A potential casino customer has a number of decisions to make.
Which casino to visit? When to go? How much money to spend? And where to spend that money—at slot machines or tables?
Similarly, a potential hotel guest must decide where to stay, when to go, how many nights to stay, what services to use—such as room service or the mini bar—and if and when he or she will return.
Predetermined answers to these questions could prove invaluable to businesses. And now, two Penn marketing professors and a former Penn Ph.D. student have developed a tool that will enable companies, specifically casinos, to answer these questions and identify and target their most lucrative customers.
Jehoshua Eliashberg, the Sebastian S. Kresge Professor of Marketing at the Wharton School, Raghuram Iyengar, an assistant professor of marketing at Wharton, and former Wharton Ph.D. student Sam Hui have devised a model for predicting gamers’ revenues.
In their study, Eliashberg, Iyengar and Hui plugged demographic data provided by a major casino operator into a scientific formula, and then assessed certain skills of different players. By doing this, the researchers have been able to provide casino operators with a better characterization of players and give them the ability to more accurately predict each player’s future revenue.
Their study, “A Model for Gamers’ Revenues in Casinos,” began in 2006 after a chance meeting with a casino executive. The team eventually examined more than 1,500 customers who made nearly 9,000 casino visits between 2004 and 2007.
Eliashberg says when they plug historical data into the model, they can answer questions about customers’ habits in the future.
“In other words, once the model is trained, we can use it now and ask when is [a customer] going to come next, to what casino, how much will [he or she] wager,” he says.
The key element their model exemplifies and captures, Iyengar says, is what they call a “latent skill” parameter. Unlike some skills that are obvious to the naked eye, casino employees can’t tell with any accuracy who’s going to be a good card player, and who’ll be a bad one. But these latent skills are important and can affect the quality of decisions a gamer makes, such as when to hit or stay in the game of blackjack.
For a casino operator to precisely predict future expected revenues, the study says the operators must separate “chance” from “skill”—and identify low-skill players who would most likely provide higher expected casino revenue.
“Only by modeling it and truly understanding how that skill level relates what you wager to how much you win or lose, and capturing that in a quantitative fashion, can uncover how skilled a person is,” Iyengar says.
If casino operators want to capture how much money they will keep from table games, Iyengar says they must understand that different people might have different levels of skill, and take that into account—along with house advantages and the split in slots and tables—before they can start predicting a customer’s value.
Some parts of their model can be used in other industries, such as the hotel business, says Iyengar, because, like the casino industry, hotels want to know when and where a person is going, and predict when a customer will return.
“So in some sense, while we don’t have things like how skilled is this person [in the hotel business] … many of the other questions that we’re asking also translate to these other industries,” Iyengar notes.
Eliashberg says some have criticized their study for aiding casino operators—but that is not the ultimate goal.
“For us, it’s a much more higher-level type question, which is how a company [should] identify the most valuable customer it has, and approach and target them,” he says. “This is a tool that helps quantify that value of your customer.”
Originally published on June 11, 2009