How much content is too much content? Research shows that the average U.S. adult takes 7.4 minutes to make a selection on over-the-top services. According to further research, the average U.S. household has three streaming services and 70% of survey respondents agreed that they often struggle to figure out what to watch next. So, it’s not so surprising that 21% of all viewers decide to give up and not watch anything at all.
How can providers put an end to endless scrolling and provide accurate recommendations that put relevant content in front of viewers, so all they have to do is press play? It has to be a two-way street.
While OTT providers can make greater strides to using accurate metadata to give the algorithm and recommendation engine the best information to work with, viewers also need to reconsider how much they’re willing to share with providers in order to get the accurate recommendations they’re looking for.
There was a time when U.S. viewers had very limited TV channels to choose from: the “Big Three,” CBS, NBC and ABC, dominated broadcasting and decisions on what to watch came from the radio, newspapers or magazines. From this, viewers were given the ability to record programs, first on a VCR and subsequently a DVR, and then watch them at any time, creating far more choice.
Navigating Peak Content
This vast library of content that suddenly became available quickly presented another challenge: How could this new volume of choice be made easier for viewers to navigate? It was clearly time for operators to address this, but with so much content to present on either a TV screen or a mobile device, this was no easy task.
One way to solve this was through introducing algorithms that, when based on a set of rules, could determine which content to present to viewers. For example, if one viewer had watched a film with a certain actor, the algorithm could determine that they might also like other films or TV shows featuring that actor. However, this is not a sophisticated or accurate method of recommending content.
These algorithms have the ability to increase their accuracy by adding more rules, but this model still makes a lot of assumptions. In fact, in 2009 Netflix awarded a $1 million prize to a developer team for an algorithm that increased the accuracy of the company's recommendation engine by just 10%. Technology has moved along a lot in 11 years, but the challenge remains the same.
Furthermore, this only gets more complex when subscription credentials are shared between family or friends. One-third of U.S. subscribers admit sharing their streaming service subscriptions with one to two other people, and more than a quarter of subscribers with three to four people.
While a recommendation engine can form patterns between content watched at certain times of day — for example, cartoons between 4 p.m. and 6 p.m. — when the lines get blurred and cartoons are watched at different times of the day, and suddenly cooking shows are watched at 4 p.m., there is only a degree of intelligence that the engine can refer to.
These algorithms and rules have little potential to create the experience viewers crave without rich metadata behind them. The deeper and richer that metadata can be, the more chance the recommendation engine will have of finding a piece of content that the user will want to watch. However, to surface a sufficient recommendation, users also have to be willing to give a little something back.
It’s predicted that by 2023, there will be 257 million U.S. social media users. In this current age, consumers share everything about themselves on social media. But when it comes to sharing data or information with a TV provider, data capture suddenly becomes out of the question. A 2019 survey of U.S. consumers showed found respondents are less likely to share personal data than they were the year before.
Yet these are the same consumers that seek these personalized recommendations. If consumers want to be able to discover content more easily, they must be prepared to share personal information with their TV provider. It’s a two-way street; the provider needs to be able to use viewing history to suggest new programs or films line with a viewer’s tastes. But in order to do this, the operator needs to be able to convince the viewer that they will get a better experience as a result.
A Time-Consuming Chore
There is only so much content that can be presented on a TV or mobile screen, and if that content isn’t what the user is looking for, then searching behind the scenes for something more relevant can be a very long process.
When a program possibility is found and if the short trailer isn’t enough, the viewer could then commit to as much as an hour spent finding out if they like the program or film. That’s a big ask. By giving the user more of a snapshot of what they’re about to watch, or creating shorter form content like recently launched Quibi, users can spend less time scrolling and more time watching content to decide if it’s a good fit.
There are multiple ways to fix the content discovery conundrum, but one thing will make a big difference: confidence.
Confidence must come from the viewers that the operators are doing their best to give them accurate recommendations and they must trust they are using their data in the right way. Operators must have confidence in the quality of the metadata behind the algorithms to ensure that quality recommendations will be given as a result. λ
