Why Do Companies Still Make Decisions With Data They Don't Fully Trust?

In one of our projects, every system appeared to be reporting correctly. The ERP showed one picture. Other business systems showed another. Each number looked reasonable on its own - until we compared the data.

That was when the more important question became clear: which version of reality should the business trust when making a decision?

That matters more than simply asking how much data a company has.

Why isn't “we have the data” the same as “we can trust it”?

Companies already have extensive data for most business decisions: order histories, inventory levels, customer records, financial information and years of operational history.

The challenge often appears elsewhere.

Reliable information can sit next to data with known weaknesses, while the system presents both with the same apparent level of confidence.

For example, a CRM system may be completely accurate about a customer's name and address while the “last contacted” field tells a different story because one department stopped updating it after a system change.

When both fields enter a report or model, they can look equally reliable.

Data quality isn't always a property of the dataset as a whole. It can vary by field, process or business event.

Where does the knowledge of what to trust actually live?

Often, it lives with a person rather than in the system.

Someone in operations knows that a particular field has been unreliable since a system migration. Someone in finance checks one specific figure every month because they have seen it drift before.

That knowledge can be accurate and valuable.

But when it stays with the person who knows the history, it rarely reaches the report, forecast or model using the same data.

This creates an unusual situation: someone in the business knows that a particular data point is unreliable while the system continues to treat it as reliable.

Companies often respond by creating additional spreadsheets, manually adjusting figures, adding informal rules or relying on individual experience.

The decisions still get made.

The uncertainty simply moves outside the system.

What does verifying trust actually look like before making a decision?

It doesn't require auditing every dataset in the company.

A more useful approach is much narrower: verify the specific data the decision depends on.

For example:

• When was this field last updated?
• How consistently is it populated?
• Does it match another source that records the same business event?
• Has the process producing this data changed?
• Is there someone who knows why this particular number may be unreliable?

Comparing the data with an independent source can be especially valuable.

A stock forecast can be compared with actual warehouse scans.

A “last contacted” field can be checked against recorded communications.

A reported process milestone can be compared with the operational events recorded elsewhere.

This is close to what process mining does: it analyses event data generated during process execution and can compare recorded system information with how the process actually unfolded.

Its value comes from checking the evidence rather than simply accepting the system's version.

What changes when data trust becomes part of the decision?

The question changes from:

“Is our data good?”

to:

“Can we trust this particular data for this particular decision?”

Different fields can carry different levels of confidence.

A company needs to know which data is reliable, which requires additional verification, and which decisions depend on it.

This becomes even more important when AI enters the process.

An AI system can process far more information than a person could review manually. That makes the quality and context of the underlying data even more important.

When questionable data enters an automated decision process, the system can apply the resulting conclusion consistently and at scale.

The result can be questionable data producing highly confident decisions.

So: why do companies still make decisions with data they don't fully trust?

Because trust is often assumed by default rather than checked when the data is used for a specific decision.

The data may be available.

What is often missing is the context that tells the business which parts of that data deserve confidence right now.

Our project experience showed how easily each system can appear correct while the combined picture describes a very different reality.

So before the next forecast, report or AI-driven decision, it is worth asking a more precise question than “Is our data good?”:

“Which specific data are we relying on for this decision - and who in the business would know if that data no longer reflected reality?”