BI in a Box: what to agree before building dashboards

Start with the meeting the dashboard should improve

A dashboard brief is more useful when it names a decision. “Show sales performance” could mean revenue recognition, bookings, margin or sales activity. Each requires different definitions and potentially different source data.

Choose one recurring review and ask what the participants decide, what they dispute and which manual preparation slows them down. This gives the first release a clear job. Additional charts should earn their place by supporting that job.

Create a KPI contract

• Name the measure and the business question it answers.
• Define the calculation, inclusions, exclusions and reporting period.
• Identify the source system and accountable business owner.
• State how late records, cancellations and corrections are handled.
• Agree the refresh target and a way to show when the data is stale.

Set the level of detail before joining data

Microsoft’s Power BI guidance describes fact and dimension tables and stresses consistent grain in fact data. This supports understandable filtering and summarisation. See Microsoft’s star-schema guidance.

Consider a monthly sales target joined to daily order lines. If the target is repeated for every order and then summed, the dashboard can overstate it. The visual may look convincing while the underlying measure is wrong. Specify the grain of each dataset and test the calculation at multiple levels of aggregation.

Use a small acceptance pack

For a first finance or sales dashboard, prepare a short set of known cases: a normal transaction, a cancellation, a late adjustment and a restricted user. Ask the business owner to sign off the expected figures before checking the finished report.

Reconcile totals to a trusted source, then test filters and drill-downs. A correct total can hide incorrect allocations between teams or products. Include access checks and failed-refresh behaviour so users can distinguish missing data from a genuine zero.

Define what comes in the box

Our suggested BI in a Box scope combines a selected source set, a documented metrics layer, agreed dashboard views and an operating handover. It should identify excluded sources and future requests, not imply that every connector or metric is available immediately.

The handover should name who handles refresh failures, who approves new measures and how changes are tested. Without that ownership, a reporting accelerator can become another spreadsheet replacement that nobody trusts enough to use.

Begin with a focused reporting use case. Expand after the team has demonstrated that the numbers reconcile, the definitions are understood and the dashboard actually supports a recurring decision.