Newton.Vaureal Consulting is a consultancy company specializing in operations management since 1998. Newton has a team of Operations Research specialists to handle all modeling needs that arise in business projects such as logistics network design optimization. With 6 network design projects carried out by relying on LocalSolver last year, Newton optimized €160M of logistics costs overall and obtained 12% savings for its clients on average. In this case, Newton’s team sought to optimize the network design of one of its clients.
Network design problem
Every year, Newton accomplished a number of consulting projects to optimize dense logistics networks for B2B or B2C retailers. The project presented below concerns a French retailer of frozen food delivered at home (€70M in revenue). The retailer operates 2 preparation sites to prepare customer orders, and further transports them to 70 distribution sites. Given an expected increase in sales of 15% in the next 3 years, the goal was to challenge the economic efficiency of the logistics network of this retailer: optimal number and location of distribution (level 1) and preparation sites (level 2).
- 300,000 customers grouped into 3,700 geographical clusters
- 70 distribution sites + 250 greenfield locations for new sites
- 2 preparation sites + 250 greenfield locations for new sites
- 1 million decision variables
Find the optimal number and location of preparation and distribution sites.
- Each customer is assigned to a distribution site
- Each distribution site is assigned to a preparation site
- Lower limit on the activity level of each distribution site
- Capacity of preparation sites
- At most one additional preparation site
- Minimize transportation costs for downstream flows
- Minimize fixed and variable costs for preparation sites
- Minimize fixed costs for distribution sites
- Minimize costs for incoming transportation to preparation sites
The first optimization level, about the optimal location of distribution sites, addresses 71% of the total logistics costs and is the most challenging part. This problem was efficiently modeled and solved using LocalSolver. Newton obtained a near-optimal solution within 3 minutes of running time.
The financial savings were above 8% overall cost-to-serve and 10% on the first logistics level.
According to Renaud Lacour, Senior Modeling Expert, Newton.Vaureal Consulting has chosen to rely on LocalSolver as an optimization solver for three main reasons. First, the optimization engine is powerful and able to solve optimally or near-optimally the real-world instances that arise in their projects. Second, LocalSolver’s high-level, friendly modeling Python APIs enable them to quickly build readable and reliable mathematical models close to business problems. Last, the technical support is extremely reactive and helpful in improving the efficiency of their models.
The innovating features of LocalSolver and its unique modeling formalism have allowed us to solve numerous problems on which we could not find satisfying solutions with MIP solvers. LocalSolver’s performance surpasses greatly our expectations on routing, matching, and facility location problems we’re working on. In one year, we have optimized €160 million with an average gain of 12%. The responsive and super expert support given by Nikolas has greatly helped us in moving forward quickly on our projects.