← Commercial research

OPERATIONS · DEMAND & CAPACITY

Operational planning under uncertainty

What capacity balances service, cost and uncertainty?

Planned research

01 / BUSINESS DECISION

Start with the decision.

Choose a feasible staffing or inventory plan under uncertain demand, with a declared cost for shortages, excess capacity and changes to the plan.

02 / PRACTICAL IMPLICATION

What the evidence can support.

Forecast error matters through the decision it changes. This direction will compare service levels and realized costs against practical planning rules, rather than assuming the most accurate point forecast gives the best operational outcome.

03 / EVIDENCE

A defined question. Work still ahead.

Planned research

Study design only. No commercial planning model, evaluated policy or repository has been published for this direction.

Planned evaluation

  • Rolling-origin demand forecasts against seasonal-naive and recent-average baselines.
  • Quantile calibration and service-level coverage at the actual planning horizon.
  • Compare a fixed-buffer rule with a constrained cost-based policy on held-out demand paths, including stress scenarios.

No evaluated result, client impact or live model is claimed for this planned study.

04 / DATA

Know where the evidence comes from.

A synthetic demand process with seasonality, shocks and known capacity constraints is planned. Staffing and inventory will not be conflated; one operational decision will be selected before implementation.

05 / METHODOLOGY

Match the method to the design.

Begin with a single-site capacity decision and known lead time. Generate demand distributions, specify feasible resources and shortage/excess costs, then evaluate decision outcomes on future periods. Forecasting and optimization will be tested together.

06 / LIMITATIONS

Conditions that matter.

  • Costs, lead times and feasible resources are stated and consistent with the decision horizon.
  • Demand changes induced by capacity or stockouts must be modeled or excluded explicitly.
  • Held-out demand paths are unavailable to the planner when decisions are made.

Simulated policies cannot establish savings or service improvements for a business. Multi-site interactions, labor rules and operational constraints require separate evidence.

07 / CODE & NEXT STEPS

The next useful piece of work.

Implementation has not begun. A repository and source download will be linked when they exist and contain reviewable work.

Next step

Select one capacity setting, specify costs and constraints, and build a realistic baseline before adding model complexity.

← All commercial research