Iyengar, who is faculty director of Wharton Customer Analytics, notes that this enormous amount of data collection was made possible by the migration of video games to an online format about 20 years ago. “Lots of gaming [involved] basically selling DVDs or other physical assets to a Walmart or a game store, but much of that has completely changed…. It’s now a direct connection to the consumer.”
Gaming companies pay a lot of attention to the data they collect, structuring and using it to make business decisions and serve customers better, he notes. It’s a heavily data-driven industry. And of course, game firms are always looking to dig deeper into the data to boost their competitive edge.
One of the giants of the gaming industry, Redwood City, Calif.-based Electronic Arts (EA), recently challenged students at the University of Pennsylvania to come up with data-fueled insights for its business. The Electronic Arts Datathon, hosted by Wharton Customer Analytics and held in late February, tasked the student teams to come up with ways to provide more personalized game recommendations for players and drive subscriptions to its suites of titles.
The judging panel comprised EA executives and experts at Wharton Customer Analytics. Three winning teams received prizes including gift cards, gaming subscriptions, and a chance for some valuable one-on-one time with Electronic Arts executives.
Inside EA’s Data Analytics Approach
Speaking at the event, Ben Medler, a principal data scientist at EA, provided a window into how his company uses customer analytics. “We do something pretty simple: We listen to and observe our players,” he said. “We can then transform the way EA develops and markets our games.” The company accomplishes this through its Global Analytics and Insights (GAI) division, which contains quantitively focused teams that do marketing analytics and data science, and qualitative teams that focus on consumer insights and user research.
Medler spoke about the interaction between the GAI department and EA’s game designers. He said that while the game designers are very good at “creating large immersive experiences” and at coming up with interesting features based on their intuition and knowledge, they don’t always fully realize the potential impact of those features on players. The GAI teams, said Medler, are there to “look at a player population and figure out what they expect and how they’ll react.” Those insights are shared with the designers to help them tweak and improve what they add into the platforms.
EA’s data science capabilities are also applied on an individual customer basis. For example, the company tracks how well customers are performing in the games. It can then generate targeted recommendations, such as more competitive playing modes for high-skill players, and tutorials for those who are struggling. EA has a dynamic messaging team that uses various channels to communicate directly with players.
Medler described how EA’s data analytics functions were outsourced in the past. Then, even after the capability was brought in-house, each data scientist was siloed to work only on one particular game or franchise. That changed for the better, he said, when data science was brought under one roof as the GAI division.
The company began investing more in improving training and communication among all the different teams, beefed up its data infrastructure, and started creating more intellectual property around data, he noted. “That has all resulted in further insights that will lead, hopefully, to even better games.” Medler added that the analytics people now all have the same strategic goals and “speak as one voice across the company for all our stakeholders.”
“We do something pretty simple: We listen to and observe our players.”–Ben Medler
The Winning Teams
The first-prize winning team’s presentation was titled, “User Preference Prediction and Game Recommendation.” The students used data science to create a Game Attractiveness Timeline tracking the popularity of different game genres such as action, adventure, racing, family entertainment, flight, and sports over the past decade. They were able to make predictions about customers’ favorite genres and their likely preferred games. The team also compared the numbers of players attracted to various games and the numbers of hours people spent playing them. Among the team’s insights was that “shooter” games typically draw a greater number of players, but sports games inspire users to play for a longer duration.
For customer game recommendations, they suggested enticing new users with the games revealed to have been the most popular over the past year, and for registered users, the games in their apparent preferred genre that they haven’t yet tried.
On the topic of subscription strategies, the students observed that few users are likely to select an annual subscription as their initial choice. One-month plans were found to be the most popular, so the recommendation to EA was to increase promotional efforts around that option.
The second-place team focused specifically on customer retention. In their model, they segmented users into two groups: those who had been subscribers for fewer than 50 days, and those who had subscribed for 50 to 360 days. They used a statistical model known as logistic regression to estimate the individual effects of selected game features on user retention.
Among their findings were that certain games—namely The Sims, Mass Effect, and the FIFA series—were among those with the highest positive impact on user retention. They recommended that EA invest more in customer service on the top “retaining” games, and to diagnose pain points of lower-retaining games to identify weaknesses.
“It’s not just about the model you built or the results you have; it’s really about the ‘so what.’”–Jon Delikat
They also found that users who play a greater number of unique games are more likely to continue their subscription, and suggested that EA find ways to incentivize users to try out more new games on the platform.
The third-place team, in its presentation “Game Recommendations: A Story-Driven Approach,” asserted that a user’s decision to play a particular game is prompted by certain fixed, traceable drivers. “A good recommendation engine should not be one that overlooks these characteristics but actively accounts for them,” they stated. They identified four human factors to study: Popularity: wanting to play “the hottest game on the street”; Social: wanting to play games that one’s friends and social circles play; Content: wanting to play games of a similar genre to those you played before; and Favorability: wanting to play games that have received good ratings and reviews. The team delved into the available data to try to isolate these factors, and made company recommendations accordingly.
Asking the Right Questions, Telling the Right Stories
Jon Delikat, a senior manager in EA’s GAI division and a Datathon judge, noted that for any data analyst’s presentation, “It’s not just about the model you built or the results you have; it’s really about the ‘so what.’” He encouraged the student participants to always weave a narrative that draws to a reasonable conclusion, know their audience, and strive to communicate their findings successfully to people outside the field.
Wharton’s Iyengar agrees. Even though there are talented individuals who understand data science, can run the models, and do all the analysis, many have trouble when it comes to “standing in front of decision-makers and telling a story.” His hope is that events like the Datathon will help train potential data analysts to speak the language of both analytics and the executive suite.
This article was first published in Knowledge@Wharton
Stay updated with all the insights.
Navigate news, 1 email day.
Subscribe to Qrius