Four Key Factors in Developing Recommendation Technology

People are often disappointed in AI and personal assistant technology. Siri can’t read my mind yet, my lunch isn’t magically making itself, and my TV still doesn’t know what I want to watch. Ironically, sci-fi movies and shows are partly to blame for building up our expectations for a personalized future. Humans are complicated, so understanding what they like and predicting what they want is very challenging. Still, we’re close to realizing that dream—at least regarding entertainment. 

As the CEO of a company building recommendation and discovery technology, I can tell you that this is the fun (and challenging!) part of working on an entertainment recommendation engine. Everyone is unique, expectations are high, and there’s no shortage of opinions about any movie, show or song.

Even how we want to experience that discovery process is unique. Some users prefer to sit back and receive predictive suggestions from a powerful system, while others are disappointed because entertainment discovery experiences aren’t interactive enough.

Numerous pathways of discovery exist, each valid in different situations. This is why pay-TV providers offer on-demand programming and over-the-top (OTT) content provider Hulu is expanding its subscription service to let users stream live network and cable TV feeds.

What Users Really Want

Striving to please all people all the time is difficult. Ericsson recently found that half of consumers watching linear TV claim they can’t find something to watch at least once per day, so they default to something familiar.

Some personalization engines dangle low-hanging fruit for users to watch when all else fails, reinforcing this behavior. A bad feedback loop only intensifies as that data drives development and licensing decisions, filling catalogs with more of the same old content.

Existing movie and TV services will survive as long as users are satisfied enough to stick around. But as Ericsson’s survey shows, people want more than just a reason not to cancel. New or expanding services must differentiate discovery to attract subscribers looking for a way out.

To keep audiences engaged, next-generation TV experiences have to provide multiple pathways through social, predictive and interactive discovery. Otherwise, new services will pop up and snatch away users who can’t find content to enjoy.

No Unifying Theory of Recommendation Relativity

Believing a magical algorithm developed by a genius Ph.D. could deliver great results every time might be comforting. Unfortunately, the truth is much more complicated.

Four critical factors must be considered when developing a personalized media platform:

1. Catalog: Content matters — a fact we are reminded of every year after upfronts. New bundles, releases and licensing deals create so much buzz when they’re first announced — especially if live TV, exclusives, or original programming are involved. Catalog is just one factor, but it gets more attention because it’s expensive, contentious and most likely to be mentioned in marketing.

2. Device: Pay-TV has always been competitive on catalogs, but it’s mostly been tethered to only one device until recently. OTT services are winning on product today, but with pay-TV providers expanding the types of devices their content can be viewed on, they’ll soon catch up. Advances in hardware and data will make great personalized multichannel video programming distributor (MVPD) products possible.

3. Data Model: Entertainment is expansive and highly fragmented. We live in a world of “big data,” but how you model that data often matters more than how much you have. New graph approaches that model data as nodes and relationships unlock the potential of recommendations that account for social factors, as well as genre and mood.

Companies that leverage this data to provide the most efficient browsing experiences will succeed, especially if they can make connections between different types of content in their increasingly diverse catalogs. Some services will expand their offerings like Hulu, while others may home in on a niche—but all will benefit by going back to the whiteboard to draw a new underlying connected model to guide UX design.

4. User Experience: The crucial user experience (UX), through which the catalog is explored and taste profiles are cultivated, is often an overlooked facet of the TV conversation. UX is usually an afterthought when it comes to developing media platforms, which makes it the easiest factor to succeed with—if you give it the kind of forethought and focus it deserves.

Staying Competitive in the TV 3.0 Era

Subscriber retention and average revenue per user will always be relevant, but we also need to focus on new analytics to deliver a new TV experience. It’s not just time-of-day and device details that entice viewers—this is a problem for Freud as much as Einstein.

To move users from the living room couch to the therapist’s couch and find out what drives their decisions, the math of large data sets needs to be combined with the critical theory that explains why we care about and crave entertainment in the first place.

Preferred genres, themes and styles mirror users’ fears, hopes and obsessions, and these revelations help improve every aspect of personalized media platforms.

Entertainment recommendation tech should challenge users by presenting them with unfamiliar content. This approach requires highly connected data and a creative team with the vision to present intuitive but less obvious connections to users.

Personalization isn’t just exciting because it makes life easy. The promise of personalization extends beyond basic suggestions to include the type of deep understanding that expands your horizons and helps you find the unexpected. Recommendation tech that provides context and invites users to participate in the process provides a superior overall experience that builds rich taste profiles, fueling better results to create a virtuous cycle. Companies that are bold enough to accept this challenge—and committed enough to back it up with better data—will win the future of TV.

Addison McCaleb is the founder and CEO of MediaHound. Its platform, The Entertainment Graph, powers personalized discovery and recommendation experiences across all media. One of MediaHound’s apps, Date Night, recently won a Webby Award.