Why Facebook’s Open Graph Will Fall Short In Converting “Social Proof” Sales
A question I’m often asked is “What’s strategic for Facebook?” My answer is – think of the most lucrative online consumer markets, and assume that Facebook, at some point, will try to bring those markets within its “walled garden.”
Last week Facebook announced the launch of “Open Graph,” a protocol which allows any website to add Facebook’s “like” buttons to their content, as well as tap into Facebook’s social graph data. The “catch” is that all data from said “liking” will revert to Facebook, adding to the meta data that it already stores about each user. This is somewhat troubling, as the quantity of data collected by Facebook dramatically enhances their ability to accurately predict your tastes and influence your future purchasing behavior. (As Chris Dixon points out in a blog post from last year, systems get smarter not by inventing new algorithms but by creating new sources of data). Facebook has created a wonderful new source of data – your browsing history and preferences throughout the whole internet. Many bloggers are predicting that Facebook’s internet domination is now unstoppable. While I agree that Facebook is almost certain to be a dominating force for years to come, I think it’s important to point out its weaknesses, some of which could be exploited by existing competitors or startups, and put a serious damper on what many are concerned will be a private data monopoly.
Facebook’s underlying assumption is that social proof (items that your friends have “liked” or purchased) will make you more likely to purchase a particular item. But on Facebook’s website such recommendations are out of context; they bring purchasing/liking data to users while they’re in “social networking mode” instead of the much more effective reverse – bringing social purchasing/liking data to users while they’re engaged in behavior that shows purchasing intent (e.g. they’re on an e-commerce site). Open Graph solves this problem by distributing social to all points of the web where social proof could influence user behavior.
Does this mean it’s “game over” for anyone trying to compete with Facebook in this space? I say no. While this is an aggressive step forward for Facebook, there’s still a big weakness –“social trust”– relating to who’s in your social graph and why.
Social proof generally works in one of three ways. Either a product or service is 1) best selling in its category, 2) recommended by someone with relevant expertise, or 3) recommended by someone who you recognize as sharing similar tastes/preferences. Method 1 is irrelevant to this post, as this data is readily available, so I’ll speak to the other methods.
Relevant expertise (in humans) is something that both Facebook and Google are missing – something which Aardvark attempted to address before they were acquired by Google. It’s an important key to converting via social proof – which leads to the real question- is there really that much expertise in your Facebook social graph to harness? My guess is no, it’s too limited. Facebook is more of a graph of people you know, whereas Twitter is a graph of people you know and people you wish you knew, which makes it a lot more useful in determining who has thought leadership in which domains. Facebook might ultimately decide to pull down the barriers of your siloed graph for purposes of conversion – if they can determine who among their 500 million users has enough relevant expertise within a domain to monetize traffic most effectively. Until then, I’m not sure I’ll buy my next computer based on what my Aunt Gertrude bought.
The third type of social proof is when someone who you recognize as sharing similar tastes/preferences recommends a product or service. Amazon should have been all over this a long time ago. Currently Hunch.com is doing a great job tapping into the “taste economy.” Their AI is smart because, as Chris says, they’ve created new sets of data from which to draw. This is also Facebook’s biggest strength. They’ll have billions of data points with which to create terribly accurate taste prediction algorithms. But – the output is a recommendation from a computer program, not from a person. Recommendations from algorithms aren’t bad, but I think the conversion rate is much higher if the recommendation is associated with a person who you recognize as having similar tastes to your own. While the chances of these people being in your Facebook Graph vary, the chances of you recognizing them as people who have similar tastes as your own is pretty low. In fact, I’d argue that on Twitter, which is more of a “peer graph,” these people are more likely to exist on your graph AND you’re more likely to recognize them as having similar tastes to your own, as these preferences are often revealed through dialog.
These issues of trust, both the trust users assign to a recognized “authority” within a domain, as well as the trust users assign to those with tastes similar to their own, are integral to making meaning of any social graph. This is why the biggest hurdle faced by Facebook or any platform trying to improve monetization of social transactions is the determination of who within a graph owns the trust of a user, and in what domains a user trusts that person.
This is the key to converting social sales. The recommendation engines will get the clickthrus, but only trust will optimize the conversions.