machine learning and ai
machine learning and ai
Thu 31/10/2019 - 17:43

Machine Learning Applications in Global 3PLs

Fueled by the rise of e-commerce, and the demand for expedited delivery times, the logistics market is currently amid a transformational renaissance. This change in the market presents new challenges, such as demand planning, network optimization, and real-time visibility for 3rd party logistics (3PL) companies. At the same time, this disruption in the logistics industry is serving as a catalyst for new technologies that facilitate and optimize 3PL processes. One emerging area of innovation is artificial intelligence (AI). Accompanied by a shift from theoretical research to commercial products and service, AI has grown to encompass nearly 340,000 inventions and over 1.6 million scientific publications. [1] Machine learning is the fastest-growing AI technique, comprising more than one-third of all identified inventions. [1] In other industries, machine learning has been established as a decision-making tool with great potential. However, it has yet to be embraced by 3PL and its overall impact on efficiency has not been fully investigated. Here we present a brief introduction into AI with a focus on machine learning and its benefits at GEODIS and other 3PL companies.


Overview of AI 

The term artificial intelligence was coined by John McCarthy, a computer science professor and Turning Award winner from Stanford who organized the first conference on the topic. In the last 60 years, improvements in computational power, data storage, and network capability, along with the advent of big data, has allowed for AI to grow into what many see as the next industrial revolution.

AI’s functional applications include everything from computer vision to natural language processing. Given its complexity, AI often gets thrown around as an all-inclusive term. Generally, AI refers to any man-made agent or machine which attempts to mimic human thinking by analyzing its environment and making decisions to successfully achieve its goals.

The rise of AI is accompanied by a shift towards consumer data collection and personalized content to better suit the needs of the customer. All these factors make AI a key factor in remaining competitive in today’s market.


Machine Learning Taxonomy

Machine learning (ML) is a branch of AI which is focused on the idea of building machines that process data to learn on its own without continuous supervision. It is characterized by a process of data analysis and pattern extraction to make decisions,

Machine learning is loosely subdivided into 3 sections: supervised learning, unsupervised learning, and reinforcement learning. Most of machine learning utilizes supervised learning. Supervised learning is the process of analyzing a labeled dataset to create a function that will make predictions about an unlabeled dataset. Conversely, unsupervised learning uses an unlabeled dataset to extract general rules about the data features.

Reinforcement learning and deep learning are some of the emerging subsets of machine learning. Reinforcement learning creates models based on a system of rewards and costs without using expert datasets. Deep learning is an extension of machine learning that adjusts the model with new insights from a neural network architecture. Deep learning models are typically created through a computationally expensive process that may require powerful CPUs or Graphics Processing Units (GPUs). However, they have a higher accuracy rate compared to traditional machine learning.


Machine Learning for GEODIS and 3PLs

Machine learning provides an ideal solution to logistics problems which is often constrained by time, cost, and resources. This is because machine learning’s greatest asset is its ability to identify patterns in data and pinpoint the specific factors that influence a 3PL company’s success. These models provide key knowledge and insight with the benefit of continuous improvement through learning. Machine learning supports the use of algorithms for real-time supply chain planning and execution while also allowing companies to explore what-if options using modelling capabilities.

There are a number of areas in supply chain where machine learning can be applied. For example, machine learning can help when predicting anomalies in supply chain performance and anticipate where Industry 4.0 automation can best be applied. Machine learning can also provide insights on how to use inspections and improve product quality and visibility. For example, it can be used to determine when auditing a product should be mandatory based off employee performance. This leads to an improvement in quality while maintaining warehouse efficiency. ML can facilitate demand planning and network optimization; it also can enhance customer relations using big data. Sentiment analytics and chatbots are just some examples of ML involved in customer interactions.

Here at GEODIS, we are integrating machine learning models into order demand forecasting, labor planning, and optimized scheduling modules. These forecasting models can provide accurate estimates for how many workers need to be scheduled on any day while also maintaining transparency for operational managers who will utilize these tools. Furthermore, this project allows GEODIS to implement new in-house software solutions that address specific business needs for each warehouse. Other retail companies are already implementing similar labor forecasting solutions in their delivery processes with astounding success.

In addition to these efforts, GEODIS is investing in data collection efforts from edge devices on the floor and at various points along the supply chain. This will unlock many opportunities soon to drive predictive programs like preventive maintenance on machinery, product quality audits, etc.

Machine learning models come at an affordable price now as most tools are open source. 3PL companies can make these investments by spreading out the costs between their diverse customer base. In this way, applied machine learning provides adaptive, value-driven solutions for complex supply chains.


Conclusions and Benefits

3PL companies have amassed a huge volume of untapped data from its operations. As the amount of data increases, 3PL companies need to make investments in technology that will allow them to analyze operations, reduce overall transportation costs, improve asset utilization, and provide better services. A 2019 study highlights that contemporary technology tools such as network modeling and optimization, web portals, cloud-based systems, advanced analytics tools, and data mining tools as frequently cited technologies by 3PL companies [2]. More than half of the 3PL companies in this study also stated that they would be investing in predictive analytics [2]. Shipping companies are utilizing data to optimize their networks and drive supply chain decisions [2]. As a result, data science applications have become a key selection criterion differentiating 3PL companies as potential customers [2].

The success of a supply chain company hinges on its ability to manage its resources efficiently amidst several unpredictable variables. The current market demands greater predictability of consumer needs and the capability to meet those needs appropriately and efficiently. Various research endeavors have already highlighted the major applications of AI in supply chain management (SCM). With the appropriate tools and mechanisms, 3PL companies can adopt the AI standard to solve modern warehousing, inventory, and consumer demand problems and improve their warehouse operations beyond what can be achieved by human intelligence alone.


Written by: Akhila Ashokan, Shibu Raj, Siddharth Krishnamurthy, and Jerimy Capps




[1] WIPO (2019). WIPO Technology Trends 2019: Artificial Intelligence. Geneva: World Intellectual Property Organization.

[2] John Langley Jr., C. and Infosys (2019). 2019 23rd Annual Third-Party Logistics Study: The State of Logistics Outsourcing.


Share it


Need help ?

Stay tuned