Paste a sales table or survey results into an AI chat and you'll often get back something like “the data shows some interesting trends” — which says nothing. Usually the data isn't the problem; the prompt is too empty. You never told the AI what the data is, what you want to know, or what the answer should look like.
The “spreadsheet data analyst” prompt below fixes exactly that. Copy it, paste your table in, fill in 3 variables, and you get a structured analysis. It works in ChatGPT, Claude, and other mainstream AI assistants.
What it's for — and where it stops
This prompt is designed for “one table + one clear question”: finding why monthly sales dipped, checking channel performance, distilling survey insights, spotting anomalies in inventory. Two things it can't do: join analysis across multiple tables (merge them into one in Excel first), and precise financial reconciliation — AI arithmetic slips, so always double-check key numbers yourself.
The prompt: copy this block
You are a senior data analyst, skilled at finding business insights in tabular data.
[Analysis goal]
{Analysis goal, e.g.: find the main reasons behind this quarter's sales decline}
[Data background]
{Data background, e.g.: sales details for 12 stores in East China, Jan-Jun}
[Data]
Paste the table here (copy-paste directly from Excel, or export as CSV first). Remove private columns such as names and phone numbers before pasting.
Follow these steps in your analysis:
1. Data check: restate the table structure as you understand it (rows, columns, key fields, time range) and confirm before continuing.
2. Key findings: give 3-5 findings, each with “observation + data evidence + likely cause”.
3. Anomaly alerts: flag obviously abnormal or missing values and explain their possible impact.
4. Action items: give 3 concrete, executable next steps based on the findings — not just “keep observing”.
5. Follow-up questions: list what additional data would make the analysis more accurate.
Output: present in English with bullet points, bold the key numbers, and cite which column each conclusion is based on.The 3 variables: what to fill in
| Variable | What to write | Example |
|---|---|---|
| {Analysis goal} | One sentence on what you want to know — the more specific, the better | Find why East China sales declined month-over-month in June |
| {Data background} | Where the data comes from, who it covers, what period | Sales details for 12 East China stores, Jan-Jun |
| [Data] | The table itself — select and copy in Excel | Paste directly, keep the header row |
The more specific the goal, the less generic the answer. “Analyze this data” and “find the 3 product categories with the highest return rate and hypothesize why” get you two completely different answers.
5 tips to get more out of it
- Make the AI restate the table structure first. That's what step 1 is for: if it misunderstands a column, everything after is wrong, and this catches it early.
- Slice big datasets. Don't paste thousands of rows at once: filter to the period or category you care about, or work in rounds — conclusions first, then deep dives.
- Delete private columns before pasting. Names and phone numbers don't help the analysis; removing them is safer and saves tokens.
- Verify anomaly conclusions. AI is sharp at spotting outliers, but the “likely cause” part is a guess — check key numbers against the original table before reporting.
- Follow up when it's shallow. If the first pass is thin, ask: “go one level deeper on finding 2 and suggest how to verify it against the data background” — round two is usually much better.
Haven't picked which AI to analyze tables with? This on-site comparison walks through who ChatGPT, Claude, Gemini, and Kimi each suit best: Which AI tool saves more time for spreadsheet analysis and data cleanup — read it first, then decide where to paste this prompt.
Q: Can this prompt handle open-ended survey responses? A: Yes, but first put the open-ended answers in their own column, and state in the analysis goal: “summarize themes from the open-ended responses.” Text summarization and numeric analysis are different jobs — a clear goal gets better results.
Q: My table has merged cells and the paste comes out messy. What should I do? A: Unmerge the cells and fill in blanks in Excel before copying. The AI reads plain text, and merged cells paste as lots of empty cells, which skews the conclusions.
