Publication Date
Topic(s) Covered:
  • Urban Logistics

Brazil’s largest online retail platform, B2W Companhia Digital, is implementing a reengineered network design, based on a model developed in collaboration with the MIT Center for Transportation & Logistics (MIT CTL). For the first time, the company is using a network optimization model to assess its operational footprint and provide actionable insights on how to make the network more cost efficient and improve service quality.

In addition to helping B2W evolve in line with the dynamic online commercial environments, the project provides some important lessons about modelling last-mile operations.

B2W delivers packages to tens of millions of e-consumers in urban, suburban and rural areas of Brazil. The company began operating in 2006, and since then has developed four different online platforms. It commands around 27% of the Brazilian e-commerce market.

The market has evolved rapidly over recent years. Package volumes have increased – reflected in B2W’s 30% year-on-year growth – and customers have become more demanding. Competition is fierce.

The enterprise needed to address the complex tradeoffs in urban last-mile distribution much more systematically, using detailed information on delivery service performance, customers, and the infrastructure environments in which it competes.

Concentrating initially on its top market, São Paulo, the company worked with MIT CTL to build a last-mile distribution optimization model.

The model analyzes large volumes of internal transactional data blended with external data such as road network and traffic reports. It enables B2W to identify optimal network designs and fleet compositions, as well as the most efficient route configurations in the company’s territories.

After calibrating the model, the team carried out an in-depth analysis of B2W’s last-mile distribution network in São Paulo. Based on this analysis, the project team identified several key revisions that would improve the network’s performance and cost-effectiveness. The measures included closing 30% of the company’s transshipment centers, and reducing the number of vehicles in its delivery fleet by 15%.

The two-phase plan was projected to reduce the cost of B2W’s last-mile operations by 6%. Implementation was completed in May 2017.

For the first time in its history, B2W now has the analytical capabilities required to fully leverage its own operational data combined with public data to improve network performance and control costs.

It’s been a steep learning curve, but there are three key takeaways from the experience that can benefit any enterprise faced with a similar change management challenge.

Detail and data integrity are critical. Building, calibrating and continuing to develop a model of this sophistication requires high-resolution data. Collecting tons of operational data is not enough; the input from the field must be extremely granular. For example, B2W already collected GPS location data from delivery vehicles but these traces were collected every few minutes, which was inadequate. And the story told by the operational data was too shallow. It’s not enough, for example, to indicate that a vehicle has stopped – the data must indicate why a vehicle is stationary. Did the driver come to a halt at a traffic light or stop to serve a customer? Were multiple customers served during a single stop, and if so, how long did it take the driver to visit each one? And, of course, the data must be of high-quality and free of noise. The project team developed a mobile app for collecting delivery data.

In-house technical expertise is another critical success factor. From the outset, B2W’s analytics team was fully engaged with the project. Their ongoing involvement ensured that team members understood the structure of the model and its capabilities. Moreover, their expertise helped in the model’s early development. B2W’s inhouse analytics team will continue to support the model, and help the company to develop and refine it going forward. This is extremely important. It means, for example, that B2W can scale the model in line with the company’s growth and future market demands without the need to call on the original developers. Also, the company’s in-house IT experts will make sure that the input data for the model continues to meet the required quality standards.

Acquire the right tools. Creating, operating and developing a sophisticated network optimization engine is not possible without the right IT tools. Companies that are not familiar with this type of technology – especially those that are accustomed to working primarily with spreadsheets – probably don’t possess the commercial-grade software required to build and fully use such a high-level resource. Be prepared to invest in the right toolbox before embarking on such a development project.

B2W’s new last-mile distribution strategy is being implemented in other Brazilian cities including Rio de Janeiro and Salvador. There are plans to extend the model to include additional delivery options such as customer pick-up and drop-off solutions. Furthermore, to help it keep pace with rising customer expectations, B2W wants to include alternative delivery times and windows in its logistics analyses, and to study the impact of these additions on network design, logistics costs over the last mile, and pricing policies.

Importantly, the gains made so far have solidified the support of the company’s leadership for using advanced analytics and operations research tools and methods to achieve further logistics improvements.

This article is based on a column written by Dr. Matthias Winkenbach, Director of the MIT Megacity Logistics Lab, and published in the journal Supply Chain Management Review. For more information on CTL’s last mile research contact the author at:

Back to Supply Chain Frontiers issue #65