Many of you have heard about Facebook announcing a new business model: organizing instant census. Thanks to their wide adoption, most marketters can expect to find their market on that website; even better: thanks to Facebook strong enforcement of their “real person” policy, most information is accurate; in addition to that, it could relie on a relevant social graph, which we know since Huberman & Adamic’s study on (Amazon) recommendations, offers priceless insights through word-of mouth; but, most importantly, thanks to the intense traffic, such questionnaires can polls an astounding number of person in a matter of minutes. I’ve worked at a official statistical insititute, and that many responses are worth millions; the possible level of insight is unmatched: even within focus groups where you can’t expect to ask half of what Facebook knows; the possible inclusion of social graph makes this the marketting equivalent of a nuclear explosion.
Am I scared on the privacy side? No: unless Facebook starts to sell its information to marketters with a market of one — and it would not be worth it to do it otherwise then directly through that person — it is in everybody’s interest to respect privacy through statistical secrecy. As shown recently by Monica Chew, Dirk Balfanz and Ben Laurie, statistical secrecy is far from perfect, but there are several orders of magnitude that can be used as a cover. When Facebook will sell polls to less then a thousand respondant with narrow focusing and several identifiable questions, we might have to worry.
This doesn’t resolve the issues around the fact that a private company holds that data — but I’d rather have insights on people’s opinion shared without the usual gaming that pollsters love to add, and I don’t know any better solution for preventing the social graph to end in the wrong hands than to have competent people (knowledgeable about how to reveal statistical secrecy) control it.
This makes far more obvious why Facebook refused to erase data from users that left: they still need information about common friends for a better network-based interpretation — not to show to anyone, but to consider, aggregate and make sense out of larger samples. It is still rather worrying to have to trust them with the security of everything, including information clearly marked as obsolete, though.
I used to reply a lot that Facebook didn’t sell users’ details; they don’t sell contact information, but now they do sell information-to-go — and it’s the first time. Their offer as an ad agency was worryingly detailed, but they never sold information per se: they screened ads.
What makes me most happy about this? My advisor has this great general model about Digital Business model, where a provider has to balance betwen his role as a aggregator, match-maker and information manager. I’m not sure any manager at Facebook read it, but this announcement wouldn’t be any different if it was directly inspired by that talk. I present it often (as I want to apply it to Web 2.0) and the lack of a diversity of information managers is now compensated: with SAP, Yahoo!, Google and Facebook now in the same boat, the third option is the cool kid again.
What makes me most unsatisfied? Well , I haven’t seen any demo yet, so — most of it: basically, the annoncement lacked many essential details (I guess a live demo would make some of these clearer):
- What does the questionnaire look like? Does it encourage users to fill in? Is it annoying, does it give points? Can it be long, multimedia, interactive, etc.?
- It is possible to grant access to delicate information already known by Facebook, rather then reveal them by typing it in again? Or is it possible to reply without Facebook storing the information; in that case how does Facebook guarantess the accuracy of the information without the social filter?
- How are the participants chosen? Are all the users in the considered section sampled until a threshold is reached? Is there corrective methods applied, based on demographics? What about previous inquiries?
More basically, it could includes a lot of bias; among the many questions I wouldn’t trust to ask:
- Do you like Facebook, on-line Social Networks, computers, answering to polls?
- If a polls is taken at 6AM; do you wake up early? Same for 11PM.
- Would you answer (truthfully) a question from someone whom you do not know?
and — and there’s the real issue — all the questions correlated to that: altruist, early bird, tech-lovers are not representative of the general population.
I guess that Facebook will not bannish people who never reply: however, they need to think of a carrot that is neutral to the polling. Without it, most won’t reply, and those who do won’t be representatives of them — not that this isn’t any different then current telephone-based samples, but at least those aren’t perceived as miracles like this Facebook tool seems to be. What might make this exclusion mechanism stronger is the widespread fear of Facebook knowing too much; if you want to know about a compnay who isn’t afraid to go straight into the face of all it’s loudest critic. . . Let’s say they didn’t try to address their reputation issue. It might cost them everything, or just like NewsFeed, Zuckerberg might be both right and capable of changing to world to make his point.
Edit: the project is actually an extension of “Engagement ads” that had been tested with a few selected announcers and that I’ve seen. It’s a discreet, well-designed feature with only room for one question. It might be subject to framing, doesn’t seem ready for very interactive, multimedia, longer polls, but could be modified in that direction. The real question will challenge two explanations for the lack of click-through on Facebook ads: do people neglect them because their aren’t as juicy as party photos, or because they aren’t as interactive as commenting? If the project is lauched after testing, I guess this means that at least some users decided that a dialogue was a good thing in a merchant/customer relation. That would make me glad.