Last year we introduced a concept called the Database of Affinity — a catalogue of people's tastes and preferences collected by observing their social behaviors — and proposed that the greatest marketing value of social media won't come from marketing to people on social sites, but rather using this database of affinity to improve the marketing that happens everywhere else. And in 2013, several social networks started to pursue this opportunity: For instance, Facebook launched an artificial intelligence research team and Google started selling "affinity segments" targeting on its properties.
But are social sites going too far in their effort to build the database of affinity? Perhaps. Recently we've seen reports that some social networks are tracking not just the information that you choose to share, but even information you choose not to share. For instance, Facebook has admitted to studying "aborted posts" — the things people type into Facebook (as status updates, in comments, and on other people's timelines) but then choose not to post. Likewise, both Google and Foursquare apparently use their mobile apps track users' locations at all times, even when people aren't actively using those company's apps.
Earlier this year, we introduced the Database of Affinity: a catalogue of people's tastes and preferences, collected by observing their social behaviors, that could be the Holy Grail for more-accurate brand advertising. And since then two of the companies we featured in our research -- Facebook and Google -- have been working hard to realize this vision:
In June, Google introduced Affinity Segments -- a tool that allows marketers to target audiences based on the products and categories for which they've expressed preferences. We think Google has room to add more and broader affinity data to these segments, and to do richer analysis on that data. But Affinity Segments blends multiple signals into a single targeting tool -- which makes this an important step forward from the simplistic affinity targeting most social sites now offer.
More recently, Facebook built a team to analyze its affinity data. MIT Technology Review reports that Facebook has assigned eight people to its 'AI' team. Their goal? To address one of the key shortcomings we'd identified in Facebook's business: its inability to bring meaning to its data. It's always been clear that Facebook has one of the largest collections of affinity data online; we hope this move will help the company better leverage that data on behalf of marketers.
Last month I published new research on the Database of Affinity — a catalogue of people’s tastes and preferences collected by observing their social behaviors on sites like Facebook and Twitter — and how that database will change marketing. And I'm pleased to say I've gotten a lot of great feedback on that research. So I'm excited to be presenting the idea on stage at our Marketing Leadership Forum in London later this month.
What is the database of affinity?
I hope you'll be able to join us in London on May 21 and 22.
Recently we described an idea called the database of affinity: A catalogue of people’s tastes and preferences collected by observing their social behaviors on sites like Facebook and Twitter. Why are we so excited about this idea? Because if Facebook or Twitter or some other company can effectively harness the data from all the likes and shares and votes and reviews they record, they could bring untold rigor, discipline, and success to brand advertising.
But exploiting the database of affinity won’t be easy. Any company hoping to turn affinity data into something marketers can use will need three things:
Lots of affinity data from lots of sources. The raw data required to build a functional database of affinity doesn’t live in just one place. Facebook controls the most "like" data, recording more than 80 billion per month at last check. But Twitter records more "talking" than anyone else (1.5 billion tweets per month); Amazon collects the most reviews (well over 6 million per month); and Google’s YouTube and Google Display Network have data on how a billion people prefer to spend their time.
The ability to bring meaning to that data. It’s easy to draw simple conclusions from affinity data: If you ‘like’ snowboarding you might like to see an ad for energy drinks. But the real value in affinity data won’t be unlocked until we can find hidden combinations of affinity that work for marketing. That’ll require technologies and teams that can do some serious data analysis — as well as a real-time feedback loop to determine whether people really are interested in the ads targeted to them based on such complex assumptions.
For years, brand marketers have guessed at people’s affinities from the barest of demographic, geographic, and contextual clues. We deduce that Midwestern men prefer pickup trucks and that people watching extreme sports like energy drinks, and then we spend billions advertising to these inferred affinities.
But today, we no longer have to guess. Every day huge numbers of people online tell us what they like. They do this by clicking a ‘like’ button, of course — but there are many other ways people express affinity: talking about things on Twitter and in blogs; reviewing things on Amazon and Yelp; spending time with content on YouTube (and telling us where they’re spending their offline time on Foursquare); and sharing things through both public and private social channels.
People’s rush to post their affinities online recalls another flood of data that began a decade ago: the explosion in online searches. John Battelle once described the data created by search as the “database of intentions,” which I’d define as “a catalogue of people’s needs and desires collected by observing their search behaviors.” In the same way, the result of all these online expressions of “liking” has created the “database of affinity,” which Forrester defines as:
A catalogue of people’s tastes and preferences collected by observing their social behaviors.