Making recommendations based on what a user likes is only part of the solution to the discovery problem. Popularity measures that reach a national scale are essential too.
Recommendations have limitations
Recommendations are a great tool to help get viewers to the content they want as quickly as possible. However, consumers still rely far more on the tried and trusted the guide. In a recent survey of UK, German, and Swedish viewers, nScreenMedia found that 40% or more of online viewers rely on the service guide while 15-20% rely on service recommendations. Though TV device manufacturers are investing in their cross-service recommendation features, fewer than 5% of viewers rely on the function.
What keeps people coming back to the guide? It could be they are afraid of missing what’s hot. Most recommendation solutions leverage consumer viewing habits to figure out what other shows or movies they might want to watch. Unfortunately, that means customers might not see the latest hot show or sleeper hit movie because it doesn’t fit their usual viewing style. Gracenote wants to change that with its new Video Popularity Score (VPS.)
Bringing popularity to recommendations
The VPS technology is part of the company’s Advanced Discovery suite of metadata products aimed at helping pay TV operators, online TV providers, and CE devices improve their interface experiences. Earlier this year, Gracenote enhanced its metadata with more extensive video descriptors. It is now using that technology as part of the VPS solution.
VPS draws on several data sources to allocate a popularity score between 0 and 1 for movies and TV shows. The score is calculated using a proprietary algorithm. The data sources include:
- TV, VOD, and online TV service consumption data from Nielsen Total Content Ratings
- Movie box office data from Gracenote
- Nielsen Social Content ratings
- Other trusted sources.
Answering the basic personalization question
According to Kamran Lotfi, VP of Product Management at Gracenote, VPS providers an answer to the fundamental question every video service interface designer has:
“When you get a list of 50 items, how do I put them into an ordered list to give back to the user?”
Having a single score representing the popularity of content means it is trivial to list the titles from most to least popular.
For the burgeoning online free ad-supported TV (FAST) services, providing a customized viewing experience is difficult because viewers often don’t need to login to watch. VPS can be helpful here because an unknown viewer can be shown the most popular content first.
Newest is not always the most popular
The inclusion of social content ratings in the algorithm for calculating VPS helps ensure the titles of the shows and movies driving the national zeitgeist receive an appropriately high score. Mr. Lotfi says the Video Popularity Score always reflects the most up-to-date data because it is recalculated daily. Services using it are also updated daily with the new scoring dataset.
However, VPS doesn’t always rank the newest content the highest. Mr. Lotfi gave the following example to illustrate the point:
“There are two versions of the movie Overboard. There’s the one with Goldie Hawn and Kurt Russell and the newer one with Anna Ferris. Based on our popularity score, the newer one actually has a lower score than the older one.”
Differentiating between what is popular and what is new is essential. For example, when a viewer asks a voice system to “Play the movie Overboard,” starting the most popular version is probably the better bet.
Tools like VPS can be beneficial to service providers and devices manufacturers by improving the usefulness of the recommendations they provide. Moreover, better recommendations mean a much-improved experience, and experience is a critical factor in keeping a user happy.
Why it matters
Content recommendations are often based on the viewing behavior of the individual viewer.
Recommendations must also consider the broader popularity of content to be effective.
Without popularity, search and discovery tools such as voice search, can’t do an adequate job.