With Predictive Analytics, Companies Can Tap the Ultimate Opportunity: Customers’ Routines

If knowing what customers need is marketing gold, pinpointing exactly when they need it may just be platinum.

Services that become part of a customer’s routine may deliver advantages beyond repeat business for a company, Harvard Business School Associate Professor Eva Ascarza and colleagues find in a new working paper.

“We find that routine customers have higher value to the organization, even after controlling for their level of consumption,” Ascarza says.

These customers may also tolerate price increases better and even stay loyal longer when things go wrong compared with customers who haven’t made a service part of their routines, the authors find.

These findings come as companies such as Procter & Gamble, Adidas, and McDonald’s are trying to collect more consumer data to hone their marketing messages. With artificial intelligence (AI) opening new possibilities in the noisy world of digital marketing, companies are looking for new ways to gain an edge with fatigued customers. Harnessing customers’ routines may offer a compelling new opportunity.

When services such as ridesharing are part of a routine—even if that routine isn’t obvious to the user—firms may be able to pinpoint a customer’s motive more precisely than for people who use the service casually or merely as a preference. That may help companies carefully tailor both marketing and service for their most valuable customers, the authors find.

Ascarza teamed with Ryan Dew from the University of Pennsylvania’s Wharton School as well as Columbia Business School’s Oded Netzer and Nachum Sicherman to develop the model that identifies routine users and their value.

Not all rides are routines

To track how targeting routines may work, the authors teamed up with a rideshare company in New York City and tracked some 2,000 users, homing in on passenger usage data between January and November in 2018. After a rider had been active for three weeks, the authors tracked how—and, more important, when—customers used the service.

The researchers then ranked service use times and plugged them into a statistical model to create a “routineness score” by layering them on top of seven-day periods. For example, the model groups someone ordering a car at 3 p.m. and someone ordering a car at 4 p.m. as more likely to have similar routines than two people who call for rides at 3 p.m. and 3 a.m., respectively. The authors used that information to predict how often and when a customer may request a car as part of their routine.

The model could drill into specific kinds of routines, too: The model identifies seven clusters of typical ridesharing routines in these data, including “the morning commuters,” who tend to use the service Monday through Friday morning, or “the weekenders,” who mainly use Friday and Saturday nights and Sunday all day. With that information, the authors could then figure out who likely booked a ride as part of a weekday commuting routine versus a ride to get to a class on Tuesday and Thursday afternoons, and even which rides were not part of any routine.

All kinds of service companies may benefit

One reason routines—and the ability to predict them—is important is that routine customers may tolerate service changes, like higher prices, or glitches, such as a delay in pickup, better than customers who don’t use the service as part of a routine, the authors find.

“Services from streaming video to online retailing might profit from using the model because detailed data can be closely tracked and parsed to detect routines.”

“And the findings don’t just apply to ridesharing. Services from streaming video to online retailing might profit from using the model because detailed data can be closely tracked and parsed to detect routines,” Ascarza says.

More services and products than ever are now sold online. That means the method lends itself to companies trying to identify the way customers interact with a service. Companies like Netflix or Amazon, for example, might be able use the model for more effective timing or services, Ascarza says.

“Imagine, for instance, how you connect to e-learning. When you connect to the platform, [a company] can see when you have a routine behavior versus a sporadic one,” Ascarza says. “That could be very valuable for managers, who can then make sure that [firms] engage with customers the right way—to make sure that they actually develop these expectations and they fit the needs of the customers.”

Firms can explore how to tailor a service—be it exclusive viewing or a specific product or service sale—to draw in more routine customers. More research on how that may work best is needed, Ascarza says.

Not just companies

The method of tracking routines may have implications for policymakers, too. “Once you have a model for routines, of course, you can better predict what will happen in the future at an exact point in time,” she says.

“Understanding patterns of behavior in transport could be super relevant for policymakers.”

Take transportation. Knowing when drivers are taking a route as part of a routine—like commuting to work—rather than a sporadic trip to the office could prove invaluable in designing flows in and out of a city’s center.

Another example: Designers developing new infrastructure for electric vehicles might benefit from knowing driving routines in a geographic area. Even bicycle share services could use the method to plot when enough bikes are needed to cut congestion or improve bike traffic flow. The next step, Ascarza says, may be to study how people who are traveling as part of a routine would respond differently to policy interventions—like the introduction of green alternatives or a new road’s design—versus those who merely use the system from time to time. “Understanding patterns of behavior in transport could be super relevant for policymakers,” Ascarza says.

You Might Also Like:

Feedback or ideas to share? Email the Working Knowledge team at [email protected].

Image: iStockphoto/kate_sept2004