Almost everyone in the civilized world knows that content consumption – more concretely, watching TV – has shifted from the “live” lean-back experience that was really successful for the broadcast industry during the last century to an “on-demand” experience: people now expect to be able to watch what they want, when they want, and where they want.
Traditional broadcasting is still relevant when it comes to live events like sports matches, concerts, news programs, and prime time TV Shows. Certain content needs to be watched live in order to reap advertising benefits; such programming includes the Super Bowl or high-profile programs that will be aired only in exclusivity windows to maximize viewer numbers.
The linear broadcasting business has certain clear rules that have been part of the industry since almost as far back as TV’s beginning in the 1930s. In this paper we’re not going to focus on this side of the business. There are already plenty of books on this subject and professionals who make their living from specialized expertise in this area. We’ll leave this topic to them.
Instead, we’ll focus on an analysis of Video on Demand (VOD). Although this model has been around for almost 10 years, it is still in the early stages of its development and there’s lots of room for improvement. Specifically, we will look at solutions for content recommendation and enabling direct content discovery as methods to enhance the customer experience.
The “on demand” lean back experience
It is curious that in the early days of VOD all service providers including Netflix, the pioneer of “on demand” TV, aspired to emulate the broadcasting industry’s traditional “lean back” experience. At first they offered their customers content based on past viewing history, what was top-listed, or new content.
They also offered search engines, usually filterable by genre, actor/actress, director, or title of the movie or series.
The recommendations process has evolved so that as a consumer you now have your own profile. Smart recommendation engines learn from this profile, how the profile acts, and what it does. Content related to what you normally watch is then recommended.
Let´s take a look at how Netflix both recommends content and enables the discovery of content on their service.
Netflix recommendations:
First of all, we don´t normally change profiles when watching programs with our family. As a result, the service is equally likely to recommend Narcos as Pokémon, as shown in the example above. This happens because most of the time you watch adult series and kids cartoons under the same profile.
Now let´s look at an example of someone using the same service’s search functionality.
Netflix search engine
In this scenario you don’t want to use the recommendations and would rather search the content for yourself – I encourage all of you to put the search engine to the test! In this example, I tried looking for pirate movies, so I typed “pirates” into the search engine.
The results varied. They returned everything from vampire movies to comedies, from Asterix to Matrix and The Fast and the Furious; only the first result was related to pirates. I find it hard to understand the basis for some of these recommendations, other than that they are big, blockbuster Hollywood movies.
We know that the most advanced VOD service in the world invests millions in recommendation and search engine tools, yet with just a couple of quick real-life examples we find that there is still a lot of room for improvement in both of these areas.
It is worthwhile to take a deeper look at the TV industry. In doing so we find significant opportunity for innovation, because we have observed that many VOD services merely copy Netflix. Specifically, the industry needs new approaches to the problem of finding the right movie in a library with more than 20,000 titles.
The “New Paradigm”
Even with all the major changes the television industry has undergone in the past five years, there is one simple fact that has changed the rules of the game and introduced a new business paradigm: in online TV every “click of the button” is registered and can therefore be tracked.
It´s needless to say that this fact is providing enormous value to companies that are using the huge amount of available data (Big Data) in an effective way and transforming Big Data into useful data (Smart Data). Companies like Amazon, Google, Netflix, and Spotify have a long history of using data-driven approaches to yield spectacular results.
The latest trend in the television industry is how to use machine learning and algorithms to make the business grow and help to create a differentiator in an already crowded environment: the TV market.
Let’s return to the subject of this paper – content discovery and recommendations –
to see if we can extract some applicable conclusions from our analysis.
As we described previously, there are two ways that people usually interact with online TV services and VOD stores:
Recommendations from the service:
Consumers turn on the TV application and wait for recommendations that the VOD service provides based on the customer profile and preceding usage.
Consumer Discovery:
Consumers opt for navigating a bit, looking for hidden content that is oftentimes great quality, but not always very easy to discover (similar to the “zapping” or “channel surfing” experience)
There are many vendors in the TV industry that focus on providing solutions to the first approach; companies like Jinni, Think Analytics, Contentwise, Digitalsmiths.
On the contrary, with regard to the area of “Smart Discovery” we have not identified many solutions on the market that help consumers find the content they want in the fastest way possible.
The Search Engines Example
In order to better explain the problem and the various alternative approaches to address it, we’d like to draw a comparison to the explosion of Internet search engines back in 1998-2000.
In their infancy, Internet search engines were pages full of images and recommendations: what to search for, what to find, where you had been navigating previously, and so on.
All types of videos, pictures, and news were shown in these initial search engine pages. Search results consisted of a mixed offering – video web pages, news, and YouTube-type related content – delivered up by the Terras, Altavistas, and Yahoos of the world.
They were worried about how to recommend navigation topics to keep people on their pages longer, not to mention that in many cases advertisements were displayed all over the page confusing the issue even more.
Ten minutes by car from Yahoo’s headquarters in Sunnyvale, there was this company called Google that changed the paradigm: they provided a simple blank space where internet users just typed in what they wanted to search for; it was as simple as that.
Of course, now we all know about the complexity that lies behind Google’s search engine and all the underpinning technology required to make it happen, but it sure started as a simple idea: to make searching easy and fast.
JUMP: Smart Discovery
In the case of content discovery, Jump wants to take a different approach to the ones already in use by top-VOD stores today. Jump wants to focus on “Smart Discovery”, which entails learning from people’s behaviors. When they search for content, the offering should be adapted to their specific mood or situation; the time of the year or the day of the week; if they are alone, or with family or friends.
If you think about it, the way each of us searches for series, movies, sports events, and other TV programs (previously known as “channel surfing” or “zapping”), is directly related to several variables that are hard to evaluate beforehand – unless we could read minds (not yet, but coming soon!).
Examples of such variables include: who you are with – are you alone, with a companion or spouse, with friends or family; your mood; the weather outside; if you are on holidays or in the middle of the workweek; if it’s Christmas or Thanksgiving, and your kids are off-school and at home etc.
These aspects and many others influence your feelings about what you want to watch. We think it’s possible to adapt content discovery to the individual so that the search results are tailored person to person and won’t necessarily be the same for everyone using search functionality to find something to watch.
So, even though we can’t read your mind or guess your mood, if we have information about your past consumption habits, we can make quite accurate predictions. The more you use the service, the more we´ll learn, and the better the results of content searches will become.
When someone uses a certain service, it leaves patterns that can easily be tracked. By applying machine learning algorithms we can make the VOD service offer you more accurate content when you’re looking for something on TV. Looking for something to watch becomes much more than a search: it becomes a discovery.
In subsequent articles, we will provide more information about Jump´s approach to Discovery.
