The Question
Case study · Supply chain · Inventory policy
Client's brief · reconstructed excerpt · values held in confidence
(withheld), a leader in managed clothing solutions, keeps (withheld) garment SKUs in stock: every style in every size and color its customers' crews wear. Supply comes from (withheld) suppliers on lead times measured in weeks. Demand moves in jumps: a new customer installs, a crew adds wearers, a site changes uniforms. Planners run several reorder policies across those products, and the company wants to know which policy suits each one. It also has a new candidate: an AI policy that decides in real time whether to order, from which supplier, and how much. Trying any of them on the live supply chain means months of waiting, and real costs if the policy is wrong.
Bars cover values SimWell holds and is not releasing. The released figures stand as printed.
§ 1
History shows how one policy performed on one stretch of demand. It can't show how a different policy would have done on the same months, and every product behaves differently: a fast-moving work shirt and a rarely ordered size of flame-resistant pant need different rules. Trying a new policy on the real supply chain takes months to read, and a bad one shows up as an empty shelf on a customer's install date or as stock nobody needed. With thousands of SKUs, trial and error is not an option.
§ 2
Design a scenario and play it out in minutes. SimWell built the supply operation in AnyLogic with an input screen for the policy and its parameters, the products, and the time window. As the model runs, cost and inventory levels update on screen. Results flow into a Power BI dashboard that lines up every scenario on the same measures, so planners can compare policies product by product on equal terms.
§ 3
Only after it has earned it somewhere safe. The same simulation is the training ground for the AI policy, which decides in real time whether to order, from which supplier, and how much. It learns over thousands of simulated months, makes its mistakes where they cost nothing, and is then measured against the rules planners use today, on the same products and the same demand.
Representative model. Demand, policies, and costs shown are illustrative, not client data.
Cleared for release
The client's feedback from testing: the tool is central to improving its supply operations. Planners can now choose the best policy for each product before rollout, and test the AI policy against today's rules on equal terms.
Knowing the outcome before the order goes out is where the savings come from: fewer empty shelves on install dates, and less stock carried against demand that never arrives.
SimWell helps operators make better decisions under uncertainty. We build simulation and optimization models that reflect operational reality, test scenarios before committing resources, and transfer capability so teams can run decisions independently.
More case studies
A Canadian retailer tested a few dozen store-fulfillment networks in simulation and chose one that halved online delivery time and cut transportation cost 5%.
Read the case studyICP Group and SimWell built a refreshable model of a network assembled by acquisition, sized the DC footprint, and found a 7% supply chain cost-saving opportunity.
Read the case studyA leading spice manufacturer simulated whether one packaging line could absorb another site's volume, then consolidated for a 35% increase in return on investment.
Read the case studyWant to learn more? Subscribe to our newsletter for the latest updates and industry news.
