Sramana Mitra: Let’s double click down into that field. I would like to hear three use cases of customers, where you use these kinds of predictive analytics to solve problems or achieve meaningful business goals.
Matti Aksela: We can start with one of the most traditional use cases in the mobile operator space – finding ways to predict and prevent. We had been doing solutions on churn prevention and churn prediction already. One of the things we added from an analytics view point is that we are trying to utilize all of the information operators have, take all that information and get valuable insights out of it, then use that to fuel predictive analytics and reach the best possible prediction accuracy.
SM: What do you learn from a social network from among subscribers, for example?
MA: There is a lot of information that is indirect. People tend to interact with similar people. On the predictive model side, we have also been able to see that taking features from your social network “neighbors” – the people you are communicating a lot with – helps us do better predictions. Another use case we have been doing is demographics prediction, especially when looking at prepaid dominant markets, where there is not that much information available on subscribers. This information has a lot of value. The age and gender of a subscriber, for example, is information you can obtain from people they have been communicating with. You are using the interactive algorithm over the social network to refine this information until it converges into better information.
SM: How much of this information are you able to dig into? If you have a subscriber, are you allowed to track whom they are communicating with on Facebook, for example?
MA: In our case, we are focusing more on the mobile operator data. How much we are allowed to track usually depends on the local legislation – we always abide the local legislation, of course. In some countries you are not allowed to use the social network, call detail records, etc. For some legislation it is quite a valid thing to use those resources to provide better services. It is not just about predicting for marketing activities. We also have use cases that focus on being able to understand what the problems are in a network. We thus also provide that kind of value to customers, which are the mobile operators, and help them serve their own customers better.
SM: How much revenue has this predictive analysis product generated for you?
MA: I am sorry, but I am not at liberty to discuss the product line revenues. Since we are a public company, there is a policy that I can’t discuss revenue numbers.