The prompt
Score 10 of 10, Excellent
Use it when You need trends, outliers and the main takeaways from a spreadsheet in plain language.
- {{data}}
- {{column_notes}}
- {{question}}
- {{audience}}
Full prompt and why it works
Prompt
The data (pasted below, or attached as a file): """ {{data}} """ What the columns mean, with units and anything unusual (for example, "amounts are in thousands" or "the last row is a total"): """ {{column_notes}} """ What I want to know: {{question}} Who the findings are for: {{audience}} Task: Analyze this data and explain what it shows in plain language, for {{audience}}. The goal is findings they can act on and that I can defend if someone checks the numbers. Work in this order: 1. Describe the data first: how many rows and columns you read, what one row represents, the date range, and any quality problems, such as blank cells, duplicates, text in number columns, or subtotal and total rows mixed in with the data, which would count things twice. If the data seems cut off, say how many rows you could see. 2. Answer my question with 3 to 5 findings. For each: one plain sentence stating the finding, the numbers behind it, and the calculation on one line, for example "North: 412 of 1,030 orders = 40%". 3. Outliers: values far from the rest, in a table with the columns Row or item, Value, Typical range, Possible reason (or "unknown"). 4. Caveats: what the data cannot tell us, such as small groups, missing periods or missing columns. 5. Next steps: up to 3 further cuts of the data, or charts, that would answer the question better. Rules: - Work only from the data above. Do not bring in outside benchmarks, and describe patterns as patterns, not as causes the data does not show. - Give every percentage with its count, so small groups are visible. - If you can run calculations on an attached file, do so and say that you did. If you are reading pasted text, recalculate the key totals and show them, so I can check them against the sheet. - Leave out findings that do not bear on my question, however interesting. Keep the whole answer under 500 words plus the table. If my question cannot be answered from these columns, say which column or data would be needed instead of answering a different question.
Variables
- {{data}}
- The spreadsheet data with its header row, pasted as text or CSV, or attach the file. Remove names and contact details first.
- {{column_notes}}
- What each column means, its units, and anything odd, such as total rows or amounts in thousands.
- {{question}}
- What you want to find out, for example "which products are growing and which are slipping this year".
- {{audience}}
- Who will read the findings, for example "my manager" or "the weekly operations meeting".
Why it works
- Background facts. Describing the data first catches subtotal rows, duplicates and cut-off pastes, which quietly distort every number that follows.
- What good looks like. Each finding comes with its numbers and a one-line calculation, so you can defend it when someone checks the sheet.
- If unsure. If the columns cannot answer your question, it names the data that would, instead of answering a different question.