How one of the world\'s leading game companies used Socialesque to transform raw data and uninformed hunches into game-changing insights about player engagement.
Social Media Week Presentation from Social2B - Metrics, Measurements and Sca...
Online attrition in MMOGs
1. CASE STUDY
THE RULES OF ENGAGEMENT
How one of the world’s leading game companies used Socialesque to transform raw data and
uninformed hunches into game-changing insights about player engagement.
OVERVIEW CHALLENGES
When the game developers behind one of the gaming The scale of the data was the first major challenge. With
world’s most well-known franchises saw attrition rates more than 250,000 avatars in the game, identifying useful
in their MMO continuing to rise week after week, they trends and targeted data analysis required a highly scalable
knew they had a problem. They just didn’t know what it and powerful system for managing all that data. The tools
was. Everyone had their own theory, but without hard required to dig into such massive amounts of data were
data to guide decision making, paralysis set in. Each well beyond the resources that the developers had on their
team continued down the same well-worn paths without desktops. And while commercial tools such as Google
addressing the core problem. Analytics have their place, they’re simply not adaptable for
the needs of game developers.
While the marketing team pushed developers for the
information they needed, the development team was simply Apart from handling massive amount of data, here are a
overwhelmed by the day-to-day demands of keeping the few challenges that we think are relevant to gaming.
game up and running. And even if they’d had the luxury of
• When data is constantly evolving in a dynamic
more time, it wasn’t entirely clear where to start. Just like
environment, even a week-old snapshot of the game
any successful MMO or social networking application,
becomes out-of-date so quickly that the insights it
the game was producing so much raw data that merely
offered would be of little use.
getting it into a workable format was a challenge.
• Measurement is an ongoing process, not a one-shot
More importantly, the team didn’t know the right
proposition. Making repeated measurements based
questions to ask. While the data can tell many stories,
on all users for the entire lifecycle of the application
there are only a handful that really matter. And that’s
introduces new obstacles.
where things get really interesting.
• Not knowing which information you’ll need in
In a span of two weeks, starting from raw data files and
the future makes extracting the right data from the
a few discussions with the right people, we helped this
beginning that much more important.
company identify and understand the factors contributing
• The ongoing collection and maintenance of data
to player attrition. Today, they are able to closely monitor
generated by tens of thousands of users isn’t enough—
attrition-related data using a series of simple, targeted
mining the data for meaningful metrics is what it’s all
tools that give them the visibility they need, when they
about in the end.
need it.
2. complicated trifecta for most development projects. This
WHAT WE DID
means that you either measure less and expect mediocre
We started with a series of massive log files containing analysis or you measure everything but don’t know where
avatar names, basic activity reports, “Whispers”— to look when problems arise. Our approach allows you
messages from one avatar to another, including time to move fluidly between all three, using the right tools
stamps, “Active friends”—links between avatars, and within the right context.
more. Some of these files had over 1.5 million records.
From here, we analyzed the data using common— RESULTS TO DATE
and some not-so-common—social science principles
Social network analysis is even more intensive than the
designed to uncover useful nuances in the data. As an
traditional statistical analysis. This is partly because of
example, while we were able to map friendships using
the inherent network effects that must be addressed to
links between avatars, we were more interested in
understand the topography of the social graph. Rather
understanding the strengths of those friendships. We
than just counting ‘number of friends,’ the company
created a measure of “cohesiveness” using a number of
now understands the true drivers of their game. Today
other data inputs, allowing us to more accurately gauge
they can monitor the health of the game using a set of
the real impact of social connections within the game.
advanced dashboards that we built for them. Instead
The most important question for us to answer was of reacting to subscription drops, the company is able
whether or not we could assess and monitor the health to predict and prevent attrition within the game—and
of individual social networks to detect which were directly improve their bottom line.
in danger of fizzling out. Put simply, if your friends
begin to leave the game, does that mean you probably
will, too? While the answer may seem self-evident,
being able to determine which social networks are at
risk opens the door to even richer insights. Are specific
content challenges leading different networks to fail?
Are there shared attributes among failing networks that
can be addressed by the game? Hardware problems?
Understanding these linkages is crucial to understanding Socialesque specializes in measuring
audience engagement.
engagement, with a focus on gaming
and social networking applications. Our
OUR APPROACH proprietary large-scale data mining and
visualization technology helps deliver the
Our “Measured Design” approach helps you tackle this
challenge. In our view, there are three kinds of analyses insights developers need to make richer,
or searches: naïve, primed, and focused. Naïve search is more engaging content—maximizing the
exploratory—when you don’t necessarily know exactly
business value of virtual human networks.
what you’re looking for, but are following a hunch.
Primed search is when you have a general idea where We’d love to talk with you so we can help
something interesting might be, but you don’t know
you understand how to engage your
exactly what it will look like. Focused search is when
audience and get more from your game.
you really know exactly what you’re looking for and want
specific information about something—and you have a Please contact Varun Nayak at
very good idea where to look. varun@socialesque.com.
Building sufficient infrastructure to let you measure www.socialesque.com
everything without overwhelming your resources or
hamstringing your analytic options is an exceptionally