The gradual delinearization of content broadcast across different screens has led to a surge in popularity for online recommendations, also known as content curation. This trend can be directly linked to television channels and content broadcasters hoping to retain their respective audiences. Today, recommendation tools and systems help content broadcasters to better understand their viewers and build loyalty.
The exponential growth of online video content worldwide—both on-demand and by subscription—has led to a fragmentation of supply and an increased number of platforms. Legal access to films has also become more complex because of the rights purchased by platforms. This abundance of content has given rise to what psychologist Barry Schwartz calls the “paradox of choice”. Choosing an online film or television show has never been so hard. The popularity of the ”automatic play of the next episode” feature on Netflix is therefore unsurprising.
In order for a recommendation system to work, it requires a critical mass of content, which means it needs detailed information on both its movies and its audience. The ancestor and inspiration for many contemporary recommendation engines is eBay, whose tools have been copied and reused by many startups. eBay’s recommendation system was the first to provide a database on its users’ purchasing behaviours. Current independent recommendation tools (TasteKid, Jinni and Clerkdogs) integrate various recommendation technologies and algorithms. But Netflix’s success is partially due to its innovative filtering of the user data it has compiled over the years.
The two main models for recommendation systems
- Collaborative filtering uses notes and recommendations from users, suggesting content based on what users with similar profiles liked or bought. This solution requires a large amount of data, and platforms that have been around longer reap more benefits. Such manual curation relies on humans, as opposed to algorithmic curation, which uses only keywords. This explains the success of Songza and other music recommendation sites that rely on content curators.
- Content-based filtering (algorithmic curation) requires in-depth knowledge of catalogues, as each element of its content must be itemized in order to create a profile from its characteristics. Users’ tastes are gleaned from their behaviour, notes and any preferences recorded.
Many recommendation systems use social networks to uncover links between people and content. These platforms act like dating sites, matching viewers with TV shows, movies and documentaries. Dating apps already use Amazon and Netflix data, in addition to Facebook “Likes”. User profiles continue to expand and are becoming an essential part of personalized recommendations.
Platform competition means that there is little overlap in content. Users frequently return to the same platforms, which naturally accelerates the trend towards “filter bubbles”, where some content is simply not delivered and only a limited amount of content is widely shared and made popular.
Recommendation and public organizations
Public broadcasters and regulatory bodies alike fear the consequences of television based on profiles created by algorithms. In France in September 2014, CSA chair Olivier Schramek discussed this threat. Past preferences, purchasing habits and professional characteristics are filtered, resulting in a “bubble” that promotes “being closed off” and an “extension of private space”, instead of “opening a window to new realities”.
Institutions are organizing to promote and protect national content, and some algorithm regulation is emerging. In France, the Conseil d’État has acted in line with the CSA, tackling algorithm regulation in its September 2014 report on digital content. The report suggests “incentivizing French and European works” by “implementing a dedicated window in the recommendation results”. As for the content displayed, it plans to “require the actors involved to modify their algorithms, favoring the inclusion of criteria that encourage cultural diversity”. However, it is still too soon to know how platforms and content editors will react to these demands.
In the United States, with a primary goal of preventing piracy, the MPAA launched its Where to Watch platform in November, a search engine that tells internet users where they can find content online. This cooperative service is one of the most comprehensive and provides information on where films can be found (to rent or buy on DVD or Blu-ray). Where to Watch combs through the catalogues of Netflix, Hulu, iTunes, Amazon, Xbox Video and more than a dozen other service providers to determine which will allow a user to watch specific content. However, this recommendation tool works only in the United States and not Canada. The delayed arrival of this platform is raising doubts about its adoption, as internet users have developed their own habits in recent years.
The increase in online recommendations raises an underlying question that all online broadcasters must ask themselves: Should we give preference to national content? What type of dialogue should be initiated with users? The current trend towards “hyper choice” hides a lack of information on cultural products and makes it more difficult to discover shows that you may enjoy. In fact, access to content and technological support are not enough to meet the challenge of discoverability: appropriate promotion of online consumption has therefore become necessary.