How to use ChatGPT for data analysis and competitive research

Open your data in Google Sheets or Excel.

Headers: clean, descriptive, no spaces or special chars. "customer_email" not "Customer Email (verified)."

No merged cells. Each row = one record, each column = one field.

Remove sensitive PII if you can: full names, full addresses, payment data. Hash or anonymize IDs.

Export as CSV (or .xlsx). File should be under 100MB for ChatGPT to handle.

Save the file. You will upload to ChatGPT in the next step.

Open a new ChatGPT chat. Click the paperclip icon (top of input box).

Upload your CSV/Excel file.

ChatGPT processes the file and confirms: "I have loaded the file. It has X rows, Y columns. The columns are: [list]. Sample rows: [shows 3-5]."

Verify the parsing is correct: are column names right? Are data types correct (numbers as numbers, dates as dates)?

If parsing is wrong (e.g., dates parsed as strings), tell ChatGPT: "Re-parse the date column as datetime, not string." It will retry.

Example questions for ad spend data: "What is the average CPA by campaign? Show top 5 by total spend." "Which campaigns have CPA more than 2x the average?"

Example for customer data: "What is the LTV distribution? Show me the 80/20 — what percent of customers drive 80% of revenue?"

Example for behavior data: "Group sessions by traffic source. Average session duration by source. Identify any source with abnormally short sessions."

ChatGPT writes Python code, runs it, returns the result + a chart.

Verify the answer: does it match your intuition? If not, push: "Walk me through the calculation step-by-step."

After getting a text answer, prompt: "Plot this as a bar chart, sorted descending by [metric]."

ChatGPT generates the chart inline. You can download as PNG.

For comparison: "Plot CPA over time for the top 5 campaigns, one line each." Time-series visualization makes trends obvious.

For distributions: "Plot a histogram of session duration. Highlight the median and 95th percentile."

Visualizations often surface insights text answers miss.

Open a new ChatGPT chat. Make sure the model is GPT-4o or higher with browsing enabled (visible in the model selector).

Prompt: "Visit [competitor URL] and summarize their pricing model. Compare to typical pricing in the [category] space."

ChatGPT browses the URL, extracts the data, and analyzes.

For broader research: "Search for recent benchmarks on [metric] in [industry]. Summarize the top 5 sources and the consensus number."

Always cite sources: "List the URLs you used so I can verify."

For every finding, ask: "Does this match what I already know about the business?"

If ChatGPT says "Campaign X has the highest ROAS" but you know Campaign X had a tracking issue last month, the analysis is wrong.

Validation tactics: cross-check 1-2 numbers manually in the source data, compare against last period's known values, run the same analysis in a different tool (Excel, Looker) and compare.

AI analysis is a starting point, not the final answer. Validate before acting on findings.

After analysis is complete, prompt: "Summarize the top 5 findings from this analysis. Format as a short report with: executive summary, key findings (with numbers), recommended actions."

ChatGPT formats the output.

For visual reports: "Export the summary as a downloadable PDF with the charts inline."

For Slack/team sharing: "Format as a brief Slack message — 3-5 bullet findings, with the key numbers."

Reuse the chat as the audit trail — if anyone questions the analysis, you can show the prompts and the code ChatGPT ran.