The dashboard exists, and nobody opens it
Someone spent months on it. It lives in Power BI or Tableau or Looker Studio, it has a dozen pages and it looks professional. Yet in the Monday meeting the sales manager still brings a printed spreadsheet, and the finance director asks for the numbers to be emailed.
Ask why and you get a story: the stock figure was wrong once, the sales total did not match the accounts, a chart showed zero for a week because a feed had stopped. One bad experience was enough.
Trust breaks on the data, not the charts
Dashboards are usually judged on their visuals, but trust is lost underneath. The common causes are dull ones.
| What people see | What is usually behind it |
|---|---|
| A number that does not match the accounts | Different definition or filter, never explained |
| A flat line or zero for a few days | A data refresh failed silently |
| Yesterday's figures missing | Refresh runs before the source updates |
| Totals that change when you add a filter | Duplicate or unmatched records |
| Pages nobody understands | Built to a spec, not to a question |
None of these are fixed by redesigning the dashboard. They are fixed in the pipeline that feeds it, and by telling the reader how current and how complete each number is.
What an untrusted dashboard costs
You pay for the licences and the build, and you still pay for the spreadsheets. Staff spend time producing both. Meetings spend time on whose numbers are right. And decisions get made on the spreadsheet version, which has no checks at all.
There is a quieter cost too. Once people learn to ignore the dashboard, they also ignore it when it shows something genuinely important.
And the original builder, if they are still around, usually gets the blame. Often they built exactly what was asked for, against data that nobody had checked, with no time set aside for the dull work of reconciling it.
How we rebuild trust in the numbers
- We interview the people who stopped using it and list the specific numbers they doubt. Those become the test cases.
- We trace each doubted number from the chart back to the source system and find where it goes wrong: a join, a filter, a definition, a refresh schedule.
- We fix the data model underneath, usually in a small database or the dashboard tool's own data layer, with the business rules written down.
- We add automated checks that run after every refresh, such as totals reconciling to the accounting system and row counts within an expected range. If a check fails, the dashboard shows a warning and a named person gets an alert.
- We put a last-updated time and a short definition next to every key figure.
- We remove or merge pages nobody looks at, using the tool's usage data where it exists.
We work with whatever tool you already have if it is sound. Power BI, Tableau and Looker Studio are all capable. The problem is rarely the tool.
A dashboard that earns its place
When a number is questioned, anyone can click through to its definition and see when it was last refreshed. When a feed breaks, the dashboard says so rather than showing a misleading zero. Over time people stop asking for the spreadsheet because the dashboard has stopped surprising them.
It also becomes cheaper to add to. New pages are built on a data model that has already been checked, rather than on another fresh query against raw tables, so each addition inherits the same checks and definitions instead of creating a new way to be wrong.
The finance team also stops maintaining a parallel version, which is often the first real sign that trust has returned.
Is this happening with yours?
- Managers still ask for figures in a spreadsheet or email
- People can name a time the dashboard was wrong
- Nobody knows when the data was last refreshed
- The dashboard has pages nobody can explain
- Two dashboards show different values for the same thing