Skip to content
Orivion

Before the dashboard: can the business trust its numbers?

Dashboards fail when definitions and data quality are left for later. Start underneath.

Author
Orivion editorial
Published
Last reviewed
Reading time
5 min

Written from Orivion's delivery practice.

Conflicting sales, finance and marketing figures reconciled into one reported number

Why this matters

A dashboard is the most visible part of a data project and the least important. If finance, sales and marketing each calculate revenue differently, a better chart simply displays the disagreement more clearly.

Definitions first

Pick the ten metrics leadership actually uses. For each, write the calculation, the source systems, the inclusions and exclusions, and the owner. This single document resolves more reporting arguments than any tool.

Quality rules second

Profile the data behind each metric: duplicates, missing values, inconsistent codes and late updates. Turn the important checks into automated tests that run with every refresh.

One semantic layer

Modern platforms, Microsoft Fabric among them, bring ingestion, storage, transformation, semantic models and reporting into one governed flow. The key idea is a single semantic model that every report uses, so a metric means the same thing everywhere.

Signs the problem is underneath the dashboard

  • Two reports show different totals for the same metric.
  • Month-end reporting needs manual reconciliation.
  • People export data to spreadsheets to check the dashboard.
  • Nobody can say who owns a metric's definition.
  • Pipelines fail without anyone being alerted.

What to do next

Choose one metric that causes arguments. Define it, trace its sources and reconcile it. The process will show you what the rest of the programme needs.

Information on this site is general. It is not legal, tax or financial advice. Requirements change, so confirm current rules with the relevant authority or a licensed adviser.

Want to talk through how this applies to your business?