Almost every organization we work with has one report that people quietly work around. The numbers go up on a screen in the Monday meeting, and afterward someone asks finance for "the real figure." That doesn't mean anyone was lazy. It usually means the report once disagreed with a number people trusted, and nobody ever found out why.
You don't need to know SQL to check whether a dashboard is trustworthy. You need a handful of questions and an hour with the person who owns the report. Here's what we look at when a client asks us to review theirs.
Can you tie one number back to the source?
Pick a single figure you already know well, such as last month's revenue, patient visits, or units shipped. Then find the same figure in the system it comes from, whether that's the accounting ledger, the EHR, or the ERP.
If they match, good. If they don't, ask whether anyone can explain the gap. A difference with a clear reason is fine. "The dashboard excludes refunds and the ledger doesn't" is an answer.
"We're not sure" is where most trust problems begin.
Does the report tell you when it was last updated?
A dashboard that looks current but is running on data from three days ago is worse than one that openly says it's three days old. Look for a "last refreshed" date and time somewhere on the page. If there isn't one, ask for it. It's usually a small change, and it saves a lot of arguments.
Also check that the refresh actually worked. Scheduled refreshes fail quietly more often than people expect, and the report keeps showing the last good version as if nothing happened.
Is there a written definition for each key number?
Ask three people what "active customer" means, and you may get three answers. One counts anyone who bought in the last year. Another counts anyone with an open account. A third leaves out wholesale.
A trustworthy report has those definitions written down somewhere people can find them, ideally one click away from the number itself.
If the definition only exists in the head of the analyst who built it, the report is one resignation away from becoming a mystery.
Do the totals hold up when you filter?
Set the filters to a month you know well and look at the total. Then break it down by region, department, or product, and add the pieces back up. They should equal the total.
When they don't, it usually points to records with no region assigned, duplicate rows, or a filter that's quietly set by default. That last one catches people often. Someone sets a filter during testing, saves the report, and every viewer since has been looking at a slice without knowing it.
Can you see the records behind a number?
If a figure looks wrong, you should be able to click into it and see the rows that make it up. Most Power BI and Tableau reports can be set up this way. Being able to check for yourself matters. People trust a number more when they know they could look underneath it, even if they rarely do.
Does someone own it?
Every report that people make decisions from should have a named owner. That person answers questions, approves changes, and gets told when something breaks. Put the name on the report.
Reports with no owner drift. Source systems change, a field gets renamed, and nobody notices until a board meeting.
What to do if your report fails a few of these
Don't throw it away yet. In our experience, most dashboards that fail these checks have fixable problems: a missing definition, a broken refresh, a join that drops records. Fixing the foundation is usually faster than rebuilding the report, and people are more likely to come back when the report they already know starts giving them the right answers.
If you'd like a second pair of eyes, send us the report and the one number you trust least. We'll trace it back to the source and tell you what we find.
