Business Questions vs Data Questions — Bridging the Gap

What You Will Learn

• The difference between a business question and a data question
• A 5-step framework for translating one into the other
• Worked examples of common translations
• Why this skill separates good analysts from average ones

Why This Topic Matters

Stakeholders never ask clean data questions. They ask things like "Are our customers happy?" or "Is the new product working?" If you take those questions literally and start computing, you will produce analyses nobody asked for. The single most valuable analyst skill is the ability to translate a fuzzy business question into a sharp, measurable data question — and to do it in a way the stakeholder agrees with.

The Two Types of Questions

Notice every data question has: a metric, a time window, a comparison, and often a segment. These are the four building blocks from lesson 09.

The 5-Step Translation Framework

1. Restate the question in your own words. Say it back to the stakeholder. "When you say 'are customers happy', do you mean satisfied with delivery, with the product, or with the app experience?"
2. Identify the metric. What number would answer the question? Happiness → post-delivery rating, repeat order rate, complaint rate.
3. Choose the time window and comparison. This month vs last? This month vs same month last year? Before/after a change?
4. Decide the segment. Are we asking about all customers or a specific group (new customers, customers in a city, customers who used a feature)?
5. Write the data question in one sentence. Include metric, window, comparison, segment.

Worked Example 1: "Are Customers Happy?"

Step 1 — Restate: "Happy with what?" The stakeholder says: "With the delivery experience."

Step 2 — Metric: Post-delivery rating (1–5 stars) and the share of 1-star orders.

Step 3 — Window and comparison: This month vs last month, and vs the same month last year.

Step 4 — Segment: By city, because the delivery experience varies by city.

Step 5 — Final data question: "How does the average post-delivery rating and the 1-star rate this month compare to last month and to the same month last year, broken down by city?"

That question can be answered with one SQL query and one chart. The original business question could not.

Worked Example 2: "Is Our New Feature Working?"

Step 1 — Restate: "What does 'working' mean?" Stakeholder: "More users should be completing checkouts."

Step 2 — Metric: Checkout completion rate = completed_checkouts / started_checkouts.

Step 3 — Window and comparison: The 4 weeks after launch vs the 4 weeks before.

Step 4 — Segment: New users vs returning users, since the feature may help one group more than the other.

Step 5 — Final data question: "Did the checkout completion rate increase by at least 5 percentage points in the 4 weeks after launch vs the 4 weeks before, separately for new vs returning users?"

Notice the explicit target (5pp) — that came from asking "what would success look like?" in step 1.

The Question You Should Always Ask Back

When a stakeholder asks a vague question, the most useful thing you can say is:

"What would you do differently depending on the answer?"

If they cannot answer that, the question is not yet ready for analysis. If they can, their answer usually reveals the real metric, the real segment, and the real threshold they care about.

A Common Trap: The Vanity Question

Some business questions sound important but cannot lead to action. Example: "What is the average customer lifetime value?" Computing this precisely requires assumptions about future behavior that may be wrong. The number is impressive but rarely actionable.

Always push toward: What decision will this number change? If the answer is "nothing", suggest a different question.

Common Mistakes

1. Taking the question literally. "Are customers happy?" taken literally is unanswerable. Translate first.
2. Picking a metric before agreeing on the question. If you start computing without confirming what "happy" means, you will produce the wrong number.
3. Skipping the comparison. "Average rating is 4.2" means nothing without "vs 4.0 last month" or "vs 4.5 target".
4. Ignoring segments. An overall number can hide a problem in one city or one customer type.
5. Forgetting the time window. "Revenue" without a window is meaningless.

Practical Exercise (10 minutes)

Take each of these vague business questions and translate them into measurable data questions using the 5-step framework:

1. "Is our marketing working?"
2. "Are deliveries getting faster?"
3. "Should we launch in a new city?"

Write your translations in one sentence each. If you cannot pick a metric, list two candidates and explain which you would choose and why.

Mini Challenge

Open any news article that cites a statistic. Reverse-engineer the data question the journalist was trying to answer. Was the question well-framed? Did the statistic actually answer it? Most journalism uses descriptive numbers without comparison or segment — see if you can spot the gap.

Key Takeaways