Knowing what will happen is only half the job
“You will sell 200 units” is not a decision. “Order 240 by Thursday” is.
Between prediction and action, there is a step that is performed manually in most companies, every time, by someone who keeps the constraints in mind: available budget, warehouse capacity, supplier lead times, and customer commitments.
It is time-consuming, untracked work that produces different results depending on who does it and how much time they have that day. Prescriptive analytics makes this step explicit and repeatable.
The three ingredients of every decision
- The objective: what needs to be maximized or minimized — total cost, service level, margin.
- The constraints: what is non-negotiable — budget, production capacity, delivery times, minimum service levels.
- The levers: what the system can act on — order quantities, resource allocation, pricing, processing sequence.
The difficult part is not technical: it is making the real objective and actual constraints explicit. There are almost always multiple objectives in conflict — minimum cost and maximum service cannot fully coexist — and the value of the analysis lies in turning that trade-off into a conscious choice rather than an accidental outcome.
Where we apply it most often
- Inventory replenishment and stock level management.
- Resource, shift, and production capacity allocation.
- Pricing and promotional policy decisions.
- Production planning and job sequencing.
- Delivery route and load optimization.
Tested on the past before being applied
Before adopting a policy, we test it against historical data: what would have happened over the last twelve months if we had followed this rule instead of the one used so far? It is the lowest-risk way to validate a decision and produces a comparison that the business can review and discuss.
The system also answers scenario-based questions — “what if the supplier's lead time increased by two weeks?”, “what if demand dropped by 15%?” — showing how the recommended action changes. This helps determine how robust the recommendation is to the assumptions it relies on.
What we deliver, and who decides
Recommended action, rejected alternatives with the reason why, estimated impact, and — most importantly — the active constraints: the ones actually limiting the outcome and worth negotiating. This is often the most valuable information, more so than the solution itself.
Recurring, low-risk decisions can become automated; important decisions remain yours, but with a recommendation already prepared. We define the line between the two together, using the same method as our Human Automation approach.