What data analytics is — and the four questions it answers
Data analytics turns recorded facts into decisions. Learn the four types of analytics through one running example and see why each answers a different business question.
- Define data analytics in your own words and distinguish data, information and insight
- Name the four types of analytics and the question each one answers
- Classify a real business request as descriptive, diagnostic, predictive or prescriptive
- Explain why most analyst work starts with descriptive and diagnostic analysis
Why this matters
Every organisation records what happens: a sale, a call to customer service, a new employee, a delivered parcel. Those records pile up in databases whether anyone looks at them or not. Data analytics is the discipline of turning those records into better decisions. If no decision changes, no analytics happened — only reporting.
This lesson gives you a vocabulary you will use for the rest of the program. You do not need any prior knowledge.
From data to decision
Start from first principles. Consider a single row stored by a shop's till:
| order_id | order_date | store | product | quantity | price |
|---|---|---|---|---|---|
| 51203 | 2025-11-28 | Baku Central | Air Fryer Plus | 1 | 113.96 |
- Data is the raw recorded fact: someone bought one Air Fryer Plus for 113.96 on 28 November. On its own, it tells you almost nothing.
- Information is data organised to answer a question: Baku Central sold 412 air fryers in November, up from 205 in October.
- Insight is information that explains something or changes what you believe: Air fryer sales doubled because of the Black Friday discount, and 60% of buyers were new customers.
- A decision is the action taken because of the insight: Re-run the discount next year, but order stock two weeks earlier because we ran out on day three.
An analyst's job is to move the organisation along that chain — and to be honest when the data does not support the next step.
Data analytics is the process of collecting, cleaning, transforming and analysing data to answer questions, explain what happened, anticipate what is likely to happen, and support decisions.
The four types of analytics
Analytics is usually divided into four types. Each answers a different question, and each depends on the one before it.
| Type | Question it answers | Typical output | Example (Orion Retail) |
|---|---|---|---|
| Descriptive | What happened? | Totals, trends, KPIs, dashboards | Net revenue in 2025 was 1.14M, +12.5% vs 2024 |
| Diagnostic | Why did it happen? | Breakdowns, comparisons, drivers | Istanbul Kadikoy (+35k) and Ankara — open since April 2024, so 2025 was its first full year (+29k) — contributed more than half of the growth |
| Predictive | What is likely to happen? | Forecasts, probabilities, scores | (illustrative) December 2026 revenue is expected within a stated range, e.g. ±6% |
| Prescriptive | What should we do? | Recommendations, optimised plans | (illustrative) Increase Black Friday stock for the fastest-selling products and test a 20% rather than 30% discount |
Descriptive analytics — what happened?
Descriptive analytics summarises the past. It sounds simple, but it is where most errors happen: counting cancelled orders as revenue, mixing gross and net figures, or comparing a 28-day February to a 31-day March. A trustworthy descriptive number requires an exact definition. "Revenue" must specify: which orders (completed only?), which amounts (before or after discount?), which dates (order date or delivery date?).
Diagnostic analytics — why did it happen?
Diagnostic analytics decomposes a change into its parts. If revenue grew 12.5%, did it grow because there were more orders, bigger orders, higher prices, or new stores? You answer by breaking the total down along dimensions (store, channel, category, customer type) and comparing periods.
Revenue = number of orders × average order value. In Orion Retail, completed orders rose from 11,277 (2024) to 12,713 (2025), +12.7%, while average order value was flat (90.2 → 90.0, −0.2%). Revenue therefore rose by 1.127 × 0.998 − 1 ≈ 12.5%. The diagnosis says growth came entirely from volume (more orders), not from bigger baskets — which leads to very different follow-up questions.
Predictive analytics — what is likely to happen?
Predictive analytics uses patterns in historical data to estimate unknown or future values: next month's demand, the probability that a subscriber leaves, the expected value of a lead. Predictions are always uncertain; a professional prediction includes a range or an error estimate, not a single magic number.
Prescriptive analytics — what should we do?
Prescriptive analytics recommends actions, often by comparing options under constraints: budget, stock, staff, regulation. It combines the other three: you need to know what happened, why, and what is likely, before you can say what to do.
Business application: one request, four layers
The Head of Sales sends: "Online sales looked weak last month — can you look into it?"
- Descriptive: confirm the fact. Online net revenue for the month vs the previous month and vs the same month last year. Perhaps it was only weak compared with a campaign month.
- Diagnostic: if it really was weak, break it down: traffic, conversion, basket size, category, new vs returning customers.
- Predictive: will it recover next month on its own (seasonal pattern), or is the decline structural?
- Prescriptive: recommend an action — for example, restore a payment option that was removed — and propose how to measure whether it worked.
Experienced analysts verify the claim before explaining it. A surprising number of "problems" disappear once the comparison period, the definition or a data-loading delay is checked.
Common mistakes
Beginners often want to build a model first. If you cannot yet describe and explain what happened, a model will only automate your misunderstanding.
"Sales were 4.7M" means nothing without a definition. Always state what is included and excluded.
Diagnostic analysis finds associations. Whether something caused the change needs further reasoning (Module 7 and 12).
Practice
A regional manager asks: "Which of our stores are most likely to miss next quarter's target?" Which type of analytics is this?
Revenue fell 8% last month. Which of the following are diagnostic analyses? Select all that apply.
Which statement is an insight rather than information?
Independent assignment
Pick an organisation you know (an employer, a university, a shop you use). Write down one realistic question for each of the four analytics types. For each question, list the data you would need and who would act on the answer. Keep it to one page — you will reuse it in the requirements lesson.
Summary
- Data analytics turns recorded facts into decisions: data → information → insight → decision.
- Descriptive = what happened; diagnostic = why; predictive = what is likely; prescriptive = what to do.
- Each type depends on the previous ones; most professional work is descriptive and diagnostic, done with precise definitions.
- Predictions carry uncertainty; recommendations need constraints and a way to measure impact.