Advanced Performance Management · Data science and analytics

Data Analytics Types: Descriptive, Diagnostic, Predictive and Prescriptive

Updated 11 October 2026 · Fact-checked

The four types of analytics answer four questions. Descriptive: what happened? Diagnostic: why did it happen? Predictive: what is likely to happen? Prescriptive: what should we do? To answer an exam question, identify the decision, match the type to the question, and apply it to the scenario.

Understand Data Analytics Types: Descriptive to Prescriptive

Data analytics means examining data to find patterns that help managers decide. The four types form a ladder. Each step adds more insight and needs more data skill and better data.

Descriptive analytics summarises past and current data. It uses totals, averages, trends, dashboards and reports. A monthly sales report by region is descriptive. It tells you what happened but not why.

Diagnostic analytics looks for causes. It drills down, compares segments and finds correlations. If sales fell in one region, diagnostic work might show that a competitor opened nearby or that a key product was out of stock. Variance analysis and root cause analysis are familiar examples.

Predictive analytics uses historical data, statistical models and machine learning to estimate what is likely to happen. Examples are demand forecasts, customer churn scores and credit default risk. It gives probabilities, not certainty. Its quality depends on the quality and relevance of the data.

Prescriptive analytics recommends the best action. It uses predictions plus optimisation, simulation and rules to compare options against objectives and constraints. Examples are dynamic pricing, route planning and stock replenishment. The difference from predictive: predictive says what is likely to happen, prescriptive says what to do about it.

Key rules to remember

Descriptive: What happened?; Past and current data. Reports, dashboards, KPIs, summary statistics.
Diagnostic: Why did it happen?; Drill-down, comparison, correlation, root cause and variance investigation.
Predictive: What is likely to happen?; Forecasts and probabilities from models. Outputs are estimates, not certainties.
Prescriptive: What should we do?; Recommends actions using optimisation, simulation and decision rules, within constraints.
Value and complexity: Descriptive → Diagnostic → Predictive → Prescriptive; Insight and decision value usually rise along this order, and so do data and skill needs. This is a general pattern, not a strict rule.

How to solve Data Analytics Types: Descriptive to Prescriptive questions

Use this method for any scenario question on analytics. Link every point to the organisation's decision.

1Read the requirement. Note whether it asks you to explain, apply, recommend or evaluate.
2Identify the decision or problem in the scenario, such as falling margins, stock-outs or customer loss.
3Match each part of the problem to the type of analytics: what happened, why, what next, what to do.
4Name the data needed for each type and where it would come from, internal or external.
5Give a specific use in the scenario, with a clear link to performance, such as a KPI or a cost.
6Evaluate: note limits such as data quality, cost, skills, bias, model error, privacy and over-reliance on the model.
7Conclude with a recommendation. Say which type to start with and what to add later.
8Check professional skills: be concise, commercial and balanced, and use the format asked for.

Quickest way: Four-question shortcut

When to use it: Use when time is short and you need a clear structure within a few minutes.

Write four labels: What happened? Why? What next? What to do?
Under each, put one scenario-specific example, not a generic one.
Add one data requirement or limitation per type.
Finish with a one-line recommendation that links to the decision and the organisation's goals.

Common mistakes in Data Analytics Types: Descriptive to Prescriptive

Mixing up predictive and prescriptive analytics.
Both look forward, so they seem the same.
Fix: Predictive estimates what will probably happen. Prescriptive recommends the action to take. If the answer is a forecast, it is predictive. If it is a choice, it is prescriptive.
Giving textbook definitions with no link to the scenario.
Definitions are easy to memorise and write.
Fix: After each definition, add one example using the company's products, customers or processes.
Treating diagnostic analytics as only variance analysis.
Variance analysis is the most familiar APM technique.
Fix: Say diagnostic work also includes drill-down, segment comparison and correlation to find causes.
Presenting analytics as always accurate.
Students focus on benefits and forget limits.
Fix: State that models rely on past data, may be biased or wrong, and need human judgement. Mention data quality and ethics.
Recommending prescriptive analytics straight away.
It sounds the most advanced.
Fix: Check the organisation's data, skills and cost. A firm with poor reporting may need descriptive and diagnostic basics first.
Ignoring professional skills in the format.
Students write a list of definitions.
Fix: Write in the requested format, such as a report or email. Prioritise points, be commercial and justify your conclusion.

Worked examples

Example 1

A retail chain has 120 stores. Management receives a monthly report showing sales by store. Sales fell in 15 stores last quarter, and the board wants to know what to do. Explain how each type of analytics could help.

Show the solution

Descriptive: the report already shows what happened. Sales fell in 15 stores. Add trends by product, week and store type to show the scale and pattern of the fall.
Diagnostic: look for why. Compare the 15 stores with others on local competition, stock availability, staffing, pricing and customer footfall. This may reveal a common cause.
Predictive: use the causes found and past data to forecast sales for the 15 stores if nothing changes, and to estimate which other stores are at risk.
Prescriptive: compare options such as price changes, promotions, stock reallocation or store closure. Recommend the option that best meets the sales and margin targets within budget limits.
Limits: data quality across stores, model error and local factors that data may miss. Management judgement is still needed.

Answer: Descriptive analytics shows the fall, diagnostic finds the causes, predictive estimates future sales and risk, and prescriptive recommends the best response. The chain should check data quality and apply judgement before acting.

Example 2

A software subscription company wants to reduce customer cancellations. A director says, 'We need predictive analytics, and that will tell us what to do.' Comment on this statement.

Show the solution

Agree in part. Predictive analytics can score each customer on the likelihood of cancelling, using usage, complaints, payment history and contract length.
Challenge the second half. A churn score does not say which action works. That is prescriptive analytics, which would compare options such as discounts, support calls or product training, and recommend the best one per customer segment.
Point out the need for diagnostic work first. Understanding why customers leave, for example poor onboarding or price, makes the model and the actions more relevant.
Note the data needs: reliable customer records, usage data and results of past retention offers.
Note the limits: the model gives probabilities, may be biased, and customer data raises privacy and ethical duties.
Recommend a sequence: use descriptive and diagnostic analytics to understand churn, build a predictive model to flag at-risk customers, then use prescriptive methods to choose actions, testing results against retention and cost KPIs.

Answer: The director is only partly right. Predictive analytics identifies who is likely to cancel, but choosing what to do is prescriptive. The company should combine diagnostic, predictive and prescriptive analytics and manage data quality, bias and privacy.

Exam tips

Always tie each type to the scenario's decision. Generic definitions earn few marks.
Use the four questions as a structure, then add data needs, benefits and limits to reach the depth the requirement asks for.
When asked to evaluate or recommend, give a balanced view and a clear conclusion, including which type to adopt first.
Include ethics and data quality briefly. They are easy marks and show professional scepticism.
Write in the format asked for, keep paragraphs short, and use headings if a report is requested.

Practice questions from Data science and analytics

Data Analytics Types: Descriptive to Prescriptive in other exams

The same ground in other exams, if you are preparing for more than one or want another angle on it.

Data Analytics Types: Descriptive to Prescriptive: frequently asked questions

What is the difference between predictive and prescriptive analytics?

Predictive analytics estimates what is likely to happen, such as next quarter's demand. Prescriptive analytics recommends what action to take, such as how much stock to order. Prescriptive analytics often uses predictions as an input.

Which type of analytics is most valuable?

Prescriptive analytics usually supports decisions most directly, but it also needs the most data, skill and cost. The best choice depends on the decision, data quality and resources. Many organisations gain a lot from good descriptive and diagnostic work.

How do I answer a data analytics question in APM?

Identify the decision in the scenario, match each type of analytics to a part of it, and give specific examples and data needs. Then evaluate limits such as data quality, bias and cost, and end with a recommendation.

Is variance analysis descriptive or diagnostic?

Calculating the variance is descriptive because it shows what happened against budget. Investigating why the variance occurred is diagnostic. Explain both parts if the question asks about variances.