Upload real data from your work and ask ChatGPT to find trends, calculate summaries, create charts, and generate 3 actionable insights. You'll also learn when to trust the analysis and when to double-check it.
Advanced Data Analysis Mode
ChatGPT's "Advanced Data Analysis" (formerly Code Interpreter) is the feature that makes this possible. When enabled, ChatGPT can write and execute Python code in a sandbox to process your data. You never see the code — you just see the results.
How to enable it: In a new ChatGPT conversation, click the "+" icon to attach a file, or drag and drop a CSV/Excel file directly. ChatGPT will automatically switch to analysis mode when you upload data and ask analytical questions. Requires Plus or the free tier's file upload feature (check your account — OpenAI has been expanding this).
What file formats work
- CSV (comma-separated values) — works best
- Excel (.xlsx, .xls)
- Google Sheets — export as CSV first (File → Download → CSV)
- JSON — works, but CSV is cleaner
Privacy reminder: You're uploading data to OpenAI's servers. Do not upload files containing personal health data, financial records with PII, confidential client data, or anything your employer would consider proprietary. Use anonymized or sample data when possible.
The Analysis Prompt Sequence
Don't just upload a file and ask a vague question. Use this sequence for maximum value:
I've uploaded a CSV file containing [describe what's in it].
The columns are: [list the main columns].
First, give me:
1. A summary of what's in this dataset (row count, date range, any issues)
2. The 3 most interesting things you notice immediately
3. What questions this data could answer# Trend analysis
Show me [metric] over time. Is there a trend?
Highlight any anomalies or outliers.
# Segmentation
Break down [metric] by [category column].
Which segment is performing best and worst?
# Correlation
Is there a relationship between [column A] and [column B]?
Show a scatter plot if helpful.
# Summary statistics
Give me the mean, median, min, max, and standard deviation for [column].
Flag anything that looks unusual.Based on everything you've analyzed, give me 3 actionable insights.
For each insight:
1. What you found (specific numbers)
2. Why it matters
3. What I should do about it
Be specific. Avoid generic statements like "you should improve performance."Create a [bar chart / line chart / pie chart] showing [what].
Make it clean and presentation-ready.
Title: "[your title]"
X axis: [label]
Y axis: [label]
Export as PNG.When to Trust the Analysis — and When to Check
ChatGPT's data analysis is impressive but not infallible. Here's an honest map of where it's reliable vs. where you should verify:
| Reliable | Verify manually |
|---|---|
| Row counts, basic statistics | Complex date calculations and time zones |
| Sorting and ranking | Statistical claims like "significant correlation" |
| Filtering and segmentation | Causal claims ("X caused Y") |
| Generating charts and visuals | Forecasts and projections |
| Summary tables | Any output that drives a major decision |
The rule: use ChatGPT to find the story in your data. Verify the specific numbers before you present them to anyone.
Analyze your own real data
- Find a CSV or spreadsheet from your work — sales data, website analytics, expenses, survey results, anything with at least 50 rows
- Upload it and run the 4-prompt sequence above
- Generate at least one chart you could use in a presentation
- Get 3 actionable insights — verify at least one of them manually
Day 3 Complete
- Know how to upload data and use Advanced Data Analysis mode
- Have a 4-prompt sequence that extracts maximum value from any dataset
- Can request specific charts in a presentation-ready format
- Understand where ChatGPT analysis is reliable and where to verify manually
Go beyond ChatGPT for data
Our bootcamp teaches real data engineering: connecting to live databases, building automated reports, and deploying AI-powered dashboards.
See Bootcamp Details — $1,490Ready for Day 4?
Tomorrow: ChatGPT plus DALL-E, browsing, and code interpreter — and an honest comparison with Claude and Gemini.
Day 4: The AI Stack