Asking Better Questions
Frame analytical questions that lead to actionable insights. The question determines the quality of the answer.
Premium Course Content
This lesson is part of a premium course. Upgrade to Pro to unlock all premium courses and content.
- Access all premium courses
- 1000+ AI skills included
- New content added weekly
The Question Problem
The biggest mistake in data analysis happens before you touch the data.
Someone says “analyze the sales data” or “look at customer behavior.” These aren’t questions—they’re fishing expeditions.
Without a clear question, you’ll wander through the data, find random patterns, and produce a report that doesn’t help anyone make decisions.
Good analysis starts with a good question.
What Makes a Good Analytical Question
Good questions are:
Specific Not: “How’s the business doing?” But: “How did Q3 revenue compare to Q2 by product category?”
Measurable Not: “Are customers happy?” But: “What percentage of customers give us 4+ star reviews?”
Actionable Not: “What’s our average order value?” But: “Which customer segments have order values above $100, and can we acquire more of them?”
Comparative Not: “What were our sales?” But: “How did our sales compare to the same month last year?”
The Comparison Principle
Numbers without comparison are meaningless.
“We had 10,000 visitors” — Is that good? Bad? Average?
“We had 10,000 visitors, up 25% from last month” — Now we know.
Every analytical question should include a comparison:
| Comparison Type | Example Question |
|---|---|
| Time-based | How does this month compare to last month? |
| Target-based | Are we above or below our goal? |
| Segment-based | How do enterprise customers differ from SMB? |
| Competitive | How does our metric compare to industry average? |
Question Frameworks
Framework 1: The So-What Ladder
Start with a basic question. Ask “so what?” until you reach action.
Level 1: What were our Q3 sales? → So what?
Level 2: Sales dropped 15% from Q2. → So what?
Level 3: The drop was concentrated in the enterprise segment. → So what?
Level 4: Three large customers didn’t renew. Why? Can we prevent more? → Now we have an actionable question.
Framework 2: Who/What/When/Where/How Much
Build complete questions by filling in these elements:
WHO: [Customer segment, user type, team]
WHAT: [Metric, behavior, outcome]
WHEN: [Time period, comparison period]
WHERE: [Region, channel, product]
HOW MUCH: [Threshold, comparison point]
Example: “How many enterprise customers [who] had order values above $500 [how much] in Q3 2024 [when] through direct sales [where]—and how does this compare to Q3 2023 [comparison]?”
Framework 3: The Decision Frame
Work backwards from the decision:
What decision are we making? “Should we expand our sales team in Europe?”
What would change our decision? “If European revenue per rep is higher than US, yes. If lower, no.”
What data answers this? “Revenue by region, divided by sales headcount.”
The analysis question: “What is revenue per sales rep by region, compared over the past 4 quarters?”
Turning Vague Requests into Good Questions
When someone gives you a vague analysis request, convert it:
| Vague Request | Better Question |
|---|---|
| “Look at customer churn” | “Which customer segments have the highest churn rate, and what behavior patterns predict churn?” |
| “Analyze the marketing data” | “Which marketing channels have the lowest customer acquisition cost, and is this consistent across customer segments?” |
| “Tell me about sales trends” | “How has average deal size changed quarter-over-quarter by product line, and what’s driving the change?” |
| “Check on website performance” | “What’s our conversion rate by traffic source, and how does it compare to the previous month?” |
Quick check: Before moving on, can you recall the key concept we just covered? Try to explain it in your own words before continuing.
AI-Assisted Question Refinement
Use AI to improve your questions:
I need to analyze our sales data. My initial question is: "How are sales going?"
Help me refine this into a better analytical question by:
1. Adding specificity (what aspect of sales?)
2. Adding comparison (compared to what?)
3. Adding actionability (what decision would this inform?)
Suggest 5 refined questions.
AI can help you think through the angles you might be missing.
The Question Hierarchy
Some questions are more valuable than others:
LOW VALUE HIGH VALUE
│ │
├── What happened? │
│ (Descriptive) │
│ │
├── Why did it happen? │
│ (Diagnostic) │
│ │
├── What will happen? │
│ (Predictive) │
│ │
└── What should we do? │
(Prescriptive) │
Descriptive: “Sales were $1M” — Least valuable. Just states facts.
Diagnostic: “Sales dropped because enterprise customers churned” — Explains causes.
Predictive: “Based on trends, next quarter will likely be similar” — Anticipates future.
Prescriptive: “We should focus on retention over acquisition” — Recommends action.
Move up the hierarchy whenever possible.
Exercise: Question Transformation
Transform these vague requests into actionable analytical questions:
- “We need to understand our customers better”
- “The board wants to see revenue analysis”
- “Marketing wants to know if the campaign worked”
See possible transformations
“We need to understand our customers better” → “What characteristics distinguish our top 20% customers (by lifetime value) from the rest? What acquisition channels and behaviors do they share?”
“The board wants to see revenue analysis” → “How has revenue grown year-over-year by business line? What’s driving the change in each segment? Where are we ahead/behind target?”
“Marketing wants to know if the campaign worked” → “What was the cost per acquired customer during the campaign period vs. the 3-month prior average? What was the attributed revenue, and what was the ROI?”
Key Takeaways
- The quality of your question determines the quality of your analysis
- Good questions are specific, measurable, actionable, and comparative
- Numbers without comparison are meaningless—always include a comparison point
- Use frameworks: So-What Ladder, Who/What/When/Where/How Much, Decision Frame
- Move up the question hierarchy: descriptive → diagnostic → predictive → prescriptive
- Use AI to help refine vague requests into specific questions
Next: how to quickly explore and understand any dataset.
Up next: In the next lesson, we’ll dive into Rapid Data Exploration.
Knowledge Check
Complete the quiz above first
Lesson completed!