Quick Answer
A good data analysis prompt tells ChatGPT what the dataset represents, what decision you need to make, which columns matter, what counts as a useful insight, and how the result should be formatted. Do not ask "analyze this data". Ask for a specific analysis workflow.
This guide is for marketers, founders, operators, students, and analysts who use AI to inspect spreadsheet exports, survey results, sales data, support tickets, product feedback, or campaign reports.
Start With the Decision
Data analysis is not about producing charts. It is about reducing uncertainty. Before prompting ChatGPT, write down the decision you want the analysis to support.
Examples:
- Which marketing channel should we invest in next month?
- Which customer segment has the highest retention risk?
- What are the top reasons users cancel?
- Which products should be bundled together?
- What changed between this month and last month?
When the decision is clear, the prompt can filter noise and focus on useful evidence.
The Data Analysis Prompt Template
You are a careful data analyst. I will provide {{type of data}}. Goal: {{decision or question}}. Columns: {{column names and meanings}}. Please analyze the data in 5 steps: 1) validate the structure and note missing or suspicious values, 2) summarize the most important patterns, 3) identify outliers or segments worth investigating, 4) explain what the patterns may mean for the business, and 5) recommend next actions. Separate observations from interpretations. If the data is not enough to support a conclusion, say so.This template works because it asks the model to inspect data quality before drawing conclusions.
Prompt 1: Clean and Understand a Spreadsheet
Use this before asking for insights.
Review this spreadsheet export. Explain what each column appears to represent, identify missing values, inconsistent formats, duplicate rows, unusual values, and columns that should not be used for analysis. Return a table with Issue, Why It Matters, Example, and Fix.This prevents a common mistake: asking for strategy from messy data.
Prompt 2: Find Patterns in Customer Feedback
For survey responses, reviews, or support tickets, ask for themes and evidence.
Analyze these customer feedback entries. Group them into 5-8 themes. For each theme, include: theme name, frequency estimate, representative quotes, customer pain, likely root cause, and recommended action. Do not merge positive and negative feedback into the same theme unless they describe the same issue.If the dataset is large, analyze a sample first, define categories, then apply the categories to the full dataset.
Prompt 3: Compare Two Time Periods
Month-over-month analysis needs context, not just percentages.
Compare Period A and Period B. Identify the biggest increases, decreases, and stable metrics. For each change, calculate absolute difference and percentage difference if possible. Explain whether the change is likely meaningful or may be noise. End with the 3 questions we should investigate next.Ask for absolute and percentage differences because percentages alone can exaggerate small numbers.
Prompt 4: Turn Analysis Into an Executive Summary
Once you have findings, turn them into a decision-ready memo.
Convert this analysis into an executive summary for {{audience}}. Start with the recommendation. Then include 3 supporting insights, 2 risks or caveats, and the next decision needed. Keep it under 300 words and avoid technical jargon.This is useful when the reader does not need to see every calculation.
Prompt 5: Generate Follow-Up Questions
Good analysis creates better questions.
Based on this dataset and analysis, list the 10 highest-value follow-up questions. For each question, explain what additional data would answer it and what decision it would improve.This helps you avoid treating the first analysis as the final answer.
Accuracy Rules for AI Data Work
AI tools can misread pasted tables, lose rows, or make calculation mistakes. Use these safeguards:
- Provide a clean CSV or clear table when possible.
- Ask the model to restate column names before analysis.
- Ask for formulas or calculation logic for important numbers.
- Verify critical calculations in a spreadsheet.
- Do not paste private customer data, passwords, payment data, or confidential records into tools that are not approved for that data.
Internal Links for Related Prompts
Browse data analysis prompts for practical templates. For structured prompt design, read Prompt Frameworks Explained. For iterative improvement, use Prompt Iteration.
Final Takeaway
The strongest data prompt starts with the decision, checks the data, separates observations from interpretation, and ends with next actions. ChatGPT can speed up analysis, but important numbers still deserve human verification.