One of the biggest challenges a gaming company faces is keeping its biggest fans engaged and playing. MobilityWare has some of the most popular titles from day one of the Apple App store. With over 200 million downloads of its Solitaire game, the company claims to know one thing, that it doesn’t know what players want. So they test, make hypotheses and then test some more.
MobilityWare taps behavior insights through predictive modeling and hypothesis testing. Data analysis is a key factor for developers when trying to make changes to retain good users longer.
Predictive Scoring & Modeling
How do you figure out when a customer is likely to quit playing–and which player incentives could delay or stop “churn”?
Chris Densmore, director of analytics for MobilityWare, focused primarily on “dedicated players,” who had installed the game more than two weeks ago, but who had not played in the previous two weeks.
“Dedicated players–they’re more valuable, and we have a larger behavioral dataset because they’re more likely to have played a lot in the past,” Densmore says.
“They can be incentivized, so there is something we can do to influence the outcome, which is important,” he adds.
Densmore and his team used data and analytic features from the Mobility Ware’s CDP. Then he chose a logistic regression model, because it can still produce valid results even if some variables are correlated—as they usually are in real, live users—and because logistical regression makes it possible to interpret contributing factors as well. Another bonus: the coefficients that come out of regression can be easily placed in SQL scripts to produce predictions.
Two key consumer insights were determined:
Coins Are Cool. But Boosters are the Bomb.
The team found that gifting players with in-game currency at the right moment retained players longer. Many would use this gift to progress past difficult levels that had frustrated them, increasing their enjoyment of the game and keeping them playing longer.
The real surprise was that giving someone a “booster,” or a small assist or tool they can use to win, was almost as effective as coins in combating churn, but the big difference was that it increased ARPU (average revenue per user) by more than 450 percent for those with a probability of churning between 60 and 80 percent.
Customers Will Pay More If You Help Them
Players that were given in-game currency experienced what is known as “loss aversion”–hoarding it because they wanted to save it for the best possible moment. Players that were gifted boosters had a “use it or lose it” mindset that did not stop them from fully utilizing their gift, thus increasing their enjoyment of the game while simultaneously giving them an experience of premium items.
Offering just a little bit of help—so it wouldn’t devalue the accomplishment of winners—was the key to reducing churn and renewing player interest in the game. The effort offered customer insights that opened the door on a better customer experience and improved monetization. With it, hypothesis testing and predictive modeling pointed the way to making a positive impact on churn, a tough problem many game publishers face.