Asking the Right Questions
Framing data questions so AI gives you real answers
What You'll Learn
- Why vague questions produce vague answers
- The SCOPE framework for data questions
- How to chain questions for deeper insight
- Common question mistakes and how to fix them
Garbage In, Garbage Out
The number one reason people get disappointing results from AI data analysis? Their questions are too vague. "Analyze this data" is like telling a chef "make food." You'll get something, but it probably won't be what you wanted.
AI is incredibly capable, but it needs direction. The more specific your question, the more specific — and useful — the answer.
The SCOPE Method
Use SCOPE to frame every data question you ask AI:
S — Specific: What exactly do you want to know? Not "how are sales" but "which product category grew fastest in Q3."
C — Context: What does AI need to know about this data? Industry, time period, what the columns mean.
O — Output: What format do you want? A summary, a table, a list of recommendations, a comparison?
P — Perspective: Who is this analysis for? Your boss, your team, yourself? This shapes the depth and language.
E — Edge cases: Are there things AI should watch for? Outliers to ignore? Special circumstances?
Transforming Weak Questions
Weak: "What do you see in this data?"
Too open-ended. AI will give you a generic summary.
Strong: "This is 6 months of email campaign data. Compare open rates across the 4 campaign types. Which type consistently outperforms? Are there any months where the pattern breaks, and why might that be?"
Specific, contextual, asks for comparison AND explanation.
Question Chaining
The best analysts don't ask one question — they chain them. Start broad, then drill down based on what you find.
Round 1: "Summarize the key trends in this customer data."
Round 2: "You mentioned churn spikes in March. Break down the March churners by plan type and tenure."
Round 3: "For the monthly-plan customers who churned in March, what was their average usage in the 30 days before cancellation?"
Each question builds on the last. By round 3, you've found something genuinely actionable.
Try It Yourself
Take the same dataset from Lesson 1. This time, use the SCOPE method to write three increasingly specific questions. Feed them to Claude one at a time and watch how the insights deepen.
I have [describe data, time period, what columns mean]. I need to understand [specific question]. Please present the answer as [format] and flag any unusual patterns. This is for [audience].The SCOPE Framework
Match the Question Quality
Weak vs. Strong Data Questions
Tap one on the left, then its match on the right