Analytical thinking — from business question to analytical task
A practical method for turning vague requests into precise, answerable analytical tasks, and the analytical lifecycle that carries a task to a decision.
This lesson builds on: Data sources, structures and data quality.
- Apply a structured set of questions to clarify a business request
- Decompose a business question into measurable sub-questions (an issue tree)
- Write a one-paragraph analytical brief with metric definitions, scope and success criteria
- Describe the stages of the analytical lifecycle
The problem with requests
Requests arrive vague: "Can you look at churn?", "I need a dashboard for sales." Doing exactly what was literally asked often produces work nobody uses. Analytical thinking means understanding the decision behind the request and designing the smallest piece of analysis that improves it.
Five clarifying questions
Ask these — politely and quickly — before writing any SQL:
- What decision will this inform, and who makes it? ("Whether to extend the Family Saver campaign — the Commercial Director, next Tuesday.")
- What would you do differently depending on the answer? If nothing changes, the analysis is not needed.
- How exactly is the metric defined? (Churn = subscriber with no billed month after month M? Voluntary only?)
- What scope? Time period, segments, comparison baseline.
- What does "done" look like? A number, a table, a dashboard, a recommendation? By when? At what precision?
Send a three-line summary: "You want X to decide Y by Z; I'll measure it as M over period P; you'll get a one-page memo with a recommendation." This single habit prevents most rework.
Decomposing the question: issue trees
Break the main question into mutually exclusive, collectively exhaustive (MECE) parts. Example — "Why did telecom churn rise?"
- Is churn really higher? (definition, data completeness, comparison period)
- Who is churning? (contract type, tenure, region, plan, payment method)
- When in the lifecycle? (first 3 months vs long-tenure customers)
- What happened before they left? (support tickets, late payments, price changes)
- External factors? (competitor offers, seasonality)
Each branch becomes a query or a chart. Branches that cannot change any decision can be dropped.
Writing the analytical brief
| Element | Example |
|---|---|
| Business question | Why did churn rise in H2 2025 and what can we do in Q1 2026? |
| Decision & owner | Retention budget allocation — Commercial Director |
| Metric definitions | Monthly churn rate = subscribers churning in month ÷ active subscribers at start of month |
| Scope | All consumer subscribers, Jan 2024–Dec 2025; compare H2 2025 vs H2 2024 |
| Data | subscribers, monthly_usage, support_tickets, campaigns |
| Hypotheses | Month-to-month contracts and cash payers churn more; poor support experience predicts churn |
| Deliverable | 2-page memo + 4-visual dashboard by 15 Jan |
| Success criteria | Director can choose between 3 costed retention options |
The analytical lifecycle
- Ask — clarify the decision and question.
- Plan — brief, hypotheses, data needed.
- Acquire — find and extract data.
- Prepare — clean, validate, reconcile.
- Analyse — explore, compare, model.
- Communicate — tell the story with evidence and limitations.
- Act & measure — the decision is made; measure the result and feed it back.
The cycle repeats: measurement after the decision often raises the next question.
Worked example
"Sales are down in Ankara, can you check?"
- Clarify: down vs what? Last month, last year, or plan? → vs plan.
- Decision: whether to change the store manager's Q1 target. → precision matters.
- Define: net revenue, completed orders, plan from
sales_targets. - Decompose: orders vs basket size; category mix; days with zero sales (closures?).
- Deliver: a table of monthly plan vs fact, the main driver, and a recommendation on the target.
Common mistakes
Easy-to-reach tables tempt you to answer an easier question. Go back to the brief.
Analysing every dimension "just in case" delays the decision. Prioritise the branches most likely to change it.
Practice
A manager writes: "Send me churn numbers." Which is the best first response?
In the analytical lifecycle, which stage comes immediately after acquiring the data?
Marketing asks whether the Summer Sale 2025 (1–20 July 2025) worked. Write (in 4–8 lines) how you would define "worked", what comparison you would use, and one confounding factor you must account for.
Independent assignment
Take this request: "Marketing wants to know if the Summer Sale 2025 worked." Write a one-page analytical brief using the table above. State at least two competing hypotheses and the data that would distinguish them.
Summary
- Clarify the decision, owner, definition, scope and deliverable before analysing.
- Decompose questions into MECE branches; drop branches that cannot change the decision.
- The lifecycle runs Ask → Plan → Acquire → Prepare → Analyse → Communicate → Act & measure, and repeats.