With the mass cease and desist order leveled against the iPad news aggregation app Zite the other day, the topic of personalized content apps like Zite and Flipbook have come into the media limelight. Personalizing news by way of various kinds of behaviorally based algorithms has been a promise chronicled in these columns many times over the past several years. Try as they might, these systems never seem to catch on. But with mobility and smartphone or tablet apps, the utility of personalization techniques has become more obvious.
In fact, I would argue that the app ecosystem feels more constrained than the typical browser-based approach to digital media. Cross-linking across providers simply doesn't occur as frequently in most branded media apps, since the intention of most designers is to keep the user within the confines of the app. Opening and closing apps or apps and browsers to move laterally from media brand to media brand is more cumbersome in this environment.
Perhaps because of this, we are seeing aggregation apps like Pulse, Zite, Flipboard, and Newsy gain followings. One of the more interesting and under-reported entries in the field comes from an unlikely source, the marketing agency Organic. As part of its Product Development Group, Organic recently deployed the experimental Broadfeed app, which organizes news stories by extracting and prioritizing one's Twitter feeds. Similar products like Zite and Flipboard also use things like user behavior and social media to inform story choice. But Broadfeed uses "social weighting" to prioritize stories. "The primary way our algorithm works is rudimentary," says Dan Neumann, Organic's senior platform specialist. "We're just looking at the number of times a link is referenced in a given feed or list on Twitter. And then based on the popularity of that link and the times it has been shared, we give it a greater prominence in the Broadfeed layout."
Clearly this works best if your own feed has a lot of content coming into it and shared links to track. When you follow more people, the effect is heightened and the distinctions among story popularity can be greater. "The idea is a reimagining of a newspaper's front page and the editorial decision-making about what news becomes headlines," Neumann says.
There is a big difference between the "social graph" one accrues on Facebook and the connection one has on Twitter, Neumann says. Facebook is more often a network of friends and associates, while Twitter connections do not require mutual agreement and are more far-ranging. "On Twitter you may follow people who are not connecting back to you. There is a kind of inferred intelligence in that set of people [or feeds." For the Broadfeed app, Twitter "Lists" can be especially effective. Publishers like Huffington Post and others create lists of Twitter feeds from politicians, reporters, etc., and these lists and the links shared among their followers can give the Broadfeed engine a wide view of across stories that happen to be highly focused topically and then prioritized by the collective brain created by specialists simply sharing what they deem most important.
It is hard to detect in practice how much the sharing behavior in Twitter is influencing the feeds I have been testing with Broadfeed. In many cases, there are just one or two retweets on a story that can move it higher in the rank order. The application lets you tap a comment icon on the bottom of each story page to reveal the original tweets that helped push this story into queue. As Neumann suggests, it is the kind of inferred intelligence that seems to work best with a crowd. Interestingly, unlike some personalization engines that tend to over-filter for topical stories, there is a kind of serendipity built into this engine, because a stray off-topic re-tweet from one of a feed's followers could push a story into the mix.
Why exactly is an agency getting into the business of creating apps without any client attached or even advertising running into it? In part this is a proof of concept and experiment in developing for tablets. "We wanted to get deep experience with the platform and understand user experience, visual design and native development," Neumann says. "It is about developing our level of competence and skill sets," he says -- and about gleaning an understanding of evolving tablet habits. There is also a measurement platform in the app that will give the designers an understanding of how to tweak the app in response to usage patterns.
For over a decade I have interviewed academic researchers and companies trying to make personalized content delivery accurate and appealing enough to get consumers to buy in. In some sense the social filters on information that Facebook and Twitter, Digg and StumbleUpon have given us in recent years may have softened the ground a bit. Platforms like tablets finally seem to be a better fit for the efficiencies that personalized content promise. Of course, there is still always the question of how much better is one approach from another in predicting a reader's tastes. And how much of that process is evident enough in these applications to make personalization by algorithm any better than the personalization that comes from traditional human editors and the stories they prioritize on any page?