How to check a dashboard number before you trust it
A dashboard is only useful if the people reading it believe it. One wrong number in a meeting and it stops being opened.
These are the checks I run on every build. None of them needs special software.
Every figure needs a source you can point to
A number goes on a dashboard only if it traces to one of three things: an export I have re-computed, a screenshot of the source system, or the client's written word.
During discovery on one build, I set aside two automatically generated survey reports because their figures could not be traced to anything.
Reconcile each source before building on it
Before a source feeds any dashboard, total it and compare it with a figure the firm already trusts. If the totals differ, stop and find out why.
Tie each check to the data it checked
A check passes for one load of data, not forever. The clean tables are built only from a load that passed. If a new load fails, the dashboards stay on the last good one.
Refuse partial loads
If only some of the exports for a period have arrived, the load is incomplete. Do not build from it. A dashboard that mixes this month's jobs with last month's bills is worse than one that is a month old.
State the date
Every dashboard shows the date of the data it is built from. That is the date of the last load that passed, not the date someone last pressed refresh.
When a number changes, keep the old one
Sometimes a figure you have already quoted turns out to be wrong. Record the old figure beside the new one, with the reason. Replacing it silently costs more trust than the error did.
Read the finished page
The last check is to open the live dashboard and read every tile. On one build that read caught a unit label on the wrong tiles, so counts were showing as days. No automated check had flagged it.
If you would rather have this built for you, I do it for building and remodeling firms. It starts with a small fixed-price review.