The Four Types of Marketing Analytics: Descriptive, Diagnostic, Predictive, and Prescriptive
Marketing analytics isn't one thing — it's four layers of increasing sophistication. What each type answers, when you need it, and what you need before you can use it.
The Four Types of Marketing Analytics: Descriptive, Diagnostic, Predictive, and Prescriptive
Marketing analytics is often discussed as one thing. It's four things, each answering a different question and requiring different data maturity to execute well. Understanding the progression helps you know what to build next.
Descriptive analytics: what happened?
Descriptive analytics tells you what occurred. Page views. Conversion rates. Email opens. Revenue by channel. It's the dashboard layer — the numbers that tell you whether things are going roughly as expected.
This is where most marketing teams live. They have descriptive analytics and stop there. Descriptive analytics tells you the what. It doesn't tell you the why, the what-if, or the what-next. It's necessary but insufficient — the foundation of a house that still needs walls and a roof.
Descriptive analytics requires: clean data collection, consistent metric definitions, and a reporting cadence. Most teams have at least a version of this, even if it's messy.
Diagnostic analytics: why did it happen?
Diagnostic analytics explains what caused the numbers to move. Traffic dropped 30% — diagnostic analytics tells you it was because Google updated its algorithm and your top three ranking pages lost positions, not because your content suddenly got worse.
This requires analytical thinking more than analytical technology. Someone has to look at the numbers, form hypotheses about what drove them, and investigate. The tools help (segmentation, drill-down, cohort analysis), but the analysis itself is human work. The software can't ask "why?" — it can only answer when you do.
Diagnostic analytics requires: descriptive analytics as a foundation, plus the time and skill to investigate anomalies rather than just reporting them.
Predictive analytics: what will happen?
Predictive analytics forecasts future outcomes based on historical patterns. Which leads are most likely to convert? What will next quarter's revenue look like if current trends continue? Which customers are at risk of churning?
Predictive analytics uses statistical modeling and machine learning to find patterns in historical data and project them forward. The quality of the prediction depends entirely on the quality and quantity of the historical data. Garbage in, garbage out applies here with mathematical precision.
Predictive analytics requires: substantial historical data (typically at least thousands of observations), clean data infrastructure, and either in-house analytical capability or a tool that packages the modeling into an accessible interface. This is where most small and mid-size marketing teams hit a wall — not because the technology is inaccessible, but because the data volume and cleanliness aren't sufficient.
Prescriptive analytics: what should we do?
Prescriptive analytics recommends actions based on predictions. "This lead has a 78% probability of converting — send them to sales now." "This ad is likely to fatigue within 4 days — prepare a replacement." "This channel mix maximizes projected ROAS given your current budget."
This is the most sophisticated layer and the least common in practice. It requires predictive analytics as input, plus an optimization layer that evaluates multiple possible actions and recommends the best one.
Prescriptive analytics works well for narrow, well-defined problems with clear optimization criteria (budget allocation, bid optimization, send-time optimization). It works poorly for broad strategic questions where the variables are numerous, the data is sparse, and the definition of "best" is contested.
The practical path
Most marketing teams need descriptive analytics that's actually reliable (many have it but it's not reliable) and diagnostic analytics that's actually performed (many have the data but nobody investigates anomalies). Predictive and prescriptive analytics are worth pursuing when those foundations are solid and the specific use case justifies the investment. Building predictive models on unreliable descriptive data produces predictions that are precise, confident, and wrong.