It's Like Looking into the Future, Only Better
If you saw Minority Report, you’re probably thinking that reading the future is just science fiction. Think again. Katana® Analytics Engine will tell you the future, give you a measure of accuracy, and enable you to make decisions others can’t.
Katana focuses on the key predictives that drive your business: Churn, Conversion, and Time Spent.* You’ll see who’s going to do what in the future, and exactly how much is at stake. And you get to see it for everyone or on a player-by-player basis:
Automated Predictive Analytics Global Churn
Player Churn By Account with Expected Loss and Days Until
Here you get to know exactly what’s at stake so you can make smart cost-benefit decisions about what action to take. And when you fold in the importance of Social Value®, you see the whole picture. Not only is that player going to leave, but she’s taking $200 of friends’ spending next month with her unless you do something about it!
Katana layers predictive analytics with other power-user tools like segmentation and cohort analysis. So, you can not only see the future, but see how different actions and interventions impact that future. And, our easy Exporting tools let you select just the players you want to take action on. You can download them into Excel, or even connect directly to email marketing and push notification systems.
We’ve developed a series of machine-learning models finely tuned for games. Our game analytics engine takes in your data and runs its models daily, continually updating and improving itself. Machine-learning is exactly like it sounds--it takes data, sees the patterns, and gets better at it over time. Once it’s up to speed (typically in a couple of weeks), you get results with a high degree of accuracy. How accurate? We use a measure of confidence called an “F-Score.” For reference, churn models in the telecom market usually cap out at around an F-Score of .4. Katana typically delivers .7 or higher, with some games’ data hitting .9.
It is, but we have a team that’s specialized in it for nearly 10 years, and we’re really good at it. We love game data and predictive modeling, and we cut our teeth on it in projects for the CIA, Army and National Science Foundation. We’re the alpha geeks. But we also know you don’t want to hire a big team to tackle your game’s issues. It’s not practical. We know that having analytics that take a team of PhDs to figure out and execute is expensive and intimidating. But you know what’s not intimidating? Pressing a button and seeing clear, beautiful answers.
When you want to create a statistical model, you have to go through several arduous steps. Assuming you have a database and a team maintaining it, you assemble and clean the data, figure out what the right question is, pick the target variables, run many models and then, if you’re lucky, get a team of experts to interpret them. In contrast, Katana does all of this for you, every day.
Katana is the equivalent of an entire team of PhDs in a box. They don’t get tired, they don’t need to be paid, and they come to work every day and deliver. If you’re lucky enough to have some smart analytics cookies on staff, rejoice, because they are about to get a massive upgrade to their independent and dependent variables (don’t worry, they will understand), and they’ll deliver much more than they could have before.
The best metrics and models are based on sound scientific practices, and those are what make us tick. Our team is made up of university researchers who have been publishing on gamer data and algorithm development for years, and we always use the best and latest practices. Indeed, we invented most of them. We believe that data transparency is crucial. You always need to know how good the models are, and how much you should trust any individual score. If we tell you that a player is worth $5 and they’re going to quit next month, we always tell you the certainty. You get a model confidence number in a big font at the top of the page.*We’re working on more models every day, and will include them in Katana at no extra cost. Expect us to add predictive models for Virality (who’ll invite their friends), Spending and even Ad Clicks.