What Do You Recommend?

The old movie adage, "If you build it, they will come," has a twist in the competitive 21st century world of cable video. An operator could have the most advanced network and a massive content library, but if consumers can't find what they want to watch, they will feel there is nothing on and become dissatisfied, said Peter Docherty, founder and CTO of ThinkAnalytics.In other words, they may come, but they might not stay long, either.On the other hand, a robust recommendation engine, with personalized content discovery, can make the viewing experience more fruitful for subscribers."If you look at the linear TV schedule over the course of two weeks, there are probably 200,000 programs on," Docherty said. "Each of those programs are on multiple times. There are millions of instances of those programs being aired. Out of those that you could watch, which are the best 10 for you, and how do you know? (The engine) must understand the likes and dislikes of the customer."Viewers want to feel they are getting the most from their cable package, Docherty explained in a paper delivered at the Cable Show. While the average consumer has access to more than 400 channels, they tend to watch only eight. Broadening their horizons by truly targeting what will appeal to them will make them not only want stay with their service provider, but also may entice them into trying value-added products and services.True personalization is key, Docherty said. Current discovery uses web-like collaborative filtering or mood-based and content-to-content recommendations. These techniques use the concept of "people who watched this also watched that," for example, or suggest programming based on a genre, subject matter, or actor. However, this means that only the most popular content is pushed forward or that the user is still doing most of the work, Docherty said."(These methods) rely on the user to make sense of it all," he said. "Users (have) to lean forward and search and zap - find things themselves."Metadata on its own can only understand the content, and not the customer. A next-generation recommendation system needs to do both, and this requires better quality metadata, direct user feedback about likes and dislikes, and the ability to understand consumption data and usage patterns. "There is no one magic silver bullet," Docherty said. "It is really about having a ... platform that can understand everything."With the right mix of capabilities, it should be able to deliver content on the money to a 35-year-old male who likes romantic comedies, for example. "It is a case of understanding what you like and don't," Docherty said. "We can find that nugget of gold in the river."The other component is user experience and how the recommendations are presented. This can still be done in the guide for people who are comfortable with that method, but it also could involve other mechanisms, like a page that lists "my suggestions" or what to watch right now. "We are making navigation more intelligent, more of a lean-back experience, but are proactively presenting good content for them," Docherty said.Monta Monaco Hernon is a free-lance writer. She can be reached at [email protected].