Using algorithms to predict when each cold and flu season will hit and what medicines customers will buy.
It’s a Saturday afternoon in the cold and flu aisle at a Walgreens in New York. Rameek Benson, a Brooklyn resident, is looking for the Robitussin.
“I have a stuffy nose, congestion, cough, everything basically,” Benson says.
Like a lot of customers, Benson has never really thought about how his cold medicine got to the shelf. But it can’t be that hard to do, he thinks. “You know what your customers want, and basically the hot products always sell,” he says.
Lucy Alvarez, another Walgreens customer, says she’s coming down with a head cold. She is buying a cold medicine called Zarbees, and she isn’t too concerned with how it got to Walgreens either.
“They take care of that,” Alvarez says. “You’re the customer, you come in, you get whatever you want, and you go about your day.”
“Honestly, it probably does look invisible,” says Dustin Sullivan, category manager for cough, cold, and allergy at Walgreens. Sullivan’s work may be invisible to customers, but he is one of the unsung heroes of cold and flu season. He’s part of an elite, eight-person team at Walgreens. All they do, all year long, is try to figure out when customers will get sick and what medicines they will buy.
It’s really hard, says Mark Elter, supply chain manager at Walgreens and another member of the team.
“You don’t know what that consumer’s going to choose when they’re not feeling well,” Elter says. “They may go into the store and pick up cherry this time. [But] the next time they go in, they could pick up grape.”
The cold and flu are also different every single year. The team doesn’t know when each season is going to hit and where it will hit first. Reuben Slone, senior vice president of supply chain management at Walgreens, likens the business of forecasting to rocket science.
“It is very sophisticated algorithms,” Slone says.
Old Wives Tales
The Walgreens corporate office in Chicago is modern, with glass doors and windows. There is a gym for employees and a doctor’s office on the second floor. Slone is a high-level executive in a fancy suit, but he has a big smile and Midwestern hospitality.
He says the team tracks the cold and flu through the South and up into North America. “It comes up from the Southern Hemisphere — Texas, Florida,” he says. “And then it can make a sharp left turn, go to the West Coast, or kinda right turn and go up the East Coast.”
The team combines that information with store sales and tries to recognize patterns. They predict things right about 80 percent of the time. But they’re getting better at it. They have a few secrets up their sleeve. To figure out when people are about to get sick in a certain town, they look at sales of other products, like orange juice, toothbrushes, and chicken soup.
The first thing people do when they’re getting sick, it turns out, is try to stave off illness by changing their toothbrush, chugging orange juice, and downing chicken soup. The effectiveness of such measures is questionable, but if the team sees toothbrush or OJ sales jump, they know they have to send more medicine to stores in the area.
Heat Maps and H1N1
Even when the team tracks basic things — like sales of cough syrup — they do it at warp speed. It’s a February afternoon, and Sullivan and the team crowd around a color-coded heat map of the United States that’s projected onto the wall. The map shows sales for every cold and flu item in Walgreens 8,000 stores.
Andy Kettlewell, director of inventory management at Walgreens, zooms in on New York. “We see a couple of things,” Kettlewell says. “One, Manhattan is on fire from a sales perspective. Total sales are really cranking out there. And then we can drill down even further.”
He clicks on the store where Rameek Benson bought his Robitussin. “We have a large increase, up 25 percent, in syrups,” Kettlewell says. “We also see that cold sores are up 300 percent in the store.” Walgreens is going to have to send more medicine to New York stores, and soon.
But this is an easy day.
Reuben Slone harks back to the 2009 flu pandemic. Cable news reports put people on edge, Slone says. “No one knew that people in the United States would react so strongly to what was coming over on the news,” he says.
Walgreens customers couldn’t get enough hand sanitizer, so the team ordered more. The Centers for Disease Control and Prevention also recommended that sick people wear surgical masks. The team bought a bunch of those too.
“People were panicking,” Slone says. “The manufacturing base built all these face masks and all this hand sanitizer and then suddenly, the crisis passed and there was a glut.”
The hand sanitizer sold even after H1N1 peaked. But the face masks are still sitting in a Walgreens warehouse.
“Face masks just don’t sell when there’s not a need for [them], Slone says. “They’re not considered a fashion-forward kind of thing. So we’re trying to figure out what we’re going to do with them.”
The H1N1 pandemic was a big moment for Walgreens. That’s when it realized it couldn’t just rely on algorithms. It needed a team that did nothing but track the cold and flu.
The challenge of predictions
What does Walgreens stand to lose if the team gets it wrong?
“A lot of money,” says Sullivan.
He wouldn’t say exactly how much money. But when Walgreens doesn’t have products in stores, it’s also gambling with customer trust, says Kettlewell. “If you’re sick, Walgreens has to be in stock on cough and cold,” he says. “We can’t be out of stock.
That’s a challenge, Slone says.
“The reality is there’s always some new pattern of behavior, some unanticipated event,” he says. “It’s very hard to predict, because it’s imperfect. We don’t know what’s in your head, and our hope is we have the right things in our store that you need.”
Rameek Benson, the customer we met at Walgreens in Brooklyn, says there’s always room for improvement.
There have been occasions when Walgreens didn’t have what he was looking for: gummy bears and gummy worms.
“But other than that, they’re pretty good,” he says.