Managing Risk in the Supply Chain with AI

By identifying unknown variables and incorporating them into forecast models, machine learning makes supply chains more risk-resilient

istock 1168365115
iStock

Given the global upheaval that COVID-19 has triggered, managing supply chain risk is top of mind for everyone. Improving forecast accuracy to lower risk is the sweet spot for machine learning (ML) applications in supply chains.

ML is an AI application that looks for patterns, trends, and anomalies in data, the quality and accuracy of which automatically improves with experience of the system. Specifically, ML algorithms built into supply chain management platforms enable predictive risk management that accounts for unknown factors, which is critical to maintaining the continuous flow of goods through the supply chain.

Managing the Unknown

Strong risk management requires predicting and accommodating both known and unknown variables. Unknowns could be a pandemic, changing trade relationships with countries like China, disruptions from natural disasters, shifting commerce regulations, the solvency status of supply chain partners, or changes to their production output, to name a few.

“When you find a dimension that’s not part of the core and add it to the model, you make predictability closer to reality,” Gurumurthy says. As a result, he notes, “you might find solutions to exceptions," such as previously unpredicted transit holdups, cost, and delivery problems, and avoid their costly repercussions.

For example, there have been many unanticipated repercussions of the COVID-19 pandemic. One was a sudden dip in apparel sales, because people haven’t been able to physically visit stores to try on clothes. “What ML will help you do, is to find the dimension(s) missing from the core forecast, identify the reason, incorporate it into the model, then make suggestions for reducing inventory in stores and warehouses,” says Gurumurthy.

Working up to Automation

Companies new to ML can start by having the system propose an actionable solution to a supply chain hiccup. Then, once they are sure the system has learned enough to suggest accurate actions, they can allow the system to make changes automatically for greater operational improvement. When a significant change occurs, the system can determine the impact on a company’s own key performance indicators (KPIs) and make immediate supply chain decisions tied to desired business outcomes.

Some modern supply chains have reportedly achieved improvements in forecast error rates using ML in this way. Those improvements bode positively for lowering risk. The better you’re able to forecast demand, production and delivery times, ability to commit to customer orders, and other variables—both known and unknown—the more resilient to risk your supply chain will be.

For more information, visit www.gep.com/software/gep-nexxe.

Copyright © 2020 IDG Communications, Inc.