Understanding the True Value of Semantic Discovery

It’s well known that Netflix invests $150 million per year in its semantic recommendation technology, which covers employing 800 in-house professionals --  including 300 content experts who manually tag each and every entertainment title.

That comes out to around $3 per subscriber per year, or a quarter per month. To put the $3 per subscriber cost into perspective, pay TV operators are spending about the same amount per subscriber for their most valuable asset, which is still their linear program guide.

Netflix values recommendations at half a billion dollars per year to the company, said Neil Hunt, Netflix’s chief product officer. That’s because most subscribers spend only 1 to 2 minutes searching for content, and if they don’t find something of interest they are likely go somewhere else, as they have so many other options to choose from. Netflix places a high value on their recommendations, and it’s clear that using semantic discovery for content recommendations is essential for the company to generate more business and to keep subscribers engaged by consuming more content.

What sets entertainment content apart from other products and services is that when it comes to TV shows and movies, everyone has unique tastes, and Netflix seems to understand this. For example, two people in the same demographic, and even within the same household, may want to watch different types of programming. Semantic recommendation engines understand the type of content subscribers are watching and are able to recommend content according to the individual’s tastes and mood.

Who hasn’t become frustrated when receiving a “people who watched this, watched that” recommendation, which has nothing to do with what you are interested in watching? Netflix understood that poor recommendations could be a “show stopper” for their subscribers and moved to a semantic-based approach. Only the semantic technology can know if a user likes entertainment about a “rough one-man army,” a “race against time," “criminal heroes” or a “fall in love” movie for date night.

Since each entertainment title is unique, it’s difficult for human taggers with different moods and subjective opinions to be consistent and to accurately tag the content. However, by using an automated tagging process with natural language understanding and processing, consistency and accuracy are maintained and the overall cost of the solution is lower by comparison.

For TV operators that offer live programming, both the linear and VOD consumption data can be made available to understand the users’ tastes. Personal recommendations need to be based on subscribers’ entertainment personalities, which are constantly changing according to their consumption data. Through a deep semantic understanding of the content and the individual subscriber’s tastes, operators can provide a more personalized experience, thereby increasing subscribers' level of satisfaction and overall consumption of content.

TV operators and OTT service providers are still missing a key component that Netflix has known about for a long time: It’s not only about having the best content but making sure suitable content is recommended to each subscriber based on his or her unique tastes.

We continue to hear that TV operators are losing more and more subscribers to Netflix and other OTT services. One of the reasons for this trend is that subscribers are not watching enough content they are interested in, even though there is an endless amount of suitable content readily available. While Netflix continues to boast about its increased subscriber base, pay TV operators and OTT service providers are seeing lower VOD consumption; maybe they are losing business to Netflix due to a lack of personalized semantic discovery.

With the introduction of new streaming services from HBO, CBS, Dish and many others, Netflix’s investment in semantic discovery continues to set the bar high for new players entering the streaming space as well as for the traditional pay TV operators. The continued churn to the pay TV operators and streaming services’ subscriber base who are not using semantic discovery technology is high, creating a gap between themselves and Netflix and other competitors who are using this technology.

Does this mean that operators and OTT providers should make a large investment to develop their own semantic recommendation engines? No, there is no need to try to develop this in-house as there are semantic discovery providers on the market that are as validated and mature as Netflix and can provide accurate recommendations at a much lower cost per subscriber than Netflix.

Yosi Glick is the co-founder and CEO of Jinni, a provider of semantic taste- and mood-based discovery solutions for pay TV operators and OTT service providers. For more on semantic discovery, view his TEDx talk, The Analog Man.