Image by Denny Müller

Warehouse Location Optimization

AI Powered Simulations to Optimize Warehousing Costs




A consumer goods manufacturer wants to optimize future warehouse locations to serve all its retailers while also identifying alternatives to its existing layout.

  • Old School. Although the company is one of the leading brands in the market it still relied on outdated methods of domain experience to determine the optimum locations.

  • Hard to Scale. With the ever-growing nature of the company and its entry into new markets, it’s becoming harder to solve the warehouse location problem efficiently.

  • No Alternatives. There was no way to simulate multiple solutions, i.e., experimenting with possible warehouse locations  that would be used to estimate the cost of the logistic operation.




To obtain accurate results after deployment, a multitude of factors have been considered to achieve maximum cost-efficiency:

  • Retail Information. To predict the new locations of the warehouses, it is critical to look at not only present information of retailers, but also future predictions of growth based on historical data.

  • Data Factors. The approach also required to take into consideration a variety factors to ensure optimum results. Everything from the drop size to the replenishment schedule had to be included in our calculations.


The Simulation. By blending all the data sources alongside demographic and macroeconomic data, we created a highly complex simulator that is able to find the optimum number of distributors/warehouses for any scenario.

  • The steps repeat until an absolute minimum cost is found.


  • The algorithm starts choosing the number of warehouses to be equal to 6 for example.

  • Stores are associated according to the closest warehouse

  • The proposed warehouses move to the weighted center of association zones.

  • Association zones change according to new shortest distance.

  • The warehouses move again to the Weighted Centers 

  • The process is repeated until the optimum scenario is found and tries for another of number of warehouses.


  • Accurate Distribution. The results of the simulation yield an accurate number and distribution of warehouses across the map with the minimum cost.

  • Scalable Results. The results can be applied globally, are scalable to every region, and can be replicated to solve global similar problems in a fraction of the time.

  • Continuous Testing. The algorithm continuously runs and retrains itself to ensure that the results remain accurate and a minimum cost is being achieved at all times.


The project is successfully running, minimizing inaccuracies and increasing scoring quality as it enters phase 2: The Insight-Driven Transformation phase.