by Samantha Bookman
Five years from now, a viewer will turn on his or her television and see a host of content being offered that is tailored specifically to that individual. Not just a list of cable programs that are on, but a host of selections from TV programming to OTT content, to lifestyle and shopping recommendations -- all changing to suit the time of day or even the viewer's mood, so to speak. At least, that's what content providers want to happen.
If it seems as if the latest buzzword in the OTT dictionary is personalization, that's because this aspect of content discovery is fast becoming an important component of providers' search and recommendation platforms. The term itself has been around for a few years. But as consumer demand for content continues to grow, and as both traditional operators and new OTT entrants look to stand out in the IP video space, personalization of content is now a driving concern.
The quest for true personalization
"Everyone is trying to figure out the holy grail of personalized media content," Todd Viegut, CEO of Kannuu, told FierceOnlineVideo as he described the scenario above. Kannuu provides content discovery software and related services to pay-TV and other multiscreen providers including Telstra. Personalization is a key component among its services -- in addition to search and recommendation, Kannuu puts a lot of attention into the user experience (UX) as well as metadata -- all important facets of content listings that viewers don't, or at least shouldn't notice when they're looking for a program to watch.
What is personalization, anyway? In rough terms, it goes a step beyond search and recommendation as we know it. Think back to Netflix's introduction of "Max" in 2013. The discovery service, created by game developer Jellyvision, allowed subscribers using Netflix on the PlayStation 3 to answer some random questions such as their mood, the genre of movie they wanted and a few other components. Max would then spit out a random movie recommendation. The problem? "Max sucked at his job. If you wanted something light and breezy, he'd probably end up tossing you Hotel Rwanda," Film School Rejects' Adam Bellotto wrote in a recent ode to the service, which was eventually phased off the PlayStation and never saw wider release.
Max was just one attempt at providing a deeper, more personalized experience that would help a viewer who wants to find content that appeals to his or her taste or current mood. But it had some key flaws. One was the user constantly needing to directly input information for each discovery request; the other was that Max wasn't returning desirable content recommendations. "If it took Max a year to convince a couple of bored college kids to try even one bat cartoon, then yeah, he probably deserved to die," Bellotto wrote.
Search and recommendation providers now are continuing to look for ways to accomplish what Jellyvision intended with Max: pinpoint what consumers want to watch. But the trend today is less direct input from users, instead developing more intuitive analysis based on what users are telling providers through their behavior.
Many multiscreen providers and measurement firms like Nielsen are turning to social media to gauge viewer behavior and build more detailed definitions of each demographic. Sites like Facebook and Twitter are publicly accessible and can be quickly searched for activity occurring around specific programming such as sporting events or TV series. Programmers and advertisers can glean viewer information from YouTube as well, such as total views of a specific video, likes and comments. But combining social media with traditional ratings and YouTube likes still doesn't present a complete picture of the viewer.
Very few video distributors and providers in the industry would disagree that developing content discovery is important. "I think it will be paramount," Viegut said.
Frost & Sullivan analyst Avni Rambhia, in a presentation at Streaming Media East this spring, called content personalization and discovery the most transformative piece of the industry over the next 24 to 36 months.
A brief history of content personalization
Personalization as a distinct and key product "has been kicking around 10 years or so," said Alan Hoff, VP of strategic marketing for SeaChange, a Boston-based company that provides multiscreen and on-demand platforms for operators like Liberty Global and Rogers Communications. "We saw it first with what DigitalSmiths was doing in terms of a business. ThinkAnalytics really got momentum when they were working with Liberty Global and others. So you can think of them as one and two, in that respect. And then Jinni came along [too]."
Early on, video search and recommendation used predictive behavioral analytics, similar to what Web users were seeing on Amazon and Google, where product recommendations were made based on other products they had searched for or looked at. "It's two-dimensional, but it's what we had," Hoff said. "More the demographic data set and less psychographic."
DigitalSmiths was acquired by TiVo in 2014, which then incorporated its cross-platform, SaaS-based content search offering into its set-top boxes. ThinkAnalytics has been constantly developing its recommendation engine for more than a decade and has a number of multiplatform providers in its portfolio including Fox, Sony-owned Crackle, and ViaPlay. Jinni's engine can be found with OTT and traditional pay-TV operators worldwide including Vudu, Comcast's Xfinity, AT&T U-verse, Telus and others.
These companies charted some pioneering developments in search and recommendation -- TiVo's STBs, for example, were the first on the U.S. market to enable customers to see both their cable operator's program guide and access leading OTT services like Netflix and Hulu, with DigitalSmiths technology as a key component. But new entrants are continuing to pop up in the segment, particularly around better personalization of content across platforms.
Hoff continued: "You can now make determinations based on time of day, geographic location, the device on which content is being accessed. All this is very valuable data for marketers and content producers to see how their content is being viewed."
Improving on what Digitalsmiths, ThinkAnalytics and Jinni brought to market is a goal for many companies that are innovating on content personalization. And the market is seeing a number of new entrants who feel they can do it better, faster and leaner -- and directly monetize personalization as well.
Planting the personalization money tree
The founders of Rabt had a pretty good app on their hands. Tying into YouTube's API, it helped viewers find the kind of content they like on the OTT service. Bobby Alexis, VP of business development for Rabt, said the app quickly had tens of thousands of users, all getting a personalized YouTube recommendation experience based on their profiles on the app.
The problem? They couldn't make money from the Rabt app as it was. Knowing this early on, it became "a test ground of what we could do on a larger front," Alexis said.
"That helped us figure out how successful we were at personalization. So we took this approach, [we realized] we can't make money on the app now. So let's focus 100 percent of our effort on a B2B product," Alexis said. In January the six-member team built Rabt Pro, which specifically serves large businesses that have an online video presence.
Monetizing content personalization has so far been mostly indirect. Online video providers like Amazon and Netflix wrap it into their cost of doing business, as a component of their VOD service.
When it comes to directly making money from personalized content placement, advertising is the top player. Programmatic ads are rapidly gaining market share in the OTT environment, thanks to the development of technologies like dynamic ad insertion (DAI).
"Obviously, core advertising has the biggest moneymaking potential," Kannuu's Viegut said. However, personalization is important for SVOD providers as well. "Bundles and a la carte, we're gonna see a lot more of that coming forward. And probably some unique pricing and monetization strategies that will come forward," he added, as vendors like Rabt offer new platforms that more deeply personalize content for VOD users.
Rabt, for example, said it will base its pay structure on its product's performance for online video provider clients.
The company pulls together data for its personalization algorithm using a number of methods such as collaborative filtering and multidimensional statistics that correlate data from the users to the content they watch "and everything in between."
The multidimensional system also connects one user's likes to another user's, comparing what each person likes or doesn't like, and saves those results indefinitely so that when a viewer accesses a Rabt-enabled OTT service, his or her recommendations reappear -- not just as they left them but updated as other users' preferences change.
"Their actions are influencing your [preferences], so when you come back you get personalized recommendations immediately," Alexis said.
By repackaging its personalization algorithm as a business service, Rabt says it has found a model that brings in revenue -- as long as they're delivering results.
"Our monetization is on a CPC [cost per click] basis," said Alexis. "Depending on the CPMs [cost per thousand] out there, we are taking a cut on the increase or improvement on video plays that we provide."
Rabt tested its Pro platform earlier this year with LETV, one of the largest video providers in China that has more than 50 million viewers. The company's software was put up against the search and recommendation efforts of LETV's internal team of data scientists. "We outperformed them between 5 and 22 percent better in terms of click-through rates," Alexis said.
Rabt feels its service rates are competitive, and it offers companies a trial period before committing to the solution. "Personalization and content discovery is a very difficult thing to do," Alexis said. "That's why we offer [our service] at a rate that works well for content publishers. If you want to do it right, it's very tough to do internally."
Looking beyond social media engagement
Like Rabt, TruOptik's founders are building their company based on past search and recommendation experience. While personalization is just a part of what they offer to multiscreen providers, it's an important component.
TruOptik taps into publicly accessible peer-to-peer file sharing sites to develop insight into online users' behavior and preferences. "The data we're getting, the agnostic access to content on P2P, is a very pure read of what consumers like," said Andre Swanston, CEO of TruOptik. "There's no distributor limitation like with HBO. There are no walls around P2P, and [having] no walls can tell you something about content consumers."
The data gleaned from what file sharers are downloading can, for example, help brand advertisers "monetize their ad inventory by adding an external layer of data they're missing."
Even though P2P file sharers, such as those using BitTorrent, are seen by some as illegal users of content, Swanston pointed out that file sharers "spend more money on content and subscriptions than the average consumer." Learning their behavior patterns is key to figuring out how to target each demographic with both content and dynamic advertising. "It's really an opportunity to grab the younger millennial who is spending money on content and brands," Swanston said.
Privacy versus personalization
"The ultimate goal is to try to understand -- and this has been the goal of advertising from the beginning -- each customer as close as possible as an individual," said SeaChange's Hoff. "And as we get more sophisticated with our tracking and more loose with our privatization standards, we're getting closer to the goal."
Learning how households are engaging with content is a big part of the personalization game. An adult in the household may be watching pay-TV on the living room screen, Hoff said, while a teenager may be in another room streaming video on their tablet. "Netflix is all over that. It's the trackability of all this data. It's the use of natural language processing algorithms. It's the ability to subset data so it's not so big." Processing all that data is also available at much lower prices due to the size and number of data centers worldwide, he added.
How much does that increasing ability to track consumers infringe upon their privacy? While none of the companies addressed this issue directly, Hoff noted that privacy issues that would have baby boomers raging barely register with millennials. "And among the so-called generation Z set aged 7-19, it's even less of a concern. They understand that so long as they're getting access to this content they want to watch, it's a small thing."
If true, that's a huge change from less than a decade ago, when Netflix said it would not hold a repeat of its Netflix Prize competition (which offered a $1 million award to any individual or company that could create the best collaborative filtering algorithm to predict user ratings for movies), due to Federal Trade Commission concerns about privacy and a related lawsuit (later dismissed).
Ultimately, the consumer's privacy may be a tradeoff for a deeply personalized experience with the digital media they want to watch or listen to -- as well as a monetizable service for media and entertainment companies. "The more you know about the target demographic, the more valuable the content is," said TruOptik's Swanston.
"Once the TV Everywhere approach is done and all content is available across multiple devices, it's gonna be the user experience [that matters]," said Rabt's Alexis. And both online video and traditional providers need to up their content discovery game. "Within next year or two, if these OTT players and even cable, if they don't start personalizing their content they're going to start losing out," he added.