Recipient preferences-based assignment of parcels to service points
Prediction & classification of websites by their structural & design components
Elad Sadan, Tel-Aviv University Advisor: Prof. Tal Raviv
Abstract:
The recent surge in e-commerce requires delivery companies to adapt and economize their logistic processes. In particular, the last mile of the delivery is associated with a large share of the costs and adverse environmental effects. Many parcels nowadays are delivered via attended or automated service points (SPs), which prove to be a more efficient, flexible, and sustainable alternative to home deliveries. However, increasing demand and limited capacities of the SPs challenge delivery firms to better utilize the delivery network, possibly by collecting additional information about the recipient's preferences and capitalizing on their flexibility.
This study proposes a novel operation policy for parcel delivery firms. Our policy is based on providing information about the availability of SPs in the recipient’s vicinity and asking them to select their desired pickup location. Different recipients may have different sensitivities to delays and pickup location preferences, which leads to different selections. In contrast, the common practice in the industry is either to select the delivery SP solely based on the address of the recipient or based on the selection they made at the check process at the e-store before information about the SP availability at the delivery is known.
To evaluate our preference-based parcel assignment policy compared to benchmark policies that mimic the common practice, we modeled a disutility function that represents the recipients’ inconvenience due to the distance from the pickup locations and delays in the delivery process. The disutility function also consists of random recipient-specific components unknown to the delivery company. We constructed a simulation environment attempting to mirror the processes of parcel delivery, focusing on the daily assignment of parcels to SPs.
Our simulation experiment was conducted in a realistic parcel delivery network at Linz, the third-largest city in Austria. We tested a combination of two levels of congestion in the SP network and two levels of recipient variability, mirroring possible real-world conditions. Our findings demonstrate the effectiveness of our preference-based policy, particularly in high recipient variability or moderately congested scenarios.
These results underscore the potential benefits of our proposed policy. By incorporating recipient preferences into parcel delivery operations, we can significantly increase the recipient’s satisfaction and improve the network's load balance between the SPs. Importantly, our policy is not only effective but also easy to implement, offering a promising avenue for enhancing service levels without the need for additional physical resources.
The study is co-authored with Tal Raviv (thesis supervisor) and Margaretha Gansterer from the university of Klagenfurt, Austria.
