[[{“value”:”

Hey AI Breakers,

Most people drown in spreadsheets. But today, we’ll show you how to turn ChatGPT into your personal data analyst with the FIG framework.

With ChatGPT + the FIG framework, you can analyze any dataset like a pro, even if you’ve never touched complex Excel formulas or written a single line of code.

Let’s FIG it out 👇

🎯 Goal

Learn how to use the FIG framework (Frame, Investigate, Goal) with ChatGPT to:

  • Understand messy spreadsheets instantly

  • Spot insights you’d normally miss

  • Present clear, goal-driven findings without technical skills

🛠️ The System

1. Frame (understand your data)

Every dataset is like a locked box, until you frame it properly, you don’t know what’s inside.

This step is about getting instant clarity: What columns exist? What formats are we dealing with? What problems might block us later?

Instead of scrolling endlessly through rows, you’ll have ChatGPT summarize the essentials in minutes.

Step 1: Upload your dataset to ChatGPT.
Step 2: Use this prompt:

You are a data analyst. Look at the attached dataset and:
1. List all column names. 
2. Show 1 sample value from each column. 
3. Briefly explain what type of data each column likely represents (e.g., text, number, category, date).

Step 3: Confirm formats and spot inconsistencies with:

Take 5 random rows from the dataset. For each row, display the values in every column.
 
Explain any patterns or inconsistencies you notice across the samples.

Step 4: Run a quick quality check:

Run a data quality check on each column. For each one, tell me:
- % of missing or empty values
- Any invalid or unexpected formats
- Outliers or suspicious values
- Potential issues that might affect analysis

💡 Pro Tip: If a column is missing 90%+ of values, skip it in your analysis.

2. Investigate (ask better questions)

Once you know what’s in the box, it’s time to shake it up. Investigation is about curiosity: What can this data actually tell us?

Here you’ll push ChatGPT to brainstorm insights, spot gaps, and even test if the dataset is strong enough to answer real-world questions.

Think of this as turning data from “numbers in a table” into potential stories, trends, and strategies.

Step 1:

Based on this dataset, list 10 interesting business or research questions we could answer.

For each question, explain why it could be valuable to explore.

Step 2: Pressure test the best ones:

For the top 3 questions, explain:
- Which columns are needed
- Whether the current data is sufficient
- Any data cleaning or transformations required before analysis

Step 3: Find the limits:

What are 5 meaningful questions someone might want to ask about this dataset that we cannot answer because the necessary data is missing? 

For each, explain what additional data would be required.

💡 Pro Tip: If you have a second dataset (e.g. cost or viewership), upload it and ask:

I just uploaded a second dataset. Compare it to the first one and:
1. Identify if there’s a common key column we can use to join them.
2. Suggest at least 3 new questions we could answer once they are combined.
3. Recommend the best join method (inner, left, etc.) for analysis.

3. Goal (get to the real answer)


Read more

“}]] Read More in  The AI Break