lands on someone else later.

What does work

Start small, with a real owner and a definition you can check.

Start with one domain that hurts: usually item or customer. Name an owner with a name, not a department. Write down which fields are required and which values are allowed. Then measure how often reality deviates, and repair the entry rather than the reporting.

That takes a technical and an organisational side. See data quality has two owners for why those cannot be separated, and what a data contract is for how to make the agreement between producing and consuming systems binding.

Master data and AI

Once AI agents work on your data, this gets urgent. A human who sees two customer numbers recognises the duplicate and quietly corrects it. An agent does not: it calculates on and presents the result with the same confidence as a correct answer.

Reliable master data is therefore not a cleanup project but the condition for everything you want to build on top. To see where your foundation stands today, the foundation scan gives you a picture in a few minutes.

In short

Master data is the fixed data underneath your processes. It deserves an owner, a written definition and correction at the moment of entry. Every report, dashboard and agent above it is no more reliable than that.