Any Questions?

Case Study

How the diminishing profits of the client have been addressed through ML..

Client Background

A leading Logistics Group based in Coppell, Texas is the market leader in time-definite transportation services, both domestically and internationally. Since its inception in 1994, the client has been associated with Fortune 500 client base in providing 3rd party logistics services.


The client depends on 3rd party vendors for delivering their shipments. Many times on-time delivery of goods is mandatory to meet their customer SLAs and avoid penalties. Vendors sometimes fail to reach their scheduled delivery and this leads to potential revenue losses, increasing costs, and poor customer service – resulting in diminishing profits for the client. This uncertainty triggered a series of business and operational issues. The client has traditionally used business intelligence dashboards to monitor the performance of on-time delivery, but these systems mostly highlighted the root cause or post-mortem data analysis to identify gaps in the delivery process.


Seanergy understood that having a layer of analytics on top of standard delivery processes would enable data-driven actions in real time and this can be addressed using a machine learning model. A solution which can highlight the gaps in the delivery process and predict the on-time delivery of a shipment by considering some of the factors which influence the delivery as an input. Hence, after collecting a year's shipments data we built a proof of concept using ML which uses algorithm models such as random forest, support vector machine, and decision tree to process and train large volumes of data and help to identify the likelihood of late delivery.


The results of this ML approach was impressive: Using random forest about 90% instances of delays were accurately predicted. This percentage could be further improved by including parameters, such as vendors, service type, shipment notes, failure code and shipment location.

Solution Highlights


Increase in prediction accuracy