Asking the Right Questions
Master formulating analytical questions using the SMART framework
Why Questions Matter
Imagine setting your GPS to "somewhere nice" instead of a specific address. You'd never get anywhere useful. The same is true in data analytics: your question is your destination. If your question is vague, unclear, or wrong, all the analysis in the world won't help you.
Questions guide the entire analytics process:
- They determine what data you need to collect
- They shape how you analyze the data
- They define what success looks like
- They ensure your work leads to action
🗺️ Real-Life Analogy: GPS Destination
Vague Destination
Input: "Somewhere with good food"
Problem: GPS can't navigate—too many possibilities, no clear direction
Result: You drive around aimlessly and waste time
Specific Destination
Input: "Joe's Pizza, 123 Main St, Chicago"
Benefit: GPS knows exactly where to take you
Result: You arrive quickly and efficiently
Analytics parallel: A vague question like "How are sales?" leads to aimless analysis. A specific question like "Did Product X sales increase by 10% this quarter in the Northeast region?" gives you a clear target.
Consequences of Wrong Questions
Example 1: Too broad
❌ Question: "Why don't people like our website?"
Consequence: You analyze everything (design, speed, content, navigation) and end up with 100 possible issues but no clear priority.
✅ Better question: "What are the top 3 reasons users abandon the checkout process?"
Example 2: Solving the wrong problem
❌ Question: "How can we get more email sign-ups?"
Consequence: You succeed in getting 10,000 new sign-ups, but none of them buy anything because they're not your target audience.
✅ Better question: "How can we get more email sign-ups from customers who have purchased from us before?"
Example 3: Can't be measured
❌ Question: "Are our customers happy?"
Consequence: "Happy" is subjective—you can't measure it without a clear definition.
✅ Better question: "What percentage of customers rate their experience 4 or 5 stars in our post-purchase survey?"
Golden rule: Spend time crafting your question before you touch any data. A great question is half the work.
Imagine you're going on a treasure hunt!
Before you start digging, you need to know WHAT you're looking for, right?
- ❌ Bad question: "Is there treasure somewhere?" (Too vague! You'd dig everywhere forever!)
- ✅ Good question: "Is there a gold coin buried under the big oak tree in the backyard?"
In data analytics, it's the same!
- ❌ "How are sales?" = You don't know what to look for!
- ✅ "Did ice cream sales go up this summer?" = Now you know exactly what to check!
A good question is like a treasure map—it shows you exactly where to look!
⚡ Quick Check: Why Questions Matter
Test your understanding:
1. In data analytics, your question is like a GPS destination—it determines where you go.
2. It doesn't matter if your question is vague—you can fix it later during analysis.
3. Spending time crafting your question before touching data is time well spent.
Vague vs. Specific Questions
Let's look at side-by-side examples to see the difference between questions that waste time and questions that drive action.
Question Transformations
| ❌ Vague Question | ✅ Specific Question | What Changed |
|---|---|---|
| "How are sales?" | "Did sales of Product X increase by 10% this quarter compared to last quarter?" | Added: specific product, metric, timeframe, comparison |
| "Who are our customers?" | "What are the demographic characteristics (age, location, income) of customers who spent over $500 last year?" | Added: specific attributes, spending threshold, timeframe |
| "Improve marketing" | "Which marketing channel (email, social, paid ads) has the highest ROI for customer acquisition?" | Added: specific channels, clear metric (ROI), defined outcome |
| "Why are people leaving?" | "What percentage of customers who signed up in the last 6 months are still active after 30 days?" | Added: specific cohort, measurable metric, timeframe |
| "Is our website slow?" | "What is the average page load time for our homepage, and how does it compare to the industry standard of 3 seconds?" | Added: specific page, measurable metric, benchmark |
| "Understand customer behavior" | "What is the average time between a customer's first visit and their first purchase?" | Added: specific behavior, measurable timeframe, clear metric |
| "Are customers satisfied?" | "What is our Net Promoter Score (NPS) for customers who purchased in the last 90 days?" | Added: specific metric (NPS), defined segment, timeframe |
| "Fix inventory issues" | "Which products had stockouts lasting more than 7 days in the past quarter?" | Added: specific issue (stockouts), duration threshold, timeframe |
| "Make better decisions" | "Should we expand Product Line A to Region B based on current demand patterns?" | Added: specific decision, specific products/regions, clear criteria |
| "Increase revenue" | "Which existing customers have not purchased in 90+ days and would be good candidates for a win-back campaign?" | Added: specific segment, timeframe, actionable outcome |
Notice the pattern: Specific questions include measurable metrics, clear timeframes, defined segments, and are tied to decisions or actions.
The SMART Framework for Questions
The SMART framework—originally designed for goal-setting—is perfect for formulating analytics questions. A SMART question is:
Specific
What it means: The question is clear and focused, not broad or ambiguous.
Ask yourself: What exactly am I trying to find out? Who/what is involved?
Example: Not "How is the website doing?" but "What is the bounce rate on the homepage?"
Measurable
What it means: The answer can be quantified with numbers or clear categories.
Ask yourself: How will I know when I've answered this? What metric will I use?
Example: Not "Are customers happy?" but "What is our average customer satisfaction score (1-5)?"
Actionable
What it means: The answer will lead to a decision or action—it's not just curiosity.
Ask yourself: What will we do differently based on the answer?
Example: Not "What's our revenue?" but "Should we reallocate budget from Channel A to Channel B based on conversion rates?"
Relevant
What it means: The question matters to stakeholders and aligns with business goals.
Ask yourself: Why does this matter? Who cares about the answer?
Example: Not "What's the average age of visitors to our blog?" (unless age matters for your business) but "What's the conversion rate of visitors from paid ads vs organic search?" (if you're trying to optimize marketing spend)
Time-bound
What it means: The question includes a clear timeframe or reference period.
Ask yourself: What time period am I analyzing? Am I comparing across time?
Example: Not "Are sales increasing?" but "Did sales increase month-over-month in Q4 2024?"
Practice: Transforming a Question Using SMART
Starting point: "We need to understand our customers better."
Apply SMART:
- Specific: What aspect of customers? → Focus on repeat purchase behavior
- Measurable: What metric? → Repeat purchase rate
- Actionable: What decision? → Decide whether to launch a loyalty program
- Relevant: Why does it matter? → Retention drives profitability
- Time-bound: What timeframe? → Customers acquired in the last 12 months
Result: "What percentage of customers acquired in the last 12 months made a second purchase within 90 days, and should we launch a loyalty program to increase this rate?"
✍️ Fill in the Blanks: The SMART Framework
Complete the SMART acronym for crafting good analytics questions:
Word Bank:
Specific Measurable Actionable Relevant Time-boundS = - Who, what, where exactly?
M = - Can you quantify it with a metric?
A = - Will the answer drive a decision?
R = - Does it matter to the business?
T = - What is the timeframe?
Types of Questions Analytics Can Answer
Not all questions are created equal. Analytics can answer four main types of questions, each building on the previous one.
1. Descriptive: "What happened?"
Purpose: Summarize past events and current state
Examples:
- "What was total revenue last quarter?"
- "How many new customers signed up this month?"
- "What is the average order value?"
- "Which product category sold the most units?"
When to use: You need to understand the current situation or report on past performance.
2. Diagnostic: "Why did it happen?"
Purpose: Identify root causes and relationships
Examples:
- "Why did website traffic drop 30% in March?"
- "What factors contribute to high customer churn?"
- "Why are conversion rates lower on mobile than desktop?"
- "What caused the spike in returns last week?"
When to use: Something changed and you need to understand what drove it.
3. Predictive: "What will happen?"
Purpose: Forecast future outcomes based on patterns
Examples:
- "How many units will we sell next month?"
- "Which customers are most likely to churn in the next 90 days?"
- "What will demand be during the holiday season?"
- "Will this customer make another purchase?"
When to use: You need to plan ahead and prepare for future scenarios.
4. Prescriptive: "What should we do?"
Purpose: Recommend specific actions based on data
Examples:
- "Should we increase inventory for Product X based on demand forecasts?"
- "Which customers should we target for the upsell campaign?"
- "What price should we set to maximize revenue?"
- "Should we expand to Region Y based on market analysis?"
When to use: You need to choose between options and take action.
Interactive Exercise: Match the Question Type
For each question below, identify the type (Descriptive, Diagnostic, Predictive, or Prescriptive):
1. "What was our email open rate last month?"
2. "Why did email open rates decrease by 15%?"
3. "What will our customer acquisition cost be next quarter?"
4. "Should we shift budget from Facebook ads to Google ads?"
Breaking Down Complex Questions
Sometimes the question you start with is too big to answer directly. The solution? Break it into smaller sub-questions.
Strategy: Large questions are often made up of smaller questions. Answer the sub-questions first, then combine them to answer the big question.
Example 1: "How can we increase revenue?"
Problem: This is massive—too many variables, too many paths.
Break it down into sub-questions:
- Sub-question 1: "Which products have the highest profit margins?"
- Sub-question 2: "Which customer segments spend the most?"
- Sub-question 3: "What are our seasonal sales patterns?"
- Sub-question 4: "Where in the customer journey do we lose potential buyers?"
How it helps: Each sub-question can be analyzed separately. Together, they give you a complete picture of revenue drivers.
Example 2: "Why are customers leaving?"
Problem: "Leaving" could mean many things—what specifically are we measuring?
Break it down into sub-questions:
- Sub-question 1: "What percentage of new customers don't return after their first purchase?"
- Sub-question 2: "How many active customers haven't made a purchase in 90+ days?"
- Sub-question 3: "What are the top reasons cited in cancellation surveys?"
- Sub-question 4: "Do customers who contact support have higher churn rates?"
How it helps: You discover that "leaving" has multiple dimensions (new vs old customers, cancellations vs inactivity) that each require different solutions.
Example 3: "Should we launch in a new market?"
Problem: This is a strategic decision that depends on many factors.
Break it down into sub-questions:
- Sub-question 1: "What is the estimated market size and growth rate in the new region?"
- Sub-question 2: "Who are the main competitors and what is their market share?"
- Sub-question 3: "What are the customer demographics and do they match our ideal customer profile?"
- Sub-question 4: "What are the estimated costs (logistics, marketing, operations) to enter this market?"
- Sub-question 5: "What is the projected ROI over the first 12 months?"
How it helps: You can analyze each factor (market size, competition, costs, ROI) separately, then combine findings into a recommendation.
Example 4: "How can we improve our website?"
Problem: "Improve" is vague—better design? Faster load times? Higher conversions?
Break it down into sub-questions:
- Sub-question 1: "What is the bounce rate on each major page?"
- Sub-question 2: "What is the average page load time and how does it compare to best practices?"
- Sub-question 3: "What percentage of users complete the checkout process?"
- Sub-question 4: "What do user feedback surveys say are the biggest pain points?"
How it helps: You identify specific, measurable issues instead of vague "improvements."
Pro tip: If your question starts with "How can we..." or "Why are...", it's probably too big. Break it down into sub-questions that can be measured and answered individually.
Questions That Can't Be Answered with Data
Data is powerful, but it's not magic. Some questions simply can't be answered with data alone—and that's okay. Recognizing the limits of analytics is part of being a good analyst.
❌ Ethical or Moral Questions
Example: "Should we sell this product even though it harms the environment?"
Why data can't answer it: This requires values and ethics, not numbers. Data can inform the decision (e.g., show environmental impact), but can't make the moral judgment.
❌ Subjective Preferences (Without Survey Data)
Example: "What's the best color for our logo?"
Why data can't answer it: "Best" is subjective. You could survey customers ("Which color do you prefer?") or test conversions (A/B test), but without data collection, there's no objective answer.
❌ Future Events with No Historical Pattern
Example: "Will a new technology disrupt our industry next year?"
Why data can't answer it: If there's no historical precedent, you can't predict it with data. You'd need scenario planning, expert judgment, and market research—not just historical analysis.
❌ Causation from Correlation Alone
Example: "Does ice cream sales cause crime to increase?"
Why data can't answer it: Data might show that ice cream sales and crime rates are correlated (both increase in summer), but correlation doesn't prove causation. You need additional evidence or experiments to establish causation.
Takeaway: Data is a tool to inform decisions, not a replacement for judgment. When you encounter these types of questions, acknowledge the limits of data and seek additional input (expert opinions, stakeholder values, experiments).
Interactive Question Formulator
Practice transforming vague questions into SMART questions with this interactive tool.
Question Transformation Tool
Select a scenario above
Key Takeaways
- Questions are your GPS: A vague question leads to aimless analysis
- Use the SMART framework: Make questions Specific, Measurable, Actionable, Relevant, Time-bound
- Know the four question types: Descriptive (what happened), Diagnostic (why), Predictive (what will happen), Prescriptive (what to do)
- Break down big questions: Complex questions are made of smaller sub-questions
- Recognize data's limits: Some questions need judgment, not just data
- Invest time upfront: A great question is half the analysis
📝 Knowledge Check
Test your understanding of this chapter! Choose the best answer for each question.
1. Which of the following is a SMART question?
2. What does the "M" in SMART stand for?
3. "Why did sales decrease in March?" is what type of question?
4. "How many units will we sell next quarter?" is what type of question?
5. Why is it important to break down complex questions into sub-questions?
6. Which question CANNOT be answered with data alone?
7. What's the main problem with the question "How are sales?"
8. A prescriptive question helps you:
9. Which component is NOT part of the SMART framework?
10. "What was our total revenue last month?" is an example of which type of question?