Decentralizing a delivery network

Big rig truck

The model by Jesús Madrid Varela and Christian Moller Corral can be seen in action in the AnyLogic Cloud.

Improving the distribution system of a wholesale company using agent-based simulation

Cooper Enterprises is a large wholesale company in Venezuela that sells a wide variety of goods in cities throughout the country. Products include hardware and electrical tools, veterinary supplies, and construction materials. The company has over a thousand customers and handles 800 stock keeping units (SKUs).

Copper Enterprises

Cooper Enterprises

The problem

Their current supply chain network is heavily centralized. All distribution operations are handled by a single facility located in the central region of the country. This leads to a lot of uncertainty in delivery times, a collapse of the system during periods of high demand, high transportation costs, and low customer satisfaction.

As part of a new business strategy, which aims to have a more efficient supply chain network and anticipate future market growth, the company has decided to decentralize its operations by adding new distribution centers (DC’s) in other regions of Venezuela.

Simulation vs. Trial & Error

Determining the right DC’s, where they should be located, and what resources to incorporate in these new facilities, was not something the company could afford to do by trial and error. Consequently, simulation was chosen as the most effective tool for assessing the problem. Simulation enabled the evaluation of several scenarios and the use of stochastic analysis to develop a proposal that effectively achieved an efficient and decentralized supply chain for Cooper Enterprises.

The model

The model simulates all the main activities of the Cooper Enterprises distribution system — from the moment a retailer generates an order for a product, until the goods are received at the store.

Logic model for Cooper Enterprises distribution network

Logic model for Cooper Enterprises distribution network

The model considers:

  • Order generation
    • Orders are generated at a varying rate, and come from different customers depending on each customer’s probability to order.
    • Using statistical information, an algorithm simulates which products are included in the order and in what quantities.
  • Warehouse prep-operations
    • Orders are sorted at a DC into 9 different delivery zone groups and prepared for shipping.
    • The model then selects the most suitable vehicle for each delivery batch, depending on the its volume and weight, and depending on which vehicle units are available at the DC for a given moment during the simulation.
  • Order deliveries
    • Using Statecharts the model simulates a multi-stop delivery operation (one vehicle unit serves multiple clients in a single journey). Vehicles move using a nearest-client routing policy until all customer orders are served, they then return to the warehouse.
  • Experimental design & measuring results
    • The experimental design of the model allows the user to test different scenarios; deciding whether to include a specific DC or not, and the number and types of vehicle units assigned to each facility.
    • The order frequency rate can be increased to simulate the growth of the company’s operations.
    • The system’s efficiency for each scenario is measured through a set of indicators included in the model.

Main outcome & results

Interestingly, adding just one more of the largest vehicles in the current system, significantly decreased order delays and transportation costs. Further improvements could be achieved by adding a new DC in the western region in Venezuela. And finally, delays can be reduced to almost 0% if two new DC’s are added, one in the East and the other in the West.

Cooper Enterprises distribution network with 2 additional DCs

Cooper Enterprises distribution network with 2 additional DCs

The company could have tried to avoid purchasing new vehicles by transferring units from the existing DC to new facilities. However, simulation proved that this would lead to a collapse of the system, with fleet utilization at the central DC and incurred delays becoming unacceptably high.

The results also demonstrated that two new DC’s would allow the system to perform efficiently with demand increases of up to 40%.


Through the use of simulation, Cooper Enterprises developed a roadmap for its decentralized expansion strategy. By knowing which decisions to avoid and what others to implement they could ensure the efficiency of their distribution network, both now and in the future.


You can see the model in action in the AnyLogic Cloud.

⭐ Find out more about AnyLogic and supply chain simulation modeling.

What do you think? Does decentralizing work for everyone? Let us know, give your view points below!