AI as a gamechanger in a complex logistics chain

Organizations with long production chains often struggle with setting up their processes as efficiently as possible. This also applies to the dairy industry, where perishable ingredients play an important role. The main challenge in the case at present: how do we get our fresh dairy from a multitude of farms and other parties via an optimal route to the right factory? Taking into account a multitude of factors such as the fat and protein percentage of the milk or the purchase agreements there are with individual parties? Until now, this was still largely done manually via huge Excel sheets, and partly through the instincts of the buyer or planner. A time-consuming and error-prone job, in which the freshness and bacterial sensitivity of dairy products was a significant factor. Was a tanker truck traveling too long? Then the entire cargo load was to be rejected.

Initially, the optimization question to KPMG revolved around a drive for efficiency. By deploying AI, planning was able to take into account thousands of variations in detail for the first time: from the number of locations to purchase quantity and from product differences regarding fat and protein to routes and detours. This alone resulted in a savings of millions. In practice, there was a second opportunity to add value. Besides optimizing for travel movements, other aspects could be taken as starting points, for example those in terms of ESG. What was the most sustainable route? Which route created the least amount of CO₂ emission? Which route was the most efficient one for the drivers? Moreover – thanks to the use of AI – continuous measurements could be carried out, based on newly added data such as traffic volume, the composition of the dairy or unexpected issues such as a broken truck.

We came up with a ‘constraint-based optimization model’ at KPMG. In other words, the boundary conditions of the subject determined the outcome. In this case, a thousand factors such as routes, business agreements, purchase quantities and so on led to one optimal route. For this, we built a cloud-based solution with a state-of-the-art optimizer that was simple to use. The relevant data was sent to the cloud, after which an answer was given for the optimal planning for the trucks. In this case, the client's data management was basically in good order, but often, there is a big challenge in getting the source data up to par. The result was a system that both ensured a more efficient planning for the delivery of raw materials and substantially reduced costs from the outset.