Your browser is not supported. Please update it.

19 December 2019

After riling discovery folk, Hulu tests limits of ad formats

TV advertising has come on leaps and bounds in terms of addressability and new formats, but no matter how personalized ads become, they will still fundamentally be perceived as nuisances by a large share of viewers, particularly the younger generation. Hulu is a company that doesn’t shy away from pushing the boundaries of OTT video – and now Hulu wants to rejuvenate the ad sector by – erm – showing less ads.

There is method in the madness, however. Hulu will reward binge viewers by showing less ads, a contentious move it plans to offset by allowing brands to display highly personalized offers targeted specifically at the most prolific of binge watchers.

The ambitious new ad format uses machine learning techniques to establish when a binge session is about to commence. This data is then served up to brand partners, presumably with a whole host of viewer data to personalize the experience, from where they can launch into real-time bidding. Every third episode of a binge session will spring up a message to inform the viewer they are a binge watching champion and will be rewarded dutifully for their loyal escapades by showing the next episode ad-free.

Hulu is a maverick of video streaming, particularly when it comes to ads. The company has raised eyebrows with its “pause ads” feature as well as the “stop suggesting” option which rolled out in the UI last year. Hulu’s mantra is all about giving users more control over the experience, although the latter feature could be interpreted as an insult to its own recommendation engine.

Last year, Hulu talked about treating itself to some upgraded dynamic ad insertion technology and indeed Hulu has lifted the skirt on its in-house technologies a little more than its major rivals have. Its in-house recommendation engine is built on technology from California-based content metadata company the Video Genome Project, which it owns, and according to subscriber feedback, the recommendations are regularly inaccurate.

Hulu’s recommendation engine has been built using a novel neural network-based collaborative filtering approach, called neural autoregressive distribution estimator for collaborative filtering (CF-NADE). This method is an application of deep learning in recommendation systems, using explicit feedback where users give each item an explicit rating. It says that, while in practice, explicit feedback is rare, implicit feedback like watch, browse, search and purchase behaviors are abundant. Hulu has previously said adapting CF-NADE to implicit feedback is a future focus, but the new Stop Suggesting button may have halted this.

In November 2016, Hulu acquired the Video Genome Project for an undisclosed fee, integrating its database of over 8 million categorized content titles into Hulu’s live streaming service. Unlike other recommendation engines which simply deliver suggestions based on factors such as genre, actor or director, the Video Genome Project says its own content metadata assigns “genes” to titles, driving down to a deeper level to enhance the video discovery experience – interacting with hundreds of elements like mood, theme, action level and scheduling. In reality, many recommendation products have gone down this route, so it’s far from unique.

There aren’t many moves like it. In a loosely similar situation in the rewarding sense, NBCUniversal introduced its WatchBack service in mid-2018 – offering things like gift vouchers to viewers on the free tier to encourage additional ad impressions. WatchBack was little more than a training exercise for NBC to learn about how rewards could be used to drive content discovery.

WatchBack was revamped accordingly in September 2019, adopting a model whereby watching 5 TV shows or 5 movies a month will reward viewers with a Fandango Promo Code towards one movie ticket with a value up to $12. It appears movie tickets are the now WatchBack’s only reward offering.

“As consumers’ viewing habits continue to shift, advertisers expect new and innovative ways to captivate these audience. Viewers desire advertising that is integrated and less disruptive to their storytelling experience. That’s why we’re committed to driving innovation that considers behavior and context when building new formats and experiences,” said Hulu’s VP and Head of Ad Platforms Jeremy Helfand.