Data quality is a symptom, not a cause

The same sentence comes up in almost every first conversation: our data quality is poor. What usually follows is a proposal to set up a data quality programme, define rules and procure a tool. Twelve months later the tool is in place, the rules are in place, and the sentence is still being said.

The complaint means something else

When an executive talks about poor data quality, they almost never mean faulty records. They mean they cannot trust a figure, because two reports show different values and nobody can say which one holds. That is not a technical defect. It is an unresolved question of accountability.

Checking this takes ten minutes. Take the three figures your leadership team quotes most often and ask the same four questions about each of them.

• Who in the business decides how this figure is calculated — by name, not by department?
• In which system does the value originate, and which systems recalculate it afterwards?
• Who is allowed to change the definition, and where is the last change documented?
• How long does it take to trace the value back to the original business transaction?

If any of these questions goes unanswered, you do not have a data quality problem. You have a figure without an owner.

Why tools do not solve this

A data quality tool measures rules that somebody defined beforehand. It checks whether a field is populated, whether a date is plausible, whether one total reconciles with another. It cannot decide which of two competing business definitions of contribution margin is the right one. That decision belongs to management, and buying a tool does not make it. It postpones it.

This also explains why such initiatives rarely fail and still change nothing. They deliver exactly what was commissioned. Only the wrong thing was commissioned.

A number without a named owner is not a metric. It is an opinion with decimal places.

What has to happen first

Assign one accountable person from the business to each of the twenty most important figures in your organization. Not IT, not a committee, but a person who owns the business definition and approves any change to it. Record which system the value comes from and through which intermediate steps it reaches the report.

This is unspectacular work and it can be done in eight to twelve weeks. Afterwards every discussion about a figure has an addressee, and the question is no longer which report is right, but whether the recorded definition still fits the business.

Only at that point does a tool pay off. It then measures something that somebody is accountable for.

The side effect that carries the investment

Organizations that have made these assignments move considerably faster through audits. When a supervisor or an auditor asks what a statement rests on, there is a person to talk to and a documented path back to the origin. The same structure later carries the evidence duties of the EU AI Act, because there too the point is not to explain the result but how it came about.

Knowing where your numbers come from lays the groundwork for everything that follows. Not knowing means buying tools that measure a problem they cannot solve.