Trust grows when people stop having to verify. You do not build it with a campaign, a training day or a new tool. You build it by making the numbers leadership uses most come out the same way, every time, no matter who produced them.
What distrust looks like in practice
Nobody says "I do not trust the data." They say "let me double-check that," and then they open a spreadsheet. The finance team's revenue number and the marketing team's revenue number differ by a few percent, so a meeting that should decide something spends twenty minutes on which one is right. An analyst gets asked to confirm a figure that is already on the dashboard.
Each of these looks small. Together they are a cost you can count: the hours spent rechecking, rebuilding and reconciling, multiplied by what those people are paid. That is FTE Debt, and a lot of it is the price of distrust.
Start with five numbers, not five hundred
Trust programs fail when they try to fix everything. Pick the handful of figures that leadership uses to make decisions: revenue, margin, active customers, acquisition cost, retention. For each one, fix three things.
- One definition. Write it down in a sentence. If two teams define "active customer" differently, decide which one wins and retire the other.
- One named owner. A business person, not only the data team, who answers when the number looks wrong.
- One visible calculation. Anyone who sees the number can see how it is built and when it was last refreshed.
This is unglamorous work. It is also the part that changes behaviour, because people stop checking a number once they know who stands behind it.
Measure the checking, not the sentiment
Surveys about "data trust" produce a score and no decision. Measure something you can act on: how many hours a month go to verifying, reconciling or rebuilding figures that already exist. Ask three or four business teams to estimate it. The first answer is rough and that is fine. Label it as an estimate, then track whether it falls after you fix the first numbers.
Why a tool will not fix this
A catalogue, a quality monitor or a semantic layer can help once the definitions and owners exist. Bought first, they document the disagreement instead of resolving it. The sequence matters: agree what the number means, name who owns it, then automate the check.
A test for this week
Pick one number from your last board pack. Ask two different people to produce it independently. If the results differ, you have found where the trust problem lives and what it costs to keep re-checking. If they match, ask how long it took each of them. That is the cost of trust at its best.
For a first estimate of what distrust costs in your company, try the FTE Debt calculator, or read what the Data Zombie Tax adds up to.