The best way to get and keep viewers on a video service is make sure they can quickly find something great to watch. That is easy to say, but very hard to do. IBM thinks Watson can help with the problem.
Pay TV needs help with discovery
Pay TV subscribers continue to struggle with finding something good to watch on TV. One third are overwhelmed by the number of channels they have access to. Two thirds of pay TV subscribers sometimes or always get frustrated trying to find something to watch. And just 22% report that their operator makes recommendations of things to watch to them.
Not surprisingly, this is a key factor undermining pay TV satisfaction ratings. The average pay TV company has a satisfaction rating of 65, with Comcast at 62 and Cox 59. Netflix has a rating of 76, and Amazon 83.
The mountain of video service data
To make good recommendations, operators need a lot of preference and video data. They are now sitting on a mountain of it. In addition to the set-top box interaction data operators have traditionally collected, there is now TV Everywhere client usage data. This new online data provides a level of detail about viewer behavior never available before. It also presents an analysis challenge the likes of which operators have never seen. Speaking about this challenge at INTX in 2015, Tim Connolly, Hulu’s SVP of Distribution, had this to say:*
“We collect 4 petabytes of user data a day on what consumers are doing on our service. 1 billion events a day…There’s a zillion ways we can use data. We are 7 years into it and we’re just scratching the surface.”
According to David Mowrey, Vice President of Strategic Planning & Business Development at IBM Cloud Video, that is precisely the challenge:
“How do you get actionable insights out of all the data? It’s easy to provide analytics, but providing insights is much harder.”
IBM points Watson at the problem
Watson, which has been integrated into IBM’s Video Cloud, is being used by the solution in three key ways:
- Analysis of social media to understand real-time audience reaction to live and broadcast events and television shows
- Automatically identifying scenes and themes in a movie or show, allowing more accurate video indexing
- To identify audience preferences by analyzing masses of data from usage, content information, and social media
Scene identification and audience reaction functionality provide a wealth of content descriptive data that is a key part of an overall content intelligence solution. However, it is the analysis of this and other content data along with preference information that has the most immediate value to operators.
Watson and recommendations
Watson has proven itself a powerful tool for turning huge amounts of data into actionable insights. Mr. Mowrey says IBM is trialing the use of several key Watson capabilities to deliver better recommendations. These include:
- Speech to Text
- Natural language processing (AlchemyLanguage),
- Detection of emotions, social tendencies and writing style (Tone Analyzer)
- Analysis to uncover personality traits (Personality Insights)
Combining this functionality into IBM’s Media Insights Platform should allow it to make very accurate recommendations. He says these capabilities will be released for use by clients later this year.
More than just better recommendations
Unlocking the information in the mountain of data available to today’s video businesses will not only improve content recommendations. IBM is betting that applying Watson’s cognitive capabilities could touch every aspect of a video service. As Mr. Mowrey says:
“How do you increase customer acquisition, how do you decrease churn, how do you make better content acquisition decisions, how do you make better licensing window decisions, how do you make better advertising decisions. Those decisions are worth billions of dollars.”
And in a climate of cord-cutting and reduced profit margins, Watson and IBM Media Insights Platform may be just the help pay TV operators need.
Why it matters
It is still very difficult for pay TV subscribers to find something good to watch through their service.
Though Pay TV operators have plenty of user data available, it is very hard to leverage it to create good recommendations.
Operators need help turning this data in actionable information.
IBM thinks Watson can help with this job.