Prescriptive Analytics
What Should We Do?
Predictive analytics says what will happen. Prescriptive analytics says what you should do about it. It is where analytics finally meets the decision: a forecast becomes a recommended action by weighing scenarios against the business's rules and constraints, then choosing the option that delivers the best expected outcome. A prediction nobody acts on changes nothing — prescriptive analytics closes that gap.
This is a four-chapter mini-series built around one concrete decision: how to split a fixed marketing budget
across channels. Each chapter shows the Python and the rendered result, and
the optimization is real scipy.optimize.linprog — only the expected-return inputs are
illustrative (the kind a predictive model would supply).
A predicted outcome is only an input; prescriptive analytics adds scenarios and constraints, then optimizes to a single recommended action.
The Four Chapters
Scenarios — Model the Levers
What-if analysis: compare candidate allocations and see how much intuition leaves on the table versus a deliberate plan.
Business Rules & Constraints
Encode the limits — budgets, minimums, policy — so a recommendation is not just profitable but allowed.
Optimization — Allocate for the Best Outcome
Turn the decision into an objective, variables, and constraints, and let a linear-programming solver find the optimum.
Feasible, Not Just Optimal
Sensitivity analysis and the discipline that keeps a recommendation honest — including when more budget stops helping.
Reference Code
The scenario-modeling and constrained-allocation helpers are packaged in the
Data Analytics Library's
prescriptiveanalysis module. Prescriptive analytics is the final stage of the lifecycle: it
consumes Predictive outputs and closes the loop that began with
Descriptive ("what happened").