Dashboards People Actually Open

The typical marketing dashboard is built by asking what data is available. It ends up with thirty tiles, which means it has no argument, and people stop opening it within a month. A dashboard is a communication artefact, and it should be designed like one.
Design from the decision backwards
Ask who looks at this, what they decide, and how often. A weekly channel-budget decision needs spend, efficiency, and saturation signals for each channel — perhaps six numbers. An executive monthly review needs revenue, blended acquisition cost, and growth against plan. Those are different dashboards, not different tabs of the same one.
| Audience | Cadence | Core content |
|---|---|---|
| Executive | Monthly | Revenue, blended CAC, payback, plan variance |
| Channel owner | Weekly | Spend, efficiency, volume, pacing by campaign |
| Analyst | Ad hoc | Raw exploration, not a dashboard at all |
A number alone says nothing
Every figure needs a comparison: prior period, same period last year, or target. Without one, the reader cannot tell whether to act. Add a sparkline or a simple variance indicator and the same tile becomes interpretable at a glance.
The question a good tile answers is not "what is it?" but "is that normal?"
Definitions live in the dashboard
Put the definition of each metric behind a tooltip on the tile. Separate documentation is never read, and the recurring meeting argument about whether a number includes refunds costs more time than writing the tooltip ever would.
Annotate the history
Mark significant events directly on the time series: a pricing change, a site migration, a major campaign launch, a tracking fix. Six months later, that annotation is the difference between understanding a step change and inventing a story about it.
Related articles

The Quarterly Data Quality Audit
Most bad decisions made from data are not made from the wrong analysis. They are made from the right analysis on broken inputs.

Incrementality Testing: The Only Number Your CFO Should Trust
Platform-reported ROAS answers "who touched the sale?" Incrementality answers "would it have happened anyway?" Only the second one supports a budget decision.

Predicting Lifetime Value Without Fooling Yourself
An LTV number that nobody can falsify is not a forecast. It is a permission slip for overspending.