The algorithms big companies use to manage their supply chains don’t work during pandemics

The algorithms big companies use to manage their supply chains don’t work during pandemics

May 9, 2020 Off By stuartsupplies

Even during a pandemic, Walmart’s supply chain managers have to make sure stores and warehouses are stocked with the things customers want and need. COVID-19, though, has thrown off the digital program that helps them predict how many diapers and garden hoses they need to keep on the shelves.

Normally, the system can reliably analyze things like inventory levels, historical purchasing trends, and discounts to recommend how much of a product to order. During the worldwide disruption caused by the COVID-19 pandemic, the program’s recommendations are changing more frequently. “It’s become more dynamic, and the frequency we’re looking at it has increased,” a Walmart supply chain manager, who asked not to be named because he didn’t have permission to speak to the media.

Most retail companies rely on some type of model or algorithm to help predict what their customers will want, whether it be a simple Excel spreadsheet or a refined, engineer-built program. Normally, those models are fairly reliable and work well. But just like everything else, they’re affected by the pandemic.

“When you have something like COVID-19, it’s just a total outlier,” says Joel Beal, the co-founder of the consumer goods analytics company Alloy. “No model can predict that.”

Researchers have some understanding of how shocks to the system like natural disasters can disrupt supply chains and how impacts demand predictions. Disasters like hurricanes or floods, though, are usually regional. The pandemic is impacting the entire world. Even if companies stress-tested their demand forecasting models against diseases like H1N1 and SARS, they wouldn’t have accounted for something of this size. “This coronavirus pandemic is on another level entirely,” says Anna Nagurney, supply chain model expert and professor of operations and information management at the University of Massachusetts at Amherst.

Forecasting models usually use past data to predict future trends. If a company sold a lot of lawnmowers in April, they might use that data to tell the company to keep more lawnmowers in stock in April of the following year. Models can also typically assume that lawnmowers can be produced and transported on a certain schedule.

The radical changes in people’s behavior, transportation, and production during this pandemic mean that the usually predictable ebb and flow is upended. “Now we’re gonna have so many outliers in terms of the data,” Nagurney says. “Everything is shifted.”

Because of the massive, worldwide disruptions, the normal data feeding the models — which include buying patterns over years — aren’t as relevant.

“You’re probably going to not use as much historical data or will not be weighing that as much as you expected,” Beal says. Instead, companies are likely using much more recent data: looking to last week to predict next week, for example, or just relying on the few months of information on what was purchased since the pandemic took off worldwide.

The models can still be used. “It’s the data that you input that has to be changed,” Nagurney says. Companies like Walmart and Amazon that use more complicated machine learning models will also likely ramp up the amount of uncertainty that’s built into their systems, she says.

Those adjustments allow companies to continue forecasting. The predictions they make now, though, aren’t going to be as precise as the ones they were able to make a few months ago. “They’re not going to give us the accuracy that we’ve seen before,” says David Simchi-Levi, professor of civil and environmental engineering at the Massachusetts Institute of Technology.

Instead, people who manage supply chains will have to more actively interpret the projections, Beal says. “Companies have to rely more on good demand planners and forecasting people, who will say, ‘do I believe this?’ Rather than believing these models will be able to capture everything that’s going on.”

Alloy, for example, works with a company that saw sales for its product go up by 40 percent at a major retailer in March. (Beal couldn’t disclose the names of the company or retailer.) The retailer placed a huge order for April in light of that spike in sales, but the company knew that demand for the product had already crashed back down, and the retailer wouldn’t be able to sell everything they’d ordered. “That’s what we’re seeing over and over,” Beal says. “A lot of these systems haven’t caught up.” In this case, the company told the retailer not to purchase that much of its product, and they were able to adjust.

Some companies are changing their systems to account for the pandemic, Simchi-Levi says. He’s working with a company that’s trying to combine models that predict the length and severity of the COVID-19 outbreaks in various countries with their usual supply chain machine learning models.

Supply chain models will also have to change to account for the pandemic even after it passes. “This is a period I’m probably not gonna want to be using what I’m predicting what’s gonna happen next year,” Beal says. In addition, people might continue to buy things like toilet paper and beans at different rates than they did before the pandemic, so some changes might stick around longer than the crisis, he says. “We’ll have to understand the new steady state.”

The disruptions to modeling systems during this pandemic show some of the limitations to relying on computers to predict the demand for products. “Most companies struggle with it and it’s an ongoing challenge, even in ‘normal times’,” Beal says. The pandemic might push companies to invest fewer resources in demand forecasting and to focus more on responding to what they see in front of them. “It’s a shift away from thinking that you can predict what the world’s gonna look like months down the line,” he says.