9.4 Types of Models

BI models fall into three broad categories, each serving a different purpose in the decision-making process.

9.4.1 Descriptive Models

Descriptive models summarize and characterize data as it is — they explain what happened, not what will happen (Han et al. 2011). Techniques include measures of central tendency, clustering, and association analysis. A retail chain might use cluster analysis to segment customers by purchasing behavior, then tailor marketing to each segment. In healthcare, descriptive models identify common characteristics among patient groups to guide policy.

9.4.2 Predictive Models

Predictive models forecast future outcomes based on historical patterns (Hastie et al. 2009). Techniques range from linear and logistic regression to time series analysis and machine learning algorithms. In finance, predictive models forecast stock prices; in supply chain management, they forecast demand to optimize inventory; in customer relationship management, they predict churn so businesses can act before customers leave.

9.4.3 Prescriptive Models

Prescriptive models go beyond prediction to recommend actions (Powell 2011). Using optimization and simulation techniques, they evaluate scenarios and constraints to identify the best course of action. In logistics, prescriptive models optimize delivery routes; in energy, they simulate production strategies for resource allocation; in risk management, they prescribe strategies to mitigate threats and capitalize on opportunities.

These three types form a hierarchy: descriptive models tell you what is happening, predictive models tell you what is likely to happen, and prescriptive models tell you what to do about it. Most BI projects begin with description and move toward prediction and prescription as the organization’s data maturity grows.