Challenge-based research can open new avenues of innovation

This is an excerpt of the original article. It was written for the November 2021 edition of Supply Chain Management Review. The full article is available to current subscribers.

November 2021

This is the last regular issue of Supply Chain Management Review for 2021. Normally this time of year, I look forward to what’s in front of us. That’s turned out to be a fool’s errand over the last year and a half. So, instead, I looked back to see what I wrote this time last year. My column was titled “COVID hasn’t stopped supply chain progress.”

Browse this issue archive.

Already a subscriber? Access full edition now.

Need Help?

Contact customer service

847-559-7581   More options

Not a subscriber? Start your magazine subscription.

Route planning is essentially an optimization exercise that uses modeling techniques to optimize a route for a certain objective, such as minimizing cost or distance traveled. But this traditional approach to route planning falls short in one key respect: It doesn’t sufficiently capture how driver know-how can influence the efficiency of delivery routes.

A research project we recently completed aims to bridge this gap by using data-driven methods such as Machine Learning to incorporate driver behavioral factors in route planning. The project represents a departure from established optimization modeling approaches. It also demonstrates how an innovative approach to research—the use of public challenges to engage a wide community of researchers—can unlock new knowledge about supply chain planning and performance.

Taking directions from drivers

Optimization models have done a good job in helping companies improve the efficiency of product delivery networks. However, there are variables that are difficult, if not impossible, to encode in these models. One standout example is driver behavior. The use of Machine Learning to capture drivers’ specialist knowledge may increase the quality of route plans by better reflecting the real-world environment in which delivery vehicles operate.

A typical example of a human intervention often ignored by traditional route plans is a delivery driver’s decision to avoid a certain urban area at a certain time. Even though this “off-book” change may add miles to the route, the driver probably has a sound reason for altering the official route plan. Perhaps he or she knows that parking is particularly scarce at the target location during the period concerned. Maybe a customer will be unable to receive the delivery if the truck arrives at the designated time on the official schedule.

Machine Learning methods present opportunities to leverage data to detect such behavioral patterns by analyzing how drivers operate. Traditional optimization-based analyses might not be able to identify or interpret such patterns, especially when looking at route execution data for certain areas of demand across an extended time period. A similar logic can be applied to other variables unknown to route planning systems, such as unexpected weather conditions or traffic delays.

The advance of Machine Learning techniques in supply chain management makes it possible to sensitize models in this way. Also, thanks to increased levels of digitization in the supply chain and logistics domain, today’s modelers have access to greater volumes of data to work with and these data sources continue to evolve and grow.

SC
MR