/
All articles

Good solutions

Stop cleaning data in a hub – hold a mirror up to your source systems

Many organisations spend large sums on MDM hubs that quietly correct faulty details. But when the hub cleans the data, the source systems stay wrong and responsibility disappears. Quality only improves when the people entering the details see which business rules were broken and who should fix the error.

By Jens Fahlén · Published · 4 min read

The structured monster – the information monster stacking the building blocks in order

Introduction

When an organisation discovers that information in its systems is inconsistent – units lack a cost center, addresses are incomplete and old permissions linger – a common reaction is to buy a central MDM hub or build a flow that cleans the data.

The idea is tempting: let the business systems accept whatever is entered, and correct the details centrally before they reach reports and integrations. But when errors are quietly fixed further down the chain, the person who entered the detail never notices that anything was wrong.

1. The washing machine hides the errors

Traditional MDM hubs receive details from source systems, merge duplicates, fill in missing values and pass on a “golden record”. Often it is then synchronised back to the source systems.

It looks like order, but three things happen at once:

The source systems stay wrong: Nobody in the HR or finance system sees that the detail was wrong, because the hub corrected it somewhere else.

Responsibility moves: The error becomes the integration's problem instead of the source's. When an automated process fails, everyone looks in the hub, not where the detail was created.

Trust declines: Details in downstream systems suddenly look different from what the case worker sees on their own screen.

2. Without visible errors there is no reason to get it right

If an HR administrator can save a new unit without a base cost center, and a hub later sets a default value, the next unit will be saved the same way. The error becomes part of how work is done.

Data quality cannot be outsourced to a central hub. Whoever creates the information also needs to know when it is wrong – and be able to correct it where it originated.

3. From percentages to business rules

Many data quality tools measure the share of fields that are filled in: “the register has 88 per cent data quality”. The figure says little about what needs to be done, or by whom.

What is needed are checks against business rules – rules describing how information should fit together. Some examples from the library:

Base cost center (VR-ECO-002): Does each unit have exactly one base cost center? If not, invoices and licences cannot be posted automatically.

Process roles and engagements (VR-PRC-003): Does anyone still hold a process role even though their engagement has ended? Then they can still order or approve.

Identity and accounts (VR-IT-002, VR-IT-012): Are there active accounts whose digital identity is no longer active, or accounts without an owner with an active engagement? Then there is a security gap.

Each rule says both what is wrong and why it matters. That makes it possible to prioritise.

4. A mirror instead of a washing machine

A better tool does not clean the details. It reads from the source systems, tests them against the rules and shows the result to the person who can correct the error – in the source.

Instead of a report with thousands of technical deviations sent to an architect, the right person receives a clear message:

To the unit manager: “Two employees still hold the role orderer even though their engagement has ended. Remove the roles.”

To HR: “The unit Home Care Central lacks a base cost center. Add it in the HR system.”

To finance: “The supplier lacks a GLN. Add it before the first invoice is paid.”

To IT: “The function account for the payroll system has no owner with an active engagement.”

The same rules can also show how well each source follows the model over time, and where errors most often arise – at onboarding, reorganisation or a change of manager.

5. Why tools are rarely enough

There are plenty of tools for data quality, observability and identity governance. They can run rules, but they rarely supply the rules themselves. They do not know what a unit, a base cost center or a process role is.

So organisations often spend a long time formulating the rules before the tool is useful. A shared rule set with building blocks, keys and code lists is the missing part – and the part that can be reused whatever the tool.

Conclusion: Quality is created at the source

Correcting details afterwards in a hub treats the symptoms. A more sustainable approach rests on three principles:

1. Describe business rules for how information should fit together.

2. Test the sources' details against the rules automatically.

3. Show the error to the person who owns the detail, so it is corrected where it was created.

Then quality work becomes part of everyday work rather than a separate project in a hub.

Tags

  • #master data
  • #data quality
  • #MDM
  • #source principle
  • #business rules

What did you think of this article?