Prescriptive analytics

Prescriptive analytics is the most advanced tier of the analytics maturity spectrum, extending beyond predictive analytics (forecasting what will happen) to recommend specific actions that should be taken to achieve a desired outcome or optimize an objective. Where a predictive model might output "demand for this SKU will be 2,400 units next week," a prescriptive system outputs "order 800 units from Supplier A and 400 units from Supplier B, and mark down the current overstock by 12% to clear it before the replenishment arrives." Prescriptive analytics typically combines predictive models with optimization algorithms — linear programming, constraint satisfaction, simulation, or reinforcement learning — to evaluate the action space and select the best course of action under defined constraints.

In enterprise commerce, prescriptive analytics applications include dynamic pricing engines that recommend real-time price adjustments to maximize margin given competitive and demand signals, assortment optimization tools that prescribe which products to carry in which locations, and promotion planning systems that allocate marketing budgets across channels to maximize predicted return. The business impact of prescriptive analytics is substantial precisely because it eliminates the human bottleneck between insight and action — a recommendation engine that surfaces insights a human must interpret and act on is constrained by human bandwidth, while a prescriptive system that closes the loop into automated execution can operate at machine speed and scale. Building effective prescriptive systems requires both high-quality predictive models and clear specification of the business objective function being optimized.