Tuesday, November 12, 2013
Ninja Metrics: Applying Predictive Analytics To Make Games More Profitable
In the world of online, mobile, and social games, there are now petabytes of data being collected by companies on what users are doing in those games--and trying to make sense of all that information. In a move to convert that huge pile of data into something much more meaningful, Los Angeles-based Ninja Metrics (www.ninjametrics.com) launched a new predictive analytics product for the industry. The company announced a first round of funding Monday for its technology. We spoke with CEO and founder, Dr. Dmitri Williams, on how a collection of data scientists ended up in the online, social, and mobile gaming industry, and how the company is turning those petabytes of data into what's important for companies: money.
Explain what Ninja Metrics is doing?
Dmitri Williams: NinjaMetrics is a data analytics company. We have a new piece of technology, which figures out social value, and delivers something to companies they never had before, which is the ability to measure influence among people, and to do so with hard numbers. It's not an abstract score, but an actual amount of money or sessions that those people generate for companies. This is a whole new kind of thinking. We spun this out of university research labs, where the founders and all of the team still are involved, and commercialized that research.
Talk about how your research led to this startup?
Dmitri Williams: Our team, which is still around, is the Virtual World Observatory (vwobservatory.org). It's a very successful research team, with over eighty peer-reviewed articles we've published. We started researching gamers and game data a decade ago. As we gained possession of big data sets we started putting computer scientists to work on that data, and we started doing big data before big data was cool. We started codingon this through federally funded research from the Department of Defense and intelligence agencies, the National Science Foundation, and others, and started some really efforts in big data algorithm development. We got really good at predicting what people would do, based on the data from those games. As we thought this out a few years ago, we realized that we were able to turn this into a company, because businesses were interested in predicting when people would start spending money, and how they influence each other.
How is it you do that?
Dmitri Williams: Predictive analytics is exactly what it sounds like. The thing we do is figure out social value, but lot's first talk on a generic basis about predictive analytics. Let's consider an action in the system, when someone does A, B, C, D, A, B, C, D, A,B,... you can predict what's going to happen next. Our algorithm figures out what people tend to do, what the odds are that it will happen next, and it goes back and looks at how those things happen, such as what the circumstances are, what day of the week this happens, and so on. It might predict that D happens 80 percent of the time. That's the basic idea behind predictive analytics. If you exchange those letters with actions, such as quitting a game, spending, watching, or clicking, that starts to get interesting. We take that one step further, and build in some social network analysis. Social network analysis is the science of how and when people connect, and how that impacts their lives. Peoples who are connected influence each other. You and I might become connected, and I can see that if you buy a movie ticket, and a week later I buy a ticket to the same movie. Did I influence you? Maybe. But, did I influence you? Probably not. That's one of the basic things in statistics, one thing has to lead to another. We don't ultimately care if you and I talk about things, and we don't have to track your Twitter traffic, all we need to know is that there's a connection between us, that you did something, and that I also did something. That information can help lots of companies in a lot of industries.
How difficult its this to integrate into a game?
Dmitri Williams: The integration into our system is no harder than using any analytics solution. Like everyone else in this space, we give developers an SDK, and they can add code into their system when events happen, they can send us those events via a call to us in the Cloud. So, for example, we might given them an event called Login, and the record will say you have logged in 406 times using account number 1028. That event fires to us, and we input that into our algorithm. You just have to spend some time to instrument the game, and put in the call, that's all.
You also mentioned you might apply this to areas other than video games?
Dmitri Williams: It's not just gaming where people are interested in having it. E-commerce is the next logical play, and we've started our efforts there, though we're not ready to announce it. There are many other areas besides. It's anywhere people exist where people use products and influence each other, ranging from casino gaming to healthcare. It's really a questions of events happening enough so that data is available. That's why e-commerce fits, and music, videos, and other places where people have plans. However, first things first, and we have an industry we're focused on now and a robust product built out.
What's the biggest thing you've learned so far in the process of taking this from research to product?
Dmitri Williams: I've learned about human nature and what's valuable. We found that about 25 cents on a dollar is part of this process. We didn't know how big that was. As little as 10 percent of money spent, and as much as 40 percent, is attributable to social value. We knew it was a big deal, but didn't know the opportunity was this massive. It's really exciting to see something. We're gone from a bunch of guys you'd see in a Far Side cartoon in white lab coats, to creating something which is actionable, practical, and valuable. Researchers all hope that what they're working on matters, and transfers out of the lab in some way, so when it's really happening, it's pretty exciting.
Thanks!