ChatGPT Prompts for Data Analysis: 6 Practical Tasks

ChatGPT prompts for data analysis can turn an Excel or CSV sales sheet into a cleaner dataset, a useful summary, meaningful comparisons, and clear charts without requiring you to write code. The best results come from asking for one task at a time, explaining how to handle ambiguous records, and checking important figures against the original file.

The examples below follow one sales spreadsheet from preparation to reporting. You can reuse the same prompts in ChatGPT, Claude, or Gemini, making it easier to compare their findings. For more workplace examples, explore our broader collection of AI prompts for work.

Before you start: Prepare your sales spreadsheet

Imagine you have a workbook called Sales_2026.xlsx containing order dates, order IDs, regions, products, quantities, net sales, order statuses, customer names, and customer email addresses. You want to understand which products sell well, which regions perform best, and how sales change over time.

Start by saving a separate copy of the workbook. Remove customer names, email addresses, phone numbers, physical addresses, and other personal details that are not necessary for the analysis. Replace identifying values with anonymous IDs when you need to preserve relationships between records. Remove or generalize any other information that could identify a person.

Do not upload confidential company or customer information to a consumer AI service unless your organization has approved that use. Removing names and emails is a useful first step, but it does not make every dataset anonymous. Follow your company's data-handling rules and use an approved environment for sensitive information.

Next, check whether your spreadsheet has common problems, such as North and north appearing as separate regions, mixed date formats, duplicate order IDs, blank dates, sales amounts stored as text, or cancelled orders mixed with completed orders. These are useful issues to catch before asking for totals or charts.

Keep the workbook's original column names where possible. The prompts below refer to fields such as Order Date, Order ID, Region, Product, Quantity, Net Sales, and Order Status. If your columns have different names, ask the AI to map them to the corresponding fields rather than assuming they are identical.

This guide focuses on analyzing spreadsheet data, not writing formulas. For formula-specific tasks, see our Excel formula examples.

1. Clean the spreadsheet and flag data problems

What it does: This prompt asks the AI to inspect the workbook, identify possible quality issues, and prepare a cleaned copy without hiding questionable records.

Why it matters: An analysis built on inconsistent dates, duplicate orders, or text stored in numeric columns can produce misleading results. Cleaning first also gives you a chance to decide how to handle records that cannot be corrected confidently.

Who it suits: Anyone working with a spreadsheet assembled from online orders, sales reports, exports, or several people entering data in different ways.

Copy-and-paste prompt:

Inspect the attached sales spreadsheet before changing anything. Report the worksheet names, row count, column names, likely data types, duplicate order IDs, missing values, inconsistent date formats, extra spaces, inconsistent Region, Product, and Order Status labels, and Net Sales or Quantity values stored as text. Do not delete records or guess missing values. Separate definite problems from items that need a business decision. Then create a cleaned copy if file editing is available. Standardize dates to YYYY-MM-DD, trim unnecessary spaces, normalize labels only when they clearly mean the same thing, convert Quantity and Net Sales to numeric values, preserve the original file, and provide a change log and an unresolved-issues list. If you cannot create an edited file, return the proposed changes in a table.

Review the issue list before accepting changes. For example, North and north with an extra space probably refer to the same region, but two rows with the same order ID might represent a duplicate order or separate product lines in one order. Ask the AI to flag the uncertainty rather than delete a row automatically.

The cleaned file should preserve the original information needed for analysis. If the assistant cannot generate a revised workbook in your account, use its report to make the changes in your spreadsheet application, then continue with the corrected copy.

2. Summarize the overall sales performance

What it does: Once the data is clean enough to use, this prompt produces a compact overview of the business's sales activity.

Why it matters: Raw spreadsheets are often difficult to interpret at a glance. A well-defined summary helps you see the scale of activity before looking at individual products, regions, or months.

Who it suits: Freelancers, small business owners, e-commerce operators, and sales managers who need a quick report for a weekly or monthly review.

Copy-and-paste prompt:

Using the cleaned sales spreadsheet, calculate total Net Sales, the number of unique orders, total units sold, average order value, and monthly sales totals. For this report, include rows marked Completed or Shipped in the main sales totals, and report Cancelled orders separately. Count unique Order IDs rather than spreadsheet rows when calculating order count. State the date range, currency, filters, and method used for each metric. Show the results in a summary table, followed by three concise observations supported by the data. Flag missing values or unresolved issues that could change the results. Do not invent explanations for sales changes.

The status rule is important. A spreadsheet may include cancelled orders, refunds, pending payments, and completed transactions. The sample prompt includes Completed and Shipped orders in the main report and separates Cancelled orders. Change those rules to match your business's reporting policy before using the output.

Also check how the spreadsheet records revenue. If each row represents one product line, summing rows may be correct for sales, but counting rows will not necessarily give you the number of orders. Asking the AI to state its counting method makes the result easier to audit.

3. Segment sales by region and product

What it does: This prompt breaks the overall figures into smaller groups so you can compare regions, products, and their contributions to sales.

Why it matters: A healthy overall sales total can conceal an underperforming region or a product that accounts for a large share of revenue. Segmentation helps you identify where to investigate further instead of relying on a single company-wide number.

Who it suits: Retailers, online sellers, account managers, and anyone responsible for product or regional performance.

Copy-and-paste prompt:

Using the cleaned sales spreadsheet, segment the results by Region and by Product. For each segment, show total Net Sales, unique Order IDs, total Quantity, and share of total active Net Sales. Use the same Completed and Shipped order rule as the summary. Return separate ranked tables for regions and products, showing the strongest and weakest segments where the data supports a fair comparison. Explain the calculation of sales share, flag segments with missing or inconsistent labels, and identify up to three findings supported by the numbers. Do not infer causes or recommend a business change without explaining what additional evidence would be needed.

Look at revenue and order volume together. A product may rank highly by sales because it has a high price, even if it sells fewer units. Another product may attract many orders but contribute relatively little revenue. Both patterns are worth understanding, and neither metric tells the whole story alone.

Pay attention to the size of each segment as well. If one region has far fewer orders than another, differences in revenue might reflect volume rather than better performance. Ask for the underlying counts so the summary does not turn a small sample into an overly confident conclusion.

4. Find trends and changes over time

What it does: The trend prompt organizes sales by month, compares periods, and highlights changes that deserve a closer look.

Why it matters: A single sales total tells you how much was recorded, but it cannot show whether activity is increasing, declining, or fluctuating. Monthly analysis gives you a consistent timeline for reviewing performance.

Who it suits: Business owners preparing monthly reports, sales teams tracking targets, and freelancers reviewing changes in client or project revenue.

Copy-and-paste prompt:

Analyze the cleaned sales spreadsheet over time. Group active Net Sales, unique orders, and units sold by calendar month using Order Date. Use the same order-status rules as the previous summary. Show a chronological table with each month's totals and the percentage change from the previous month when the calculation is valid. Flag missing dates, incomplete months, months with no recorded transactions, and any gaps in the reporting period. Distinguish zero sales from missing data. Identify notable increases or decreases, cite the figures behind each observation, and do not claim a cause or predict future performance from this data alone.

Check the reporting period before interpreting the trend. A partial month should not be compared directly with a full month without a clear warning. Likewise, a blank date does not mean an order occurred in a month with zero sales.

Month-to-month changes are useful for spotting questions, not automatically proving an explanation. A decline might coincide with fewer orders, lower average order value, a product shortage, or an incomplete export. The spreadsheet can show some of these patterns, but it may not explain the underlying cause.

5. Create a chart that makes the findings easier to understand

What it does: This prompt turns verified summary data into a visual that makes comparisons or changes over time easier to see.

Why it matters: Tables are useful when someone needs exact values. A chart can make a longer report easier to scan, especially when comparing several months or ranking multiple products.

Who it suits: Anyone preparing a presentation, a management report, a business review, or a short update for colleagues.

Copy-and-paste prompt:

Using the verified monthly sales summary from the cleaned spreadsheet, create a line chart of monthly active Net Sales. Put calendar months in chronological order on the horizontal axis and sales in the spreadsheet's currency on the vertical axis. Give the chart a clear title and descriptive axis labels. Use one series, show the complete reporting period, and do not use decorative effects or misleading axis scales. If the data contains incomplete months, gaps, or missing dates, make those limitations clear. Return the chart with the table of values used to produce it. If you cannot generate a chart, return the plotting data in a clean table and recommend the appropriate chart type.

A line chart is a sensible starting point for data measured over time. For comparing product or regional totals, a sorted bar chart is generally easier to read. Keep the visual focused on one question instead of combining unrelated metrics into a single chart.

Compare the chart's values with your summary table. The labels should match the reporting period and currency, and the chart should not imply missing months are zero unless that is what the source data actually confirms. If the visualization looks surprising, verify the underlying numbers before sharing it.

6. Verify the findings before using them

What it does: This final prompt asks the assistant to audit its own results against the cleaned data and state the limitations.

Why it matters: AI tools can misread a file, mishandle a filter, or produce a plausible but incorrect calculation. A confident explanation is not proof that the numbers are correct.

Who it suits: Anyone using the analysis to make a financial, operational, staffing, inventory, or client-facing decision.

Copy-and-paste prompt:

Audit the sales analysis you produced using the cleaned spreadsheet. Recalculate total active Net Sales, unique order count, total units, average order value, and monthly totals from the source columns. Verify the date range, order-status filters, treatment of duplicate IDs, currency, and missing values. Return a verification table containing each metric, the reported result, the independently recalculated result, the method used, and any limitations. Correct discrepancies and explain their cause when it can be established. Clearly label anything you cannot verify. Do not invent supporting evidence or describe an assumption as a confirmed fact.

For important reports, verify totals in Excel, Google Sheets, or your established reporting system. Check a small selection of source rows as well as the grand total. If a discrepancy remains, resolve it before presenting the results as final.

A second AI assistant can provide an additional review, but it is not an independent source of truth if it receives the same incorrect file or repeats the same assumptions. Keep the source spreadsheet and your cleaning notes so another person can follow how the final figures were produced.

Which AI assistant should you use for spreadsheet analysis?

You can reuse the prompts in ChatGPT, Claude, and Gemini. Attach the same anonymized workbook to each service and submit the same prompt when comparing their outputs. Keep the reporting rules identical, or differences in the instructions may explain differences in the results.

These limits are not permanent guarantees, and your account may have additional restrictions. Check the provider's current documentation before a large analysis. For example, a file-size limit does not mean the free plan will allow an unlimited number of analysis requests.

Claude users who are new to the service can also read our guide to Claude's free plan to understand the broader workflow. If you are still deciding which assistant best suits your needs, our assistant comparison guide provides a wider view of the available tools.

You do not need to run every prompt in all three services for routine work. Choose one assistant that accepts your file, use the prompts in sequence, and verify the output. Try a second assistant when you need a comparison or when a particular result is difficult to interpret.