Many organizations faced the most disruptive year in their history as they dealt with the consequences of the pandemic. Large swings in supply and demand for different product categories and changes in distribution patterns left logistics managers grappling with the biggest challenge of their careers. Challenges that were made all the more difficult by one of the most basic yet complex questions in logistics, “How long will it take to get there?”
Accurate transit time predictions have become more crucial than ever given customer demands and expectations, the cost impacts of late shipments, and dynamic nature of today’s supply chains. Machine learning (ML) techniques can be applied to provide more accurate transit information and estimated arrival times (ETAs) by analyzing the historical shipment data in your transportation management systems. Moreover, these ML-based transit time predictions can feed into shipment planning and execution processes to provide better decision-making that ultimately yields better customer service and cost performance.
How machine learning is transforming logistics
Shippers and logistics service providers are faced with multiple challenges when trying to determine accurate shipment transit time estimates. They need to consider the complexities of each transportation mode, diverse geography of each region and route, and different operating behaviors of each service provider, along with other external factors such as unpredicted traffic, weather, and other disruptions that can occur in the transportation process. ML looks into historical data (for example, transit time statistics of carriers) and data from impactful external factors (such as port congestion, weather or holidays) and uses this information to develop more accurate transit time estimates. The model learns continuously and can adapt to changing conditions in the network.
- Predict transit times with more accuracy: Machine learning can predict end-to-end transit time despite the complexity of the logistics process, with shipments passing through multiple transportation legs that may include multiple carriers, various modes of transport, and across different countries and regions.
- Take events into consideration: Machine learning uses historical trends and events that can impact transit times, and use this information to provide predictions. These can include traffic conditions, port congestion, storms, and holiday closures.
- Incorporate changing business conditions: Machine learning can automatically account for changing business conditions, including new ship-to locations and changes in service provider’s performance level.
- Produce customized results: Shippers can create models specific to their business scenarios (mode, geography, business unit, etc), identify influential factors and fine tune each model for accuracy and performance.
How machine learning benefits your logistics network
Increased lead time accuracy reduces risks involved in transportation and logistics, improving your overall supply chain. It allows shippers to reduce their operating costs, optimize capital, allocate resources more efficiently, and can lead to higher customer satisfaction, increased revenues, and even improve their competitive advantage.
- Make better transportation plans: Shippers and logistics service providers can use machine learning to enable better decision-making and to alleviate exceptions. For example, improved transportation lead time predictions enables more informed decisions on which carrier, route and service level to use for any given order.
- Keep customers happy: Shippers and logistics service providers are able to provide better ETA predictions to end customers, alert them of any potential delays, and take actionable measures for projected service failures. It can minimize the number of actual delayed shipments by making better planning decisions upfront before the orders are shipped
- Lower inventory costs: Increasing the accuracy of transportation lead times will reduce safety stock levels and warehousing costs by eliminating unnecessary inventory that’s used as a hedge against transportation uncertainty. This is crucial especially for expensive goods, products with a short shelf life, and items that are dependent on seasonal demands.
- Reduce supply chain risk: Lower unplanned delays means less need for expedited shipments and associated extra freight costs. Companies are able to allocate resources more efficiently.
The pandemic remains present and its effects will continue to affect the way supply chains work, but the one thing that won’t change will be the importance of accurately predicting transportation lead times. Customers will continue to be more demanding and disruptions from competitors and technology innovation will drive expectations even higher. Leveraging machine learning will not just improve your ability to predict more accurate shipment times, but also reduces potential risks that may make or break a shipper’s bottom line.
Derek H. Gittoes is vice president for supply chain management product strategy and leads the team responsible for Oracle’s portfolio of cloud products for logistics and order management. Derek has 27 years of experience in supply chain software. Prior to joining Oracle, he was vice president of product solutions at G-Log, where he was responsible for product marketing. Derek was also a founder and managing director of Transport Dynamics, a company that specialized in the development of optimization software for transportation companies.