The Filter's mission? To be a discovery and filtering service for all entertainment. Think Amazon's recommendations grafted onto Facebook; finally, a union of the granularity of a filter to the holy grail of marketing: word of mouth. If The Filter succeeds we'll read more books, listen to more music, watch more videos - or at least do so with bull's-eye accuracy.
Since the birth of Amazon, collaborative filtering (CF) has been a driver of e-commerce on the Web, an online analog to hand-selling. When Amazon added customer reviews it created communities around topics, and built a marketplace around the customer - not advertisers or marketers - giving them access to the value of products. Content itself became a vehicle for conversation, enhancing participation on the portal, and ultimately, loyalty. CF 1.0 has not been without its missteps, Wal-Mart had a forest fire to put out a few years ago when consumers purchasing Martin Luther King's: I Have A Dream were shown the Planet of The Apes DVD boxed set. But the science has evolved. In powering a new generation of news aggregators, (Reddit, Digg, NewsVine) CF has revealed stories authored by the most esoteric sources - buoyed up by timeliness and appeal. And now, the focus has expanded from news to media content.
Charter members of Netflixreveled in its Long Tail inventory, and in time, its database leveraged CF based on consumption and recommendation to offer a discovery engine for movies. Likewise Pandora's Music Genome Project - both hand-built by musician/linguists and employing algorithms - exposed listeners to music seemingly composed according to their harmonic preferences. In this manner, CF saved us from Barry Schwartz's "Paradox of Choice" - the overkill of choice that threatens to immobilize us all. Peter Gabriel couldn't agree more, viewing The Filter's mission as "Freedom From Choice."
While The Filter is in its infancy, it is enough of an advance to drive the discussion around discovery to the next level. The Filter aspires to provide recommendations across all forms of entertainment by mashing up CF and social networking. For now it is focusing on music, movies, online video, and, in short shrift, TV. It aspires to move into other areas as long as usage can be tracked electronically, so rather than the printed book think eBooks, podcasts, audiobooks, even RSS feeds. It is one of the first efforts, then, to recommend offerings across music, film, and video, looking at content not as media types, but as what Malcolm Gladwell refers to as "taste products." Consider the expanded marketing possibilities: favorites Goddard's Breathless, Aimee Mann's Save Me, yields a… Vespa. But let's not get ahead of ourselves.
The Filter aims to leverage discovery not merely by its CF technology, but also through the Website's usability, and by adding social networking -- expert word-of-mouth that can be filtered through friends, family, even celebrity. Word of mouse?
A criticism of CF is the vast amount of data required. The Filter solves this problem by permitting users to import data from partners. Currently, one can download a simple plug-in for iTunes and import from both Flixster and Last.fm. "We are very keen," offers Roberts, "to integrate other music-only, movie-only or TV-only services that make The Filter even more useful for people (accessing one's Netflix queue would be a good example)." At present, libraries of five million songs and just over 300,000 movies are available.
A more difficult challenge than it seems. Too many dials to toggle and a user will quickly parachute out. While our culture seems enamored of a binary "thumbs up, thumbs down" when it comes to movies and music reviewers, what's more common is a five-star system (with half-stars allowed). After initially having a slider that ranged from 1-100, The Filter has adopted the ten-point scale. Down the line, this specificity problem will have to be solved.
All collaborative filters are not created equally. The Filter, explains Roberts, "crunches evidence in the form of purchase data, consumption data and browsing data, to work out the strength of connections between content items." Bayesian results are a jumping off point; a number of filters are added to fine-tune the results. These filters may be based on metadata, (recency, genre, BPM) and can be either user- or environment-based. The Filter aspires to add the capacity to filter results through external inputs: friends, experts and celebrity tastes. Roberts adds that the technology is battle-tested. The engine has been used by over thirty-five online stores in more than twenty countries in Europe.
A Brave New World for Advertising?
With such robust technology, it's easy to see how agencies and brands might find The Filter compelling. The Filter shows promise in cutting through the noise, enabling messages to be matched according to one's tastes. More specifically, Roberts points out, "The challenge for marketers is to be able to target campaigns and promotions at a much more micro level. I believe that tools and services will emerge that will help in that process." While The Filter has plans to filter advertisements to members in upcoming versions, it will provide them with the ability to opt out of any personalization. While Roberts reports that brands and agencies have approached him to promote products and services, The Filter's first order of business is to improve the site while scaling its user base. Late summer and beyond we'll see the first fruits of its advertising and promotions relationships.
An Amuse Bouche Or A Meal?
Unlike Pandora, The Filter does not offer listeners complete tracks, only a thirty-second whiff of a song. Roberts grapples with the matter: "Pandora is revolutionizing radio - so they have to offer full tracks as they are competing with radio. We have focused our time and resources on developing algorithms and on making sense of people's tastes. In terms of consuming the recommendations, we have a thirty-second clip deal which enables people to get a taste of the track and then we have partnerships in place for people to click through and buy (iTunes, Amazon, Nokia Music, and soon eMusic). Having said all of that, I do know that the best way to taste a song is to listen to the whole thing. So, it makes sense for us to work on finding a way to offer full tracks in the future."
The Filter had just under 30,000 sign ups to the service since beta and 200,000 users have downloaded the plug-in that taps iTunes. Roberts emphasizes, "while The Filter's algorithms are very good at connecting pieces of content, we know that people trust recommendations from friends and experts. We want to enable people to use each other's tastes as filters to help them discover content that they wouldn't have before." Taking a page from Facebook's mini-feed, Roberts reasons: "If you have a friend who is a genius at discovering R&B acts, then why not just put their activity feed on your homepage and use them as a filter?" Coming from the world of magazines, Roberts foresees The Filter "as the next step on from editors filtering content for readers."
If anything, The Filter underlines the role of social networks. Roberts reveals that The Filter will make it easy for users to import friends and data from MySpace and Facebook. Hedging its bets, it will also make filtering widgets available on the key networks. But will The Filter succeed? Radio listeners log some sixteen hours per week, while car manufacturers report that CDs are rarely changed. Despite this, Roberts reports that he's getting twenty-five plus page impressions per visit on average. For those who study collaborative filtering, there are findings that will both fascinate and disturb marketers: Effective CF - which in The Filter's case combines wisdom of the crowds with one-to-one personalization -- reflects back on the host. The user feels that she is receiving better service. Even before Netflix harnessed its ratings, it reported that two-thirds of its rentals were the result of recommendations. Not only does The Filter offer the potential to enhance one's experience with media, but it will be useful in spotting emergent microtrends. Even as The Filter scales, it will be an ideal environment to break new talent. And of course, CF peers Amazon and Veoh will have to raise their game.