Dashboard design: start with the decision, then choose the chart

Write a decision brief

Choose one audience and one review rhythm. A sales manager allocating follow-up work needs a different view from a director reviewing the commercial model. Combining both in one crowded screen usually gives each person too much information and too little context.

For each proposed measure, complete this sentence: when this changes, this person will investigate this question and consider this action. If nobody can finish the sentence, the measure may belong in supporting analysis rather than the main dashboard. NIST’s discussion of performance measurement warns about overwhelming decision makers with too many top-level metrics. A smaller set with clear responsibilities is a useful starting point.

Show comparison and context

A number needs a reference. Compare it with an appropriate previous period, a meaningful target or a comparable group. Explain whether the period is complete and whether the definition changed. An apparent fall in weekly enquiries may simply reflect a shorter working week or delayed processing.

Separate an observed change from the explanation for it. The dashboard can show that qualified opportunities fell; it should not automatically declare that marketing failed. Provide a route to examine source, segment, timing and qualification behaviour. This preserves the dashboard’s usefulness without turning a convenient visual pattern into an unsupported conclusion about cause.

A hypothetical sales dashboard

Imagine a service business where managers receive a daily chart of new leads. The chart looks healthy, yet proposals are slowing down. A decision-led review might add unassigned enquiries, time awaiting a first response and opportunities without an agreed next step. These measures point towards work that somebody can inspect.

The director still needs a broader view of pipeline and completed sales, but the team’s daily screen can remain operational. No fictional improvement percentage is needed to justify the design. The test is whether the manager can identify a meaningful exception, assign a response and later see whether the issue was resolved.

Make uncertainty visible

Show the last successful refresh and distinguish missing data from zero activity. If a source is unavailable, a blank tile should not quietly look like a commercial collapse. Where an estimate is used, label it and explain the basis in accessible supporting notes.

Google Analytics documents different data-freshness intervals, illustrating why “today” does not always mean fully processed data. More generally, choose a refresh rhythm that matches the decision. A weekly planning dashboard may not need constant updates, while an operational queue may. Faster refreshes add little value if the team only acts monthly or the underlying information arrives late.

Review every tile before building

Use a paper sketch or simple table before investing in a finished dashboard. Ask the intended owner to talk through a plausible good week and a difficult week. The questions they ask will reveal missing context and unnecessary detail.

• What decision does this measure support?
• Who is responsible for investigating an unexpected change?
• What is the agreed definition and time period?
• What comparison makes the number meaningful?
• How fresh and complete is the source?
• Can the reader investigate an exception without exporting everything?
• What should happen when the source fails or a definition changes?
• Which information can move into a secondary view?

Judge the report by the work it enables

Track whether the dashboard is used in real reviews and whether decisions are recorded. Ask which questions still require manual investigation and which tiles nobody references. Remove redundant measures before adding another row. A dashboard that keeps expanding may be compensating for an unclear management process.

There is a legitimate tradeoff between consistency and flexibility. Stable definitions support comparison, while new business questions require adaptation. Keep a controlled core and a separate space for exploratory analysis. Review the core periodically with its owners. The objective is a dependable decision surface: clear enough to use quickly, detailed enough to investigate and honest about the limits of its data.