Large-Scale Logistics Network Planning and Optimization

Large-Scale Logistics Network Planning and Optimization

Analysts at Pitney Bowes wanted a parcel logistics network analysis tool to support decision making and help improve network performance in North America.

Pitney Bowes is a technology company based in Stamford, Connecticut, that specializes in services related to mailing and shipping. The company serves approximately 1 million customers and helps sort and process 15 billion pieces of mail annually.

The Pitney Bowes Innovation unit is a global development organization that supports the creation of secure best-in-class products based on agile design and data science.

The simulation-based logistics modeling software developed by the innovation unit's engineers has helped deliver significant savings across key metrics, such as parcel cycle time, truck utilization, and daily throughput.


Focusing on the delivery and return service of their operations in the United States of America, engineers sought to better understand four key areas of logistics network analysis:

This work was previously the job of a logistics network design expert alone but as the delivery network had scaled, a logistics planning tool was needed to support decision making. With the help of a simulation model of the parcel network, network design experts would be able to test their initial designs, identify bottlenecks, and generate better solutions.


A simulation model of the parcel network provided a platform for testing and analysis. The model could be configured with historical data for forecasting or with test scenarios for risk analysis and planning.

Graphic showing parcel network inputs, processes, and outputs for a parcel network induction hub.

Simulation model structure. (Click to enlarge)

The engineering team chose AnyLogic because of its in-built libraries, which help speed up model development, and its customizability, which allows for the accurate modeling of specific functionalities. Furthermore, thanks to the design, the model is easy to scale. For example, facilities can be added or removed using just a database entry. After a change to the number of facilities is made, the model updates and adjusts automatically.

Another factor in choosing AnyLogic was that the development engineers could deliver a standalone application for analysts and stakeholders to use.

The simulation model was useful in six different use cases:

Graphic showing icons that represent Network expansion and consolidation, Stress testing on facility capacity, Rerouting, Service guide simulation, Adding new clients, and Network fine-tuning

Parcel network simulation model use case overview. (Click to enlarge)


At present, the tool is used for network design evaluation. In the case of network consolidation, an indicative example of how the tool helps improve performance comes from the analysis of the merging of three facilities into one super center.

Chart showing results of parcel network site consolidation analysis. 10% reduction for average cycle hours, 100% reduction in rollover, 70% reduction in carry over, and 8% reduction in truck numbers.

Facility consolidation analysis results.

Working with historical data and carrying out parameter variation experiments, the team completed a robust analysis. Modeling showed that only 70 % of the initially planned capacity was required for a new super center and that the planned facility consolidation would deliver significant savings on key metrics. Most significantly, rollover at the new center would be eliminated and carry-over reduced 70%.

In the future, Pitney Bowes plans to develop the logistics analysis tool into a real-time alerting and decision support tool – to deliver predictive analytics in logistics. Simulation runs every day or hour would consider current backlogs, volume forecasts, and planned resource scheduling to determine near future needs such as extra labor, trucking, and rerouting requirements.

The presentation of this logistics analysis tool case study was given at the AnyLogic Conference 2021 by Pitney Bowes Data Scientist Cora Gao.

Similar case studies

More case studies

Get a brochure with industry case studies