Actuarial Statistics · Purpose and function of data analysis

Descriptive, Inferential and Predictive Analysis Explained

Updated 11 October 2026 · Fact-checked

Descriptive analysis summarises the data you have. Inferential analysis uses a sample to draw conclusions about a wider population, with a measure of uncertainty. Predictive analysis uses past data to forecast new or future outcomes. Choose the type that matches the question being asked, and document the work so others can reproduce it.

Understand Descriptive, Inferential and Predictive Analysis

Data analysis starts with a question. The type of analysis you choose depends on that question. The IAI Core Principles syllabus expects you to tell the types apart and say when each fits.

Descriptive analysis summarises the data in front of you. You use totals, averages, spreads, tables and charts. It makes no claim beyond the data set. Example: the average claim size in last year's motor claims was ₹18,500.

Inferential analysis uses a sample to say something about a larger population. You use estimation, confidence intervals and hypothesis tests. The answer always carries uncertainty. Example: from a sample of 400 policies, is the true mean claim size above ₹18,000?

Predictive analysis builds a model from past data to forecast outcomes for new cases. Regression, GLMs, time series and machine learning are typical tools. Judge it by how well it predicts unseen data. Related terms you may see are prescriptive analysis, which recommends actions based on predictions, and diagnostic analysis, which asks why something happened. Know these as extra labels. The three main types are the core of the topic.

Exploratory work looks for patterns and ideas without a fixed hypothesis. Confirmatory work tests a hypothesis stated in advance. Exploratory results are suggestions. If you test many patterns on the same data, some will look significant by chance. So confirm them on fresh data or with a pre-set test.

Reproducibility means someone else can take your data and method and get the same results. You achieve it by keeping the code, recording data sources and versions, noting assumptions and setting random seeds in simulations. Actuaries need this for audit, peer review and regulation.

How to solve Descriptive, Inferential and Predictive Analysis questions

Use this method for any question that asks you to classify, choose or justify a type of analysis.

1Read the question and find the actual aim: summarise, generalise, forecast, or recommend.
2Match the aim to a type: summarise is descriptive, generalise is inferential, forecast is predictive, recommend action is prescriptive.
3Check whether the data is the whole population or a sample. A sample used to conclude about a population points to inference.
4Decide whether the work is exploratory or confirmatory. Is there a hypothesis fixed before looking at the data?
5Name suitable tools, such as summary statistics and charts, confidence intervals and tests, or regression and GLMs.
6State the limits: sampling error, model assumptions, overfitting, or data-dredging risk.
7Add reproducibility: code, data source, assumptions and seeds recorded.
8Give a clear conclusion that links the chosen type back to the aim.

Quickest way: Aim-to-type matching

When to use it: Use this for MCQs and for the first line of a written answer.

Underline the verb in the question: describe, estimate or test, forecast, recommend.
Describe means descriptive. Estimate or test means inferential. Forecast means predictive. Recommend means prescriptive.
If the question mentions a hypothesis set in advance, choose confirmatory. If it mentions looking for patterns, choose exploratory.
Add one limitation for the chosen type to earn the extra mark.

Common mistakes in Descriptive, Inferential and Predictive Analysis

Calling any use of a sample inferential.
Students link the word sample with inference automatically.
Fix: Ask whether a conclusion is drawn about the wider population with uncertainty. If you only report the sample mean, it is descriptive.
Treating prediction and inference as the same thing.
Both use models such as regression.
Fix: Inference focuses on parameters and uncertainty about them. Prediction focuses on accuracy for new observations. The same model can serve either aim.
Treating exploratory findings as proof.
A striking pattern feels convincing.
Fix: Say that exploratory results generate hypotheses. Confirm them with new data or a pre-specified test.
Confusing reproducibility with getting the same answer from a new sample.
The two ideas sound alike.
Fix: Reproducibility means the same data and method give the same results. Collecting new data and finding similar results is replication.
Giving a definition without linking it to the scenario.
Students memorise labels but skip application.
Fix: Quote details from the question, such as the data size or the aim, and tie your choice to them.
Ignoring the limits of each type.
Students think only about what the method does.
Fix: Add one limit: descriptive cannot generalise, inferential needs sampling and assumptions, predictive can overfit.

Worked examples

Example 1

An Indian general insurer has a full database of last year's 12,000 health claims. (a) The actuary reports the mean and standard deviation of claim size and plots a histogram. (b) She then takes a random sample of 500 claims from this year's ongoing claims and tests whether the mean claim size has risen. (c) Finally she builds a model to forecast next year's claim costs by age band. Classify each task and justify.

Show the solution

(a) The aim is to summarise data already held. It uses summary statistics and a chart. No claim goes beyond the data. This is descriptive analysis.
(b) The aim is to conclude about all ongoing claims from a sample, using a hypothesis test with uncertainty. This is inferential analysis. It is also confirmatory, since the hypothesis is stated in advance.
(c) The aim is to forecast future costs from past data using a model. This is predictive analysis. Its quality should be judged on data not used to build the model.

Answer: (a) Descriptive. (b) Inferential and confirmatory. (c) Predictive.

Example 2

A student analyst examines 40 variables in a motor portfolio and finds that one of them is strongly related to claim frequency. She wants to report that this variable is a proven risk factor. Comment, and say what she should do for her work to be reliable and reproducible.

Show the solution

Checking 40 variables without a prior hypothesis is exploratory analysis.
With many comparisons, some will look strong by chance alone. So the finding is a hypothesis, not proof.
To confirm it, she should test the variable on new or held-out data, using a test and model set out beforehand.
For reproducibility she should keep her code, record the data source and version, list the assumptions and cleaning steps, and set a random seed for any sampling or simulation.
She should report the limits clearly, including the number of variables examined.

Answer: The result is exploratory and is not proof. She should confirm it on fresh or held-out data and document code, data, assumptions and seeds so that others can reproduce it.

Exam tips

Match the verb in the question to the type of analysis before writing anything.
Always add one limitation. Many written marks come from recognising the weakness of a method.
Use details from the scenario instead of generic definitions.
In computer-based or practical answers, mention how you would record code and data so the work is reproducible.
Keep the three main types separate from the extra labels such as prescriptive and diagnostic. Define those briefly if asked.

Practice questions from Purpose and function of data analysis

Descriptive, Inferential and Predictive Analysis: frequently asked questions

What is the difference between descriptive and inferential analysis?

Descriptive analysis summarises the data you have, with no claim beyond it. Inferential analysis uses a sample to draw conclusions about a wider population and states the uncertainty. Confidence intervals and hypothesis tests are inferential tools.

Where does predictive analysis fit in?

Predictive analysis forecasts outcomes for new cases using a model built on past data. It is judged on accuracy for unseen data. Regression, GLMs, time series and machine learning are common tools.

What is the difference between exploratory and confirmatory analysis?

Exploratory analysis looks for patterns without a fixed hypothesis and produces ideas. Confirmatory analysis tests a hypothesis set before looking at the data. Confirmatory results are more reliable because they limit chance findings.

Why does reproducibility matter for actuaries?

Others must be able to check, audit and rely on your work. Keeping code, data sources, assumptions and random seeds lets someone rerun the analysis and get the same results. It supports peer review and professional standards.