Spreadsheet analysis prompt
Get a reliable first analysis of a CSV or sheet: data profile, stated assumptions, answers with working, and the formulas to reproduce them.
Fill in your details
Values are inserted into every prompt below. Nothing you type leaves your browser.
4 field(s) left — unfilled variables stay as {{placeholders}}.
1Upload the file and describe the columns
Upload the file where the tool supports it rather than pasting. For pasted data, keep a header row. Describe units and codes the model cannot guess (e.g. 'amount is in pence').
2Run the analysis prompt
The profiling step catches most bad answers early: duplicate rows, blank values and mixed units.
Analyse the attached or pasted data to answer this business question: {{business_question}} Period covered: {{period}} Column notes: {{data_description}} Data (if not attached): {{data}} Work in this order: 1. Profile: row count, columns and types, missing values per column, duplicate rows, obvious outliers and any unit or format inconsistencies. Stop and ask me if something makes the question unanswerable. 2. Assumptions: list every assumption you are making (e.g. how you treat refunds, blanks or duplicates). 3. Analysis: answer the question. If you can run code, use it and report exact results; if you cannot, say so and limit yourself to what can be computed reliably by hand. 4. Reproduce: give the Excel/Google Sheets formulas or pivot-table setup that would reproduce each key number. 5. Chart: recommend one chart that best answers the question (type, x, y, grouping) and why. 6. Caveats: what this data cannot tell us.
Example input
Question: which product lines drove the drop in repeat purchases this quarter? File: orders_q3.csv (18,402 rows).
Example output
1. Profile 18,402 rows, 7 columns. 212 rows have blank customer_id (guest checkouts). 37 exact duplicate rows. order_date mixes DD/MM/YYYY and YYYY-MM-DD in 1,104 rows (normalised). 2. Assumptions Guest orders excluded from repeat-rate calculations; refunds (is_refund=Y) excluded; duplicates removed. 3. Analysis Repeat rate fell from 31.4% (Q2) to 26.9% (Q3). 71% of the drop is in "Home Fragrance"... 4. Reproduce =COUNTIFS(line_range,"Home Fragrance",repeat_flag_range,TRUE)/COUNTIFS(line_range,"Home Fragrance")
Troubleshooting
- Numbers change between runs: the model is estimating rather than computing. Use a tool with code execution, or ask it to show the code it ran.
- Dates are misread: state the date format in the column notes.
- Large files are truncated when pasted: upload instead, or pre-aggregate in your spreadsheet.
Known limitations
- Without code execution, LLM arithmetic on more than a handful of rows is unreliable.
- Correlation in the output is not causation; the model may present patterns more confidently than the data supports.
- Remove personal data you do not need before uploading.