Master data has a built-in flaw: there is nothing to show. No screen that impresses a board, no chart pointing upwards. A customer master looks exactly the same before and after you clean it up, and the difference sits in the number of records describing the same entity.
That is not an aesthetic problem, it is a budget one. In his column on twenty years spent in this layer, Jacek Bugajski, CEO of SNOK, described how master data spent two decades losing priority to things you can demonstrate in fifteen minutes, and how much organisations lost by never seeing the cost in one piece.
So instead of one more architecture diagram, we filmed a production line.
What happens on that line
Material drops from the hopper, and nobody likes the look of it: irregular, cracked lumps. That is what actually arrives from source systems. The same business partner appears in the ERP under its full legal name, in the CRM as an abbreviation somebody typed by hand, in the purchasing system without a registration number, and in the master record with an address from before the move. Each of those systems is locally right. None of them is completely right.
The film then shows three stations, because a Master Data Management project has three as well. In a real implementation deduplication and standardisation interleave - normalising the format before matching improves how well the matching works - but one thing is fixed: enrichment comes last.
Deduplication. Establishing how many entities a set of records actually describes. In SNOK MDM a language model proposes the match together with its reasoning, but a data steward approves the decision, because the matching result is probabilistic. An engine that merges master records without human oversight causes damage that is harder to undo than the duplicates themselves.
Standardisation. One format, one dictionary, validation rules at the point of entry. Without that last step the order you established holds until roughly the first week after go-live.
Enrichment. Filling in critical fields and connecting external sources, and only on a set that is already clean. In the reverse order you enrich the duplicates too.
At the end of the line the NOKs pack stamped ingots into a crate. Documentation calls it the master record, conversations more often the golden record: one version of the truth about a customer, material, product or employee, used by every system in the company.
Where the metaphor stops working
The film runs sixty seconds and carries three simplifications we would rather state plainly.
First: a factory line has one input, an organisation has a dozen or more, running in parallel and at different speeds. Cleaning up data is not a flow from left to right, it is reconciling versions between systems that keep operating throughout.
Second: the material does not run out. The hopper keeps feeding, because new records are created every day. Master data is a process, not a project with an end date, which is exactly why validation rules at the point of entry matter more than a one-off cleanse.
Third: the film fits one line, while there are six master data domains: employees, customers, materials, suppliers, products and financial accounts. Our experience of recent years says that one domain finished properly beats a programme covering all six at once. A programme spread over several quarters usually loses its sponsor before it delivers the first measurable result.
Why we are showing this now
Because the stakes went up, and not because of fashion. Salesforce closed its acquisition of Informatica on 18 November 2025. SAP acquired Reltio, closing on 7 May 2026, and the announcement put it plainly: the point is preparing SAP and non-SAP data for artificial intelligence. In April 2026 Gartner reinstated the Magic Quadrant for Master Data Management Solutions - the first edition since December 2021.
The reason is practical. An AI agent works with whatever it is given, and with four versions of the same business partner it will answer instantly, fluently and incorrectly. Gartner forecasts that through 2026 organisations will abandon 60 percent of AI projects that are not supported by AI-ready data (press release of 26 February 2025). Automation does not clean data. It increases the speed and reach of whatever is already in it.
Organisations facing an S/4HANA conversion have a deadline of their own. The Business Partner model is mandatory there, and the job of Customer Vendor Integration is to move records into the new structure, not to improve their quality. Duplicates and gaps come back as exceptions, usually inside the cutover window, which is the worst possible moment.
Where we start in practice
Not with a programme and not with a tool selection, but with a number. We take a sample from one of your domains and measure two things: how many records describe the same entity, and what share of items have gaps in critical fields. We deliver the result within five working days, without access to your production environment. That measurement settles whether data clean-up belongs in the project scope or stays on the risk list.
The service itself is described on our Master Data Management page, and the platform we implement on is on the SNOK MDM product page. It has been in service since 2022, covers six master data domains, can run inside your own environment, and its source code can be placed in escrow.
Anyone who wants the full context, with the anti-patterns and the order of work before an S/4HANA conversion, will find it in Jacek Bugajski’s column. This film is the shorter version of it, made to be shown to someone who has never heard the abbreviation MDM and signs the budget.
A note on the film itself
It was made entirely in-house, with generative techniques, in the rhythm of a series. The NOKs say nothing, all the humour is physical, and the factory is only scenery for a single thought: order in data comes from a line somebody designed, not from a one-off tidy-up.
This is the first episode. In the ones that follow the NOKs work a quality gate, a line that speeds up, and a new machine they have to learn. Anyone who guesses what each of them is about knows our offer better than our website does.