Load – Carrier assignment through Machine Learning approach

Industry: Retail Logistics


The client is one of the largest retail-chain store. Its logistics division has an issue that is unique to their operations.

The division is responsible for the shipment of goods from vendors to the client’s distribution centers. The majority of all the logistics communications happens through an automated system and relies on EDI message exchange. However, there are certain times when for specific truck loads the automated system cannot be used and they are forced to manually communicate – phone, email, fax – to arrange shipment of a load.

The manual process is very time consuming, and further since the Carriers know that client only contacts them under these circumstances they typically charge a premium for their services. Client also believes that a shipment getting delivered on-time is very important factor to maintain service-levels.

Client wanted us to look at how we could help them reduce the time and cost associated with scheduling these loads for transport. Client needed a ranking system for carriers based on delivery cost and service efficiency. Client can use this ranking system to assign loads to carrier.

Project journey

To address the above problem, Abzooba with its extensive expertise in data science and machine learning, created a customized solution for this unique problem.

Objective of the project:

Analyze loads which get manually assigned and develop a Dynamic Ranking Algorithm for assignment of Carriers.

To start with, we had access to following data.

  • Transportation database – load assignment, payment details for 6 months
  • Shipment status database – delivery details of shipment
  • Client – Carrier’s yearly contract details.

Approach overview:

The following are the steps Abzooba took to solve the problem.

  • Exploratory Data Analysis – understanding of data & selection of variables
  • Cluster Analysis – Identification of similar behaving Carriers based on Service performance, Capacity performance, Pricing and Load characteristics.
  • Dynamic Ranking of Carriers based on Load Characteristics and selected Performance Criteria

Using Exploratory data analysis, we have identified following metrics we use for Cluster analysis.

We have calculated these metrics for each Carrier. We used the above metrics to identify the following Carrier segments.

Workflow of Load assignment through dynamic Ranking of carriers:

Business benefits:

With this implementation, Client has benefited with the following:

  • Ranking of the carriers has become dynamic which is based on last 6 months Load delivery performance.
  • Business users can easily find the right choice of carriers based on the Load characteristics.
  • Load assignment process is based on both Service and cost performance of the carriers.