Las Vegas-based software and data-platform company OPTX has unveiled its latest AI feature. OPTX is calling Future Player Value a “game changer” for the casino industry.
OPTX’s data platform focuses on ingesting the information from the disparate systems within a casino and turning that into operational functionality in the areas of slots, marketing, and player development.
According to Steve Bright, OPTX’s vice president of data science, the new feature is a machine-learning model that uses both game session and demographic data to predict player spending.
OPTX measures the success of Future Player Value (FPV) by comparing against current casino best practices for predicting player spending. When OPTX compares model predictions with actual future player spending, it finds the model provides a reduction in the prediction error of more than 30% compared to the current business practice of using player spending during the previous period, Bright said.
“The motivation for why we created the Future Player Value model is that currently, a casino has to make marketing decisions on who should get what offer and what they should be based on, solely on their past history,” Bright said. “Almost always, there’s this underlying assumption of some free-play amount, because their expected spend is going to be $1,000, for example. They know that $1,000 was their spend in the last 28 days or on average over the last 90 days. That’s not an unreasonable starting point, but it leaves a lot to be desired.”
While people have spending patterns, they can change their behavior over time. Trying to predict individual behavior is difficult, given all the reasons people do what they do. On the other hand, casinos have a lot of information about their players — age, residency, gender, and game preference and behavior.
“They have the data, but it’s not reasonable to expect an individual casino to crack open Excel and look at over 100,000 rows of individual session behavior for all their players and make some gut-feel guess of what happens next,” Bright said. “That’s hard for a person and not something anyone would want to do. Fortunately, it’s very easy for a computer to do if it’s programmed by people familiar with machine learning and AI, which we are.”
OPTX has built a machine-learning model that uses demographic data, player history and preferences, and time-period behavior to predict the future value of that player for slot-machine and table-game play.
“We suggest to the operator that rather than deciding how to treat this individual in your marketing campaigns based on what they did last month, you can now decide how to market this player based on what you think they’re going to do next month,” Bright said.
Because OPTX has so much historical data, they can test their prediction that future behavior will be more accurate than the previous data. Bright said OPTX can “run the clock forward” 28 days and see if the prediction is correct.
“We found that our prediction is generally 30% more accurate than looking at previous behavior,” Bright said. “I think that 30% will get even better as we incorporate more data sources and get user feedback.”
Bright said not only does FPV predict behavior, but it also informs what offers a player may respond to – a hotel stay, food and beverage, or type of concert. Those are dependent, however, on integrating the data into the system.
“Once we lay the groundwork for the predictive modeling and lay the foundation for incorporating these modeling results into OPTX’s campaign flow, we have a complete solution based on machine-learning and AI that we can improve upon incrementally as we incorporate more data sources,” Bright said.
Bright predicts a “complete automation” of this marketing process. There’s still a lot of “drag and drop and point and click” in comparing tables, but Bright said they’re making headway in reducing the effort. Eventually, it will require only pushing a button.
“Our clients becoming engaged in using AI and machine learning has the added benefit of stimulating their desire to understand their own data source, which brings an additional sense of discipline and urgency in how to enter data and location specification of a machine,” Bright said. “When a client starts this journey with us, they often don’t have any processes or controls in place. When they see how errors in the data can affect results, we can prove this to them. Managing data is a source of revenue for them. That virtuous data cycle helps make the overall casino-gaming experience better for the player and that’s what it’s all about at the end of the day.”