Prompt details

Tested on

Claude Opus 5.5

Length

498 words · 3 fill-ins

Example input

[The decision this data supports]: Our head of operations is deciding whether to keep free returns in the West region next year or bring in the $6.95 return fee that the South region started charging in month 7 (M07). I need a report she can read in five minutes.

[What the data is and how it was collected]: Monthly totals by region from our online store, exported from the order system for the last 12 months (M01 is the oldest month). orders = orders placed; revenue_usd = gross sales before refunds; returned_orders = orders with at least one item returned; refunds_usd = money refunded; return_fees_usd = fees charged to customers for returns (South only, from M07). I pasted it straight from the export and have not cleaned it.

[Your data]: month,region,orders,revenue_usd,returned_orders,refunds_usd,return_fees_usd M01,North,1365,77511.46,153,7607.18,0.00 M01,South,1567,83936.09,208,10117.15,0.00 M01,East,1118,67926.49,106,5554.76,0.00 M01,West,1465,94105.01,210,11540.85,0.00 M02,North,1330,79212.18,148,7636.91,0.00 M02,South,1539,80845.10,225,10679.38,0.00 M02,East,1089,64905.62,106,6023.81,0.00 M02,West,1381,87428.97,218,12185.35,0.00 M03,North,1459,82402.14,152,7675.17,0.00 M03,South,1667,89626.92,228,11186.14,0.00 M03,East,1238,74610.73,128,7182.57,0.00 M03,West,1528,96693.86,245,14470.52,0.00 M04,North,1490,85320.41,173,8647.37,0.00 M04,South,1658,90913.34,222,10775.97,0.00 M04,East,1234,76033.74,127,7028.87,0.00 M04,West,1603,99860.43,271,15526.82,0.00 M05,North,1507,87176.32,173,9476.79,0.00 M05,South,1697,92540.55,225,11179.81,0.00 M05,East,1322,83027.85,137,7341.29,0.00 M05,West,1589,101120.00,255,14396.96,0.00 M05,West,1589,101120.00,255,14396.96,0.00 M06,North,1426,80807.85,149,8009.61,0.00 M06,South,1612,85729.82,223,11285.07,0.00 M06,East,1253,76199.97,126,7212.07,0.00 M06,West,1582,101842.61,269,15108.75,0.00 M07,North,1413,83843.17,164,8228.76,0.00 M07,South,1521,80813.07,130,6307.44,903.50 M07,East,1280,76968.52,120,6528.67,0.00 M07,West,1508,95382.16,286,16559.43,0.00 M08,North,1486,867961.10,167,8285.40,0.00 M08,South,1654,90816.35,152,7631.69,1056.40 M08,East,1333,80820.14,127,7076.33,0.00 M08,West,1542,94624.17,275,15007.44,0.00 M09,North,1486,83874.24,154,7727.31,0.00 M09,South,1593,85318.05,132,6682.02,917.40 M09,West,1587,97718.48,314,18779.26,0.00 M10,North,1557,90218.41,163,8132.51,0.00 M10,South,1681,89492.76,154,7202.85,1070.30 M10,East,1399,87648.20,140,7384.20,0.00 M10,West,1668,102101.98,326,19466.95,0.00 M11,North,2023,118715.23,255,13113.85,0.00 M11,South,2111,115853.96,229,11474.48,1591.55 M11,East,1829,109715.00,228,13191.64,0.00 M11,West,2157,138386.60,493,28698.12,0.00 M12,North,2139,124193.30,273,13504.79,0.00 M12,South,2299,122502.94,241,11963.15,1674.95 M12,East,2108,128180.85,266,15395.29,0.00 M12,West,2383,148909.68,518,28479.19,0.00

Example output

1. Data quality check

Check	What I found	How I handled it
Rows and columns	48 data rows and 7 columns. 4 regions × 12 months should give 48 rows, but one row is a duplicate and one month is missing, so there are 47 unique region-months.	All figures below use the 47 unique rows.
Duplicates	M05 West appears twice, with every value identical.	Removed one copy. Keeping it would add $101,120.00 to West's H1 revenue and 255 returned orders. West's H1 return rate would not change (16.05% either way).
Missing rows	M09 East is absent.	East H2 uses 5 months. Comparisons that involve East use the same months for each region (M07, M08, M10).
Missing values	No blank cells in the 47 rows.	None needed.
Impossible or outlying values	M08 North revenue is $867,961.10. The neighbouring months are $83,843.17 and $83,874.24. On 1,486 orders that is $584 per order, against North's usual ~$57 ($492,430.36 ÷ 8,577 in H1 = $57.41). It is almost certainly an entry error.	M08 North revenue and refunds are excluded from revenue measures. Its orders and returns look normal and are kept. Confirm the true value in the order system. This does not affect any West or South finding.
Logical consistency	In every row, returned orders ≤ orders and refunds < revenue. Return fees always equal $6.95 × South returned orders. Example: M07 130 × 6.95 = $903.50. Over the year, 1,038 × 6.95 = $7,214.10, which matches the column total.	No change. This also tells us every returned South order paid the fee, so there are no waivers in the data.
Seasonality	All four regions jump in M11–M12. Orders rise 36–63% over the H1 monthly average.	"After" comparisons use M07–M10 (non-peak) where seasonality would distort the result.

Claude Opus 5.5

Find Insights in Your Data – Data analysis prompt for ChatGPT and Claude

Q: Can ChatGPT analyze data?

Yes, ChatGPT can analyse data, and it is most reliable when it can run code on an uploaded file. Give it the decision the data supports and what each column means, and ask it to check data quality first and show how each number was calculated.

Q: What is the best prompt for data analysis?

The best data analysis prompt names the decision, explains the columns and asks for a quality check first, findings with their base and comparison, and the limits of the data. Asking only for insights tends to return generic observations without numbers you can check.

Q: Can I use this prompt in Copilot?

Yes, the prompt runs as is in Microsoft Copilot, as well as ChatGPT, Claude and Gemini, because it uses plain text and tags. Paste your data or attach the file, and check the headline figures against your spreadsheet.

Q: Can AI find insights in a CSV file?

Yes, AI can find patterns in a CSV file and explain them in plain language. Insights are only as good as the data, so have it report duplicates, gaps and outliers first, and treat any reason it suggests for a pattern as an idea to test.

^$9.99

Kept in your account
Copy filled in or download
Updates included

Pay once and keep it. Or get every premium prompt and skill with the plan, $19.99 a month. Payments are non-refundable. See our Terms.

votes views

Claude Opus 5.5

This prompt analyses a dataset for a decision. Paste the data, what it is and the decision it supports; you get a quality check, summary statistics, findings with numbers, charts to make and a one-page report.

What you get

A data quality check: duplicates, gaps and outliers, and how each was handled
Summary statistics and 3 to 5 findings, each with its figure, comparison and base
The charts to make, and what each one shows
A one-page report for the decision maker, plus caveats and next questions

How to use it

1
Name the decision
Write the decision the data supports, the options, and what each column means. Paste the data with its header row, or upload the file to an assistant that runs code.
2
Read the quality check first
Confirm each fix it made (a dropped duplicate, an excluded outlier) is right before you trust any finding.
3
Check the headline numbers
Recompute two or three headline figures in a spreadsheet, using the arithmetic it shows, before the report goes to anyone.

Better results

Explain each column in plain words, including anything you know is off. "Revenue is before refunds" changes every finding about profit.
Follow up with: "Which finding would change if the outlier you excluded were real?"
For more than a few hundred rows, upload the file to an assistant that runs code instead of pasting it, so every figure is calculated.

Questions

Can ChatGPT analyze data?

Yes, ChatGPT can analyse data, and it is most reliable when it can run code on an uploaded file. Give it the decision the data supports and what each column means, and ask it to check data quality first and show how each number was calculated.

What is the best prompt for data analysis?

The best data analysis prompt names the decision, explains the columns and asks for a quality check first, findings with their base and comparison, and the limits of the data. Asking only for insights tends to return generic observations without numbers you can check.

Can I use this prompt in Copilot?

Yes, the prompt runs as is in Microsoft Copilot, as well as ChatGPT, Claude and Gemini, because it uses plain text and tags. Paste your data or attach the file, and check the headline figures against your spreadsheet.

Can AI find insights in a CSV file?

Yes, AI can find patterns in a CSV file and explain them in plain language. Insights are only as good as the data, so have it report duplicates, gaps and outliers first, and treat any reason it suggests for a pattern as an idea to test.

Updated 8 Nov 2026