Decker, along with Jordan Brown, Director, Content Marketing, UFC, presented a session at the show, “Leveraging AI to Power Subscriber Acquisition.” By leveraging machine learning and AI technology, content rights holders and brands now have the ability to not only zoom into individual behaviors of their direct-to-consumer audiences, but predict who is the most likely to stream, what they are most likely to stream, or who is the most or least likely to churn. The session further looked into how Endeavor Streaming has partnered with the UFC to use AI in order create more efficient and effective marketing audiences.
Endeavor Streaming is the technology provider that powers some of the largest streaming services across sports and entertainment. On the audience development side, you use a variety of different ways across growth marketing services, analytic services and predictive modeling.
“The streaming industry is chaotic at this point, with all the different changes in the landscape, especially in the sports industry. That’s where we’re heavily involved,” says Decker. “All of these different sports leagues, teams, federations, are making streaming a foundational part of their media strategy. It’s a lot easier said than done and they are doing it in a variety of different ways. Some of them are using companion streaming apps or membership models. And then some are taking their premier content pieces and putting them on their own streaming services or licensing them to other streaming services. The sports market is insanely saturated and driving up CPA’s, increasing churn.”
Can you tell us about the data science initiatives that you undertook in working with the UFC?
Decker explains, “Endeavor Streaming and UFC have been partners on their OTT platform for five or so years. We have an abundance of data that we can look at and evaluate—who are our best subscribers and who are our churners? That seems all relatively straightforward, diving into advanced analytics if you have a substantial amount of data. But we got in a room with our data scientists, and we were just kind of evaluating all the different metrics that we had. Our churners all have these characteristics in common, and the people that view content have these characteristics in common. It’s great to know who your churners, streamers and best customers are, but it would be a lot better if we could predict who’s going to churn. That can inform our marketing and messaging as we start to think about growth and retention. That’s how we can use AI or machine learning. What are some of the new advanced things that we can do in terms of predictions and building these predictive models?”
That collaboration between the business and data science is abundant in coverage here at the Media Insights Conference. We all seem to be talking about it. There are varying levels of success. But what about the human dynamics?
“Essentially, we were looking for ways to collaborate with these data scientists,” he says. “Their solution was to have a more forward-looking approach, more foresight into all this churn that is being experienced within the industry. Let’s try and figure out a way that we can build some predictions—who’s going to churn and who’s not going to churn. It seems straightforward, but there’s a ton of complex dynamics at play when you’re trying to build predictive models. That partnership has been the foundation of what we’ve built on these predictive models. Here’s all the people that we’re predicting are going to churn or who are not going to churn. But then, how do we actually apply this to our main goal, which is to scale the business. The application side is where the model outputs turn and get transitioned to a different set of hands. What action points have been taken?”
Decker continues, “If we take propensity to churn for UFC Fight Pass, we are now getting outputs of all the users that we can make predictions for. Those outputs are classified into different segments—people that are highly likely to churn, medium likely to churn, and unlikely to churn. That should inform the way that we’re communicating to these people. We’ve set up connections from our back end where we receive these models and segment them where we can push them directly to UFC, fight passes, third party forms. We’re building different marketing campaigns. We’ve built out segments that are focused on targeting lookalikes from the best customers who are highly likely to stream and low likely to churn. When we tested that in six different markets, the best performing audience was built off the machine learning data. We saw anywhere from 20% to 40% CPA reduction. The cumulative CPA decreased by 19% during that time. It’s obvious that informed audience creation based off these models is extremely beneficial.”
“Ultimately, we’re just starting to scratch the surface,” says Decker. “AI is a very scary term to a lot of people. But it’s really accessible to people in our industry. Based off a lot of conversations that we’ve had so far here today, people are really starting to get behind all the different power and the different tools that data science and advanced analytics can bring in terms of machine learning.”
Watch the complete video from the Media Insights & Engagement Conference as Seth Adler and Beau Decker discuss reengaging churn customers, targeting the “middle” consumer group, sports world partnerships and more.
