Best for
- The dashboard request that is actually a vague anxiety
- Replacing the 40-chart graveyard with something consulted daily
- Agreeing what 'the metrics dashboard' means before the build argument
What you give it
- Who will use it, the decisions they face, and the metrics that exist (or need defining first)
What you get back
- The spec: per view, the question it answers, the metrics with their definitions linked, the comparisons and thresholds that make numbers readable
- The layout logic: what is glanceable, what is drill-down, what is deliberately absent — with reasons
- The acceptance checklist the built dashboard is tested against: can each named user answer each named question in seconds?
How it works
- Extracts the real questions from the intended users — a dashboard is specified by what its readers need to decide, not by what data is available.
- Specs each view as question-answering: the metric, its comparison (plan, prior period, threshold), its grain and freshness — a number without a comparison is decoration.
- Writes the exclusions with reasons: the spec's discipline is what it leaves out, recorded so the requests do not return quarterly.
- Defines acceptance as reader performance: each named question answerable by its named reader in seconds, tested before launch.
Example
You: Leadership wants 'a business dashboard'. Spec it before we build chart soup.
Result: The interviews turned 'a business dashboard' into five questions three roles actually ask weekly ('are we on plan for the quarter?', 'which channel is moving?', 'is churn stable?'…). The spec: one glanceable top row (four numbers against plan with thresholds that colour only when action-worthy), three drill-down views each owning one question, every metric linked to its canonical definition (two needed defining first — flagged as prerequisites), twelve requested charts explicitly EXCLUDED with reasons recorded ('interesting' is not a question). Built in a week against the spec; still consulted daily a quarter later — the previous dashboard's fate avoided.
Limits — please read
- Metrics without canonical definitions are prerequisites, not dashboard content — the spec names them (a metric-definition pass pairs well).
- A spec cannot force consultation; it maximises the odds by answering real questions with honest freshness.
- Tool specifics (what your BI platform renders) shape the layout section; it works within what you have.