The most common reason a spreadsheet analysis prompt fails is not that the AI cannot do the math, but that the question is too vague. Paste a sales or expense table and just say "analyze this," and you usually get the headers repeated back to you. The prompt below asks differently: it fixes the analysis goal first, sets the output order, and forces the AI to use only numbers that actually appear in the table. Copy it and it works right away.
Which tables it fits
Monthly sales tables, expense reports, inventory tables, and user feedback tallies all work, as long as the table has clear numeric columns such as amounts, counts, or shares. Pure text notes or lists without numbers do not fit this prompt: the AI can tidy the format, but that is not analysis.
The prompt to copy
You are a data analyst. Analyze the table below: state the conclusion first, then lay out the evidence.
Table data:
[paste your table here]
Analysis goal: [for example: find out why sales fell this month]
Please output in this order:
1. Core conclusion: state the single most important change in one sentence.
2. Key numbers: list the 3 numbers that changed the most, and note the comparison basis (month over month, year over year, or share).
3. Outliers: point out rows or columns that deviate from the normal range, and state how far they deviate.
4. Possible causes: list only causes the table itself supports, and mark guesses as guesses.
5. Next steps: give 2 actions that can be carried out directly.
Requirement: do not invent numbers that are not in the table. If an indicator cannot be computed, say so, and state which field is missing.How to fill in the variables
| Variable | What to put | Example |
|---|---|---|
| [paste your table here] | The whole table, or the filtered rows and columns, keeping the headers | Sales by channel, January to September |
| [Analysis goal] | The one question you actually want answered | Find out why sales fell this month |
| Time range (optional) | Add a sentence after the goal to limit the scope | Look at Q3 data only |
Three ways to paste a table
- Select a range in your spreadsheet app and copy it straight into the chat; the format usually survives.
- If the table is large, export it as CSV first, then paste it, and drop columns unrelated to the analysis to cut noise.
- If column names are abbreviations or internal codes, add one line before the prompt explaining them, for example "GMV means gross merchandise value," so the AI does not have to guess.
Four tweaks that make results more accurate
- Fix the comparison basis: write month over month, year over year, or share into the goal, and the AI will stop jumping between bases.
- Scope first, detail second: if you only want outliers, keep items 1 and 3 and drop the rest, and the answer stays focused.
- Check every number: compare the AI's key numbers against the original table, especially totals and shares, where mistakes are most common.
- One goal per ask: asking for the cause of a decline, growth opportunities, and next year's forecast at once makes all three answers thinner. Ask them separately.
Three traps to avoid
- Merged cells and multi-row headers confuse the AI about row and column relationships. Unmerge them and complete the headers before pasting.
- Goals like "analyze this" or "the more detail the better" hand all judgment to the AI, and the output is bound to be vague.
- Causes the AI marks as guesses must not be treated as conclusions. Check them against how the business actually works before acting on them.
