Notes on building a dashboard people can trust

This is a sample note. It is here to show the intended format and will change as real examples are added.

A dashboard earns trust before anyone looks at its colors. Trust begins with a shared definition of the question, continues through the data model, and becomes visible in every label, filter, and comparison.

Begin with the decision

“Show me sales” is not yet a useful requirement. Ask what decision the reader needs to make. Are they allocating budget, investigating a missed target, or choosing which customer segment needs attention? The answer determines the measures, grain, and comparisons that belong on the page.

Write the decision at the top of the working brief. If a chart does not help with that decision, it probably does not need to be in the first view.

Make the semantic layer boring

Reliable reporting depends on deliberately unexciting foundations:

• one definition for each important metric;
• explicit relationships and business grain;
• clear handling of missing or late-arriving data;
• tests for totals, uniqueness, and expected ranges;
• a visible “last refreshed” timestamp.

When those rules live in a shared model instead of individual visuals, the dashboard becomes easier to explain and harder to accidentally contradict.

Design for comparison

A number without context creates more questions than it answers. Pair important values with the comparison that makes them meaningful: target, prior period, forecast, or a relevant peer group. Use emphasis sparingly so the viewer can distinguish a signal from decoration.

The result should feel calm. A trusted dashboard does not try to prove how much data it contains. It makes the next useful question obvious.