Page 513 -

512 Part Three Key System Applications for the Digital Age

Zynga Wins with Business Intelligence

CASE STUDY

T he world’s fastest growing gaming up in a given month, that would result in a viral
company doesn’t boast top-of-the-line
coefficient of 1.05.
Expected revenue per user is an estimate of the
graphics, heart-pounding action, or
masterful storytelling. It doesn’t make lifetime revenue that a game player will generate,
games for the Playstation, Xbox, or Wii. The com- based on an estimate of monthly revenue per user
pany in question is Zynga, and if you have a and the churn rate. For instance, if the average
Facebook account, odds are you’re already well monthly revenue is $5 per user and the churn rate is
aware of its most popular games. Zynga’s explosive 50 percent, the expected revenue can be estimated
growth illustrates the potential of social gaming and as $5 the first month + $2.50 the second month +
the ability of social networks to provide critical data $1.25 the third month, and so forth, or approximately
about a company’s customers. $20.
Founded in 2007 by Mark Pincus and a group of The first wave of social gaming applications on
other entrepreneurs, Zynga is the leading developer Facebook tried to increase the viral coefficient with
of social network games, such as CityVille, Texas Wall postings advertising in-game actions by players.
HoldEm Poker, and FarmVille. These games, along This approach created too much “Wall spam,” or
with Zynga’s Empires & Allies game, are the four game-related postings that made it difficult for
most frequently used applications on Facebook. social network users to identify posts by friends.
Zynga’s games have over 290 million monthly Facebook and other social networking platforms
active users and 65 million daily players whose then demanded that gaming firms reduce their Wall
gaming keystrokes and clicks generate 3 terabytes spam.
of data every day. Since its inception, Zynga has put As a result, Zynga turned to social graph analysis.
a priority on data analytics to guide the manage- For social games, the “social graph,” or relationships
ment of its games and the business decisions of the between friends, is somewhat different from that of
company. the social networking platform itself. For example,
The company relies heavily on its data to improve in Zynga’s Mafia Wars game, players might have
user retention and to increase collaboration among two types of friends—those who actively play the
its gamers. In the words of Ken Rudin, chief of game and a more passive group that signed on to
data analytics at Zynga, to be useful, data must be help expand a friend’s Mafia organization and then
“actionable”—it has to be information that allows leave the game or play very infrequently. Players
Zynga to make noticeable improvements to its don’t always interact the same way with these two
games. Generating and storing game data is only half groups, with gifts and offers of help more frequent
of the battle. Zynga also uses two analytics teams— within the active group. Guiding game players to
a reporting team and analytics team—to work with communicate appropriately with these different
the data and make concrete recommendations for types of relationships helps increase revenue and
business improvements based on that data. virality while reducing churn. A social gaming
There are three key metrics that drive the company such as Zynga will thus try to improve the
economics of social gaming: churn rates, the viral player experience to make every aspect of the game
coefficient, and revenue per user. Churn, which we more profitable.
discuss in Chapter 9, is the loss rate of game play- Technology from Vertica Systems, an analytic
ers. Social gaming can have an extraordinarily high database management company, helps solve this
churn rate, about 50 percent per month on average. problem. Vertica’s Massively Parallel Processing
That means that half the new players signing up for (MPP) architecture enables customers to deploy its
a game today will be gone in a month. analytics platform using industry standard hard-
The viral coefficient is a measure of the ware or cloud solutions as building blocks called
effectiveness of existing game players for draw- “nodes.” Users can build clusters consisting of 1, 10,
ing new players, an important capability for social or 100 or more nodes, putting thousands of proces-
network platforms. For example, if 100 Farmville sors, terabytes of computer memory, and petabytes
users are likely to cause 5 of their friends to sign of disk storage to work as a single parallel cluster.

MIS_13_Ch_12 global.indd 512 1/17/2013 2:30:35 PM