Types of data analysis: descriptive, diagnostic, predictive and prescriptive

People use "data analysis" for very different jobs. Sorting them by the question they answer makes it clear which method and which tool you need.

Descriptive: what happened?

Totals, averages, counts and trends over time. This is the base of everything else. Examples: sales by month, orders by region, the average basket size. A spreadsheet or a dashboard handles most of it.

Diagnostic: why did it happen?

You split a number by segment, compare periods and look for what moved it. A fall in revenue might sit entirely in one product or one channel. Diagnosis needs care: a link between two numbers is a lead, not a proof.

Predictive: what is likely to happen?

Forecasts and statistical models built from past data. The honest version reports a range and how wrong the method has been on past periods, not a single confident number.

Prescriptive: what should we do?

Choosing an action from the options, usually by testing "what if" scenarios against the forecast. It depends on the three types before it, and on judgement about things the data does not hold.

A worked example

Say a shop’s revenue fell last month. Descriptive analysis confirms the fall and its size. Diagnostic analysis splits it by product and channel and finds that one channel accounts for most of it. Predictive analysis asks whether the shortfall will continue if nothing changes. Prescriptive analysis weighs the options, such as fixing the channel or shifting budget. Each step needs the one before it, and skipping ahead is how teams act on a guess.

Two other terms you will meet

• Quantitative analysis works on numbers; qualitative analysis works on text, such as survey comments or interviews, and needs different software.
• Exploratory analysis is looking for patterns without a fixed question; confirmatory analysis tests a specific claim.

Choosing a tool by type

• Descriptive: spreadsheets, dashboards, business analytics and business intelligence tools.
• Diagnostic: tools that compare periods and break numbers down by group.
• Predictive: statistical analysis tools, Python, or a data tool that tests its forecasts on past data.
• Prescriptive: scenario tools, plus the people who own the decision.

In AnalyzeIt

You ask in plain words. Descriptive and diagnostic questions run as checked calculations on your data, and forecasts come with a tested range. The working is shown with each answer so you can verify it.

Read more: Try a question on your data, or try the live demo with no account.