Dashboards fail for a boring reason: nobody agreed what “utilization” or “margin” meant before the charts were built. Useful BI starts with definitions and a decision someone needs to make, then works backwards to the data.

4–6metrics is enough for a leadership dashboard; more becomes wallpaper
2 sourcesof truth for the same number is one too many
Weeklyis the cadence where operational dashboards change behaviour

Define before you design

Write a one-line definition for each metric, including its source system and its owner. “Pipeline value” from CRM, owned by sales lead. “Estimate turnaround” from intake timestamp to sent quote, owned by estimating lead. If two people disagree on the definition, that conversation is the real work.

Pick the decisions, then the charts

Ask: what will we do differently if this number moves? If there is no answer, drop the chart. The best dashboards are small, boring and looked at every Monday.

  • Pipeline and win rate: where to focus sales effort
  • Estimate turnaround: where to add capacity or automation
  • Utilization and margin by project type: what work to pursue
  • Open requests and ageing: where service is slipping

Trust is the feature

One wrong number kills adoption for months. Invest in data quality at the source: required fields, integration instead of re-keying, and a visible “last updated” on every view.

Real-world example

Metal building fabricator, sales and production

Before

Sales and production reported different revenue numbers to the owner every month.

After

Shared metric definitions, CRM to ERP integration, one weekly dashboard. The monthly reconciliation meeting disappeared.

Key takeaway

BI is an agreement about what the numbers mean, made visible. Define, decide, then build small.