The prompt
Score 10 of 10, Excellent
Use it when You need the questions it answers, the metrics, charts and filters agreed before building.
- {{audience_and_decisions}}
- {{data_sources}}
- {{current_reporting}}
- {{tool}}
Full prompt and why it works
Prompt
Who will use the dashboard, how often, and what they decide with it: """ {{audience_and_decisions}} """ Data available: each source with its columns, a few sample rows and how often it updates: """ {{data_sources}} """ What they use today instead (reports, spreadsheets, questions they send me): """ {{current_reporting}} """ Tool it will be built in: {{tool}} Task: Write a one-page dashboard plan that the users and I can agree on before I build anything in {{tool}}. The goal is a dashboard people open because it answers their questions, not one that shows every number available. Sections: 1. Users and decisions: who uses it and the decisions it supports, in up to 4 bullets. 2. Questions it answers: 4 to 8 questions in the users' words, each tied to a decision. 3. Metrics: a table with the columns Question, Metric, Definition, Source columns, Compared with, Refresh. Use only columns that exist in my data. A metric I cannot build from them goes in a "Data gaps" list. 4. Layout: a text wireframe from top to bottom. Headline numbers with their comparison on the first row, trends next, breakdowns after that, and a detail table last. No more than 7 charts on the first screen. 5. Charts: for each one, the chart type and why it fits its question. 6. Filters: only those the users need for their decisions, with default values. 7. Out of scope for the first version. 8. Questions to settle with the users before building, up to 5. Requirements: - Every chart answers one of the questions in section 2. If a metric answers none, leave it out. - Every number should be shown next to a comparison (target, last period or same period last year), because a number alone does not tell anyone whether to act. - Avoid pie charts with more than four slices, dual axes and 3D charts. - Propose only what {{tool}} can do. If you are not sure it supports a particular chart or filter, say so. Keep it under 600 words. If the users' decisions are vague, list the questions I should ask them rather than guessing what they need.
Variables
- {{audience_and_decisions}}
- Who will use the dashboard, how often they will look at it, and what they decide with it.
- {{data_sources}}
- Each data source with its column headers, a few sample rows and how often it updates. Remove personal data.
- {{current_reporting}}
- The reports, spreadsheets or recurring questions the dashboard should replace. Write "none" if there are none.
- {{tool}}
- For example Power BI, Tableau, Looker Studio or Excel.
Why it works
- Who it's for. Starting from the users' decisions means every chart has to answer a question someone actually asks, which keeps the dashboard small.
- What good looks like. Every number is shown against a target or an earlier period, because a figure with no comparison does not tell anyone whether to act.
- Length. No more than seven charts on the first screen and an out-of-scope list keep the first version buildable.