<img height="1" width="1" style="display:none;" alt="" src="https://dc.ads.linkedin.com/collect/?pid=358169&amp;fmt=gif">

3 Practical Examples of Machine Learning Transformations in Supply Chain

 

Machine learning (ML) is critical to today’s complex, data rich supply chains.

While machine learning is not a new technique, over the past few years there has been a resurgence in its importance driven by: 

  • Cloud computing: easier and cheaper availability of computing power required to build and train ML models  
  • Increase in data volume, velocity and veracityThe sources and amount of data companies capture is growing at an exponential rateStructured and unstructured data sources such as IoT, point-of-sale, social sentiment, weather, construction data and more help accelerate better decisions and gain visibility into the supply chain  
  • Global and interconnected supply chains: We now link people, places, things and time across all layers in the supply chain creating previously unthinkable complexity 

According to Gartner “static algorithms that heavily rely on human intervention to analyze new data, look for significant patterns, arrive at insights and execute appropriate actions are woefully insufficient. What is needed is a technique that self-adapts and learns from new data – and that is precisely what an ML algorithm is capable of.” (1)  

There are many powerful and common examples of machine learning used today in the supply chain. To learn more, read our recent post Top 5 Machine Learning Use Cases in Supply Chain

Over our course of working with companies to help them grow in their use of ML to support their digitalization initiatives, we have identified five key areas where ML can really shine: 

  • Automation  
  • Improve supply chain productivity 
  • Better predict an event happening  
  • Improve supply chain data quality  
  • Advise the best “course of action” to take given a current situation  

 

Machine Learning Sets Companies Up for Success 

There is a lot of hype surrounding machine learning and this can lead to both a lack of clarity into its potential benefit and a reluctance for companies to take the next step and experiment with ML. Part of this may be due to the lack of real-world examples of ML in action to help you visualize the benefits.  

John Galt has many success stories of how machine learning advanced a company’s supply chain operations. Here are three examples:

 

Oil & Gas Company 

A company specializing in bulk gas storage and delivery, cylinder filing, and community distribution gas systems with more than 90,000 customer locations saw an opportunity to automate demand and replenishment planning, and optimize their delivery transportation routing. Working with John Galt Solutions, the company brings in a variety of data signals including IoT weatherPOS (point of sale) and customer sales history to drive the supply chain plan. Telemetry data, for example, is sent by an IoT sensor placed on top of bulk storage tanks which allows the company to measure inventory levels at customer locationsUsing automated pattern recognition algorithms driven by ML, we capture, harmonize, and look across intra-day demand readings from the telemetry units, syndicated weather data, and usage data to predict demand. With more accurate customer orders, the oil and gas company can automatically re-plan replenishment needs and optimize truck route delivery.    

The oil & gas retailer also uses John Galt’s machine learning capabilities to identify bad data. We can automatically identify the root cause of the bad data and send prescriptive recommendations to the team.

 

Food & Beverage Company 

global flavor and spice manufacturer uses multi-echelon inventory optimization (MEIO) to automatically adjust their inventory position at each stocking location and SKU level by looking at all the nodes in the entire supply chain and taking into account all the variability such as cycle times, transportation lead times, WIP, and more. 

The company’s supply chain is characterized by combinations of shared and unique ingredients, from raw materials through finished goods delivered to customers. To address these combinations, the company models its supply chain to recognize purchase and distribution points for raw materials, along with distribution centers for blends of raw materials that are subsequently incorporated into other mixes. Transfer capability within the network is also recognized in order to minimize storage and transportation costs.  

Using MEIO, the customer can identify the optimal balance of inventory at the right locations and determine the optimal inventory parameters and position by location including intermediate warehouses and retail customer destination. This allows the global flavor and spice manufacturer to optimize their transportation routing, hold on to safety stock only in the most cost-effective locations and minimize the time from raw material acquisition to customer delivery.

 

Consumer Goods Packaged Company  

The world’s largest manufacturer and distributor of ice brings in POS (point of sale), IoT and weather data to increase forecast accuracy and improve the daily transportation delivery routings.   

IoT sensors were installed in iceboxes so that the company can measure how full iceboxes are throughout the day and monitor how much ice they hold in real time. The sensors transmit data to the supply chain planning platform every 30 minutesThe software analyzes this data along with POS to predict when iceboxes will run out. Using machine learning, the company automates demand sensing and adjusts order recommendations as each location sees a rise or drop in demand. 

The near real-time nature of the data helps the company identify anomalies, such as a spike (or drop) in demand. With John Galt Solutionsthe company adjusts demand on the fly to minimize delayed shipments and stockouts, and automatically re-plans replenishment-related decisions everyday (e.g. frequency, quantity, and routes). 

 The system also calculates if the truck has enough ice to meet new deliveries, then routes a truck to new locations or suggests an alternate delivery route. In addition, the company is able to see when more ice than expected gets delivered and to forecast potential shortages for final deliveries. For example, the system might recommend using a different truck for final deliveries and allocate the remaining ice. This helps reduce waste, save time and money. The system also monitors how many stops a truck makes and, if it begins to lag, recommends maintenance.  

 

How to get started

These are three practical examples of machine learning in use at supply chains today. The tools are available today to turn the mountain of data you have access to into real insights that drive the business. Just get started! 

 

Sources:
1: Machine Learning 101 for Supply Chain Leaders Part 1: How It Works and Its Relationships to Other Analytics Techniques, Gartner, 




Tags:     Blog Demand Planning Forecasting

Subscribe