Facebook's Blind Spot

Yeah--but what I care about is: What Are You PRETENDING Not To Know?

Facebook knows--or at least ought to know--a good deal about me. It knows my exact age, where I work, and what I look like (it would've gotten the Marsellus Wallace question right first try!). My foray into data-mining with the excellent Digital Shadow app (which mines Facebook) wasn't a home-run but it did realize that I was likely to be "found at the gym" and while it underestimated by net-worth by a lot it wasn't ... I don't know ... it wasn't absurdly bad.

It also noted that I keep my FB account pretty secure. Well, yeah--but not from Facebook. They see everything. When I first looked into private FB data-mining the Wolfram Alpha Facebook report app wasn't working. Today it was--so I took a look.

The Grid of The Known

That grid above is a way of quantifying risks, knowledge assets, and insight. You can leverage things you know that you know. You can look for insight in the don't-know-that-you-know box. You want to ready yourself for personal black swan events coming out of the don't-know-you-don't-know sector. You do research in the Know-that-you-don't-know category.

What Facebook Knows It Knows

Anyone with access to Facebook should know the following:

• Name, age & location / address
• Marital status, gender (I bet anyone using the Facebook alternative genders gives them a massive edge in correlating information)
• Job, home town, etc. (if you have told it--FB keeps asking my home town and I keep ignoring it)
• Visual ethnicity
• An indication of social activity--even if only a vague one.
• A friends network

By combining this with a commercial data-mining company they should know:

• Value of home and proximate income based on ZIP code
• Approximate family size and number of people in home
• Information about magazines subscribed to (those lists are easy to get) which will give some indication of interest (my home receives medical journals and cooking magazines)
• Number of kids (estimated, at least).

This tells you the size of the "whale" you are: it should give, within a range, major-life milestones you have experienced (and will), approximate earning-value over time, and some general ideas about my capabilities. For example, if my picture shows me in a wheelchair, today, there is no reason for a computer to miss that.

What Facebook Knows it Doesn't Know

Facebook knows it doesn't know my hometown. It has some guesses and it keeps asking me. That's its method of doing "research." Notably, it doesn't incentivize me to answer its question so I haven't. I'm also not sure exactly what my 'hometown' is--is it where I was born (and lived until I was 5?) or where I "grew up"? Facebook doesn't say and I see no advantage in telling it that.

Facebook knows it doesn't know my cell phone number and other similar location data (work address?). It has a way to try to get people to ask me that (apparently: I've not seen this feature).

What Facebook Doesn't Know That It Knows

Facebook 'knows' the true identity of The Omnivore as I own The Political Omnivore's page. It 'knows' that I regularly buy political analysis books. It knows that I am interested in certain stories and publications. This material never shows up in my ad-stream for my real-person identity. Is this simply because no one is trying to sell political books? Or is it because Facebook is not connecting the dots.

Facebook has enough information that, if it correlates with what a human could do with a Google search it would find out about my hobbies. It's possible it does know something about them (see below) but if so, it is doing a fairly bad job. The Omnivore regularly likes geeky memes other people put up--but there's nothing in my ad feed that suggests that. The Omnivore goes with some regularity to Comic Con (as press)--nothing there. The Omnivore plays computer games ... nothing.

And so on. This is all quite discernable from my feed (pictures with Zach Levi, for example) but no one is trying to sell me Chuck DVDs ... which I might actually buy.

What Facebook Doesn't Know It Doesn't Know

I think Facebook has a blind spot in the don't-know-don't-know space. On Wolfram Alpha's report it looks like this:

This is hard to see--but it's a map of my friends (each dot is a friend) and how they connect to each other. When you mouse-over you see the friend's name and picture. Now, to you this looks like gibberish. To Facebook, it's a wealth of data--lots of it structured (the names, ages, etc. of each person). To me it's a country map. What do I mean?

I've divided these up by how I see them. To Facebook, I am sure, this looks very similar. Facebook, if it is looking--and it is looking--understands a lot of this. It knows who I went to high school with (it pretty much has: its best guess about my hometown is, in fact, correct).

It doesn't seem to know where I worked at my dot-com (HotOffice: we were all going to be millionaires) but I'm sure it could figure that out. If the WatchDogs application knows I can be found at the Pure Muscle gym--and it learned that from Facebook, the red-state should be pretty transparent to it.

The location of all my Alabama relatives ought to be a give-away too. And so on. These are nuts and bolts--individual facts--that Facebook has in spades.

What Facebook isn't hip to--its blind spot--is what value I put on each country. I don't think Facebook is looking at the influence value of each 'nation.'

The Value Proposition: Influence Level

What each 'country' has, as a whole, is an influence level over me. I have different status-levels in each one and different levels of emotional investment and real-life interaction. I also have aspirational goals in each country (I want my red-state relatives to read The Omnivore) and different levels of competition and such (do you want to look like a winner to your high school class? Yes, yes you do).

If Facebook understood more than just "who my friends of friends are" it could understand things like: in each nation, who's the leader? The person in Lake Martial Arts with the most friends is a high school kid (2196 friends :-0). The 'leader' of the nation? That'd be the Master Instructor. Does Facebook know that? Are they even trying to figure that out?

The Dot-Com expatriates all have had pretty good career arcs--but there are only some there that are capable of granting others high paying jobs ... Has Facebook figured out who they are? Is there some way to leverage that in advertising? 

It turns out we don't actually have to guess--we can see it: Let's go look! 

Let's Advertise The Omnivore

If you want to see what Facebook knows about people in general, that's not hard: just go launch an ad! What are we going to advertise? Why, The Omnivore. Let's take a look.

Start here: Facebook - Create An Ad!

The top of the page is text and links--but then you get down to the good stuff that we care about: targeting.

The Omnivore's Audience: Well To Do Educated Non-Partisans

If you look at the various drop-downs they have Interests (pages you have liked or items 'you have expressed an interest in). They have Behaviors--stuff you've done like ordering office supplies. You can choose connections: people with a direct connection to, say, your page--or friends of them.

That last bit--where you expand around your target's friends--that's where I think they're missing things. I think they're missing things because of the general poor quality of ads and that even if you correctly target each of my 'Friends' the 'country' they 'live in' will have very different ramifications when the ad shows up on my page.

Facebook has a fairly complex taxonomy of ads. For side-bar ads you won't know exactly why you were targeted--but for Sponsored Stories where I see an action that Friend-X took show up on my timeline (because the company promoting the action sponsored it) I will know who it came from. 

Depending on which 'country' that friend is from will have a very different impact on how I perceive the impact of that statement. The advertising taxonomy doesn't seem to recognize this. It has 'connections' but is absolutely silent as to the 'type' or 'strength' of connections.

The Dark Continent

If Facebook could see my GChat and Skype records they'd know who I talked with in that cloud most often. They'd know who I never speak to. If they had access to my mobile phone's GPS logs they'd know who I visited in real life most often. They'd know who I never visited. 

If they were smart--real smart--they'd know who on that list has kids my kid's age and whom I hang with when I can. They don't seem to.

The demographics will let you target, say, Hispanic Moms making over 100k per year. They don't let you target "influencers" in the lives of those moms. They don't have a section for "close friends of your target audience" or "real-life friends of your target." (At least not that I see there--I do believe you can get fancier if you get more heavily invested in Facebook ads).

The Omnivore would like to target people who follow David Frum on Twitter: there's probably no perfect way for them to do that--but there might be a way for them to guess (you can see followers, you can try to match them to FB users--I bet there's a correlation that is 90%+ if Facebook is really trying).

This dark-continent of the strength and importance and emotional-weight of relationships is something Facebook has the capacity to know. Is it trying?

Conclusions

The thing you most want from Big Data is insight into the Don't-Know-Don't-Know box. Insight in the Didn't-Know-You-Knew is where you can drive innovations from. The DKDK box is where you get caught with your pants down and some up-and-coming ap eats your lunch. Facebook has one of the largest data collections in the world and one of the most unique--so its DKDK box is smaller than most--but I suspect that one of the drivers for its purchases is to fill that box in with stuff like real-life interaction data that it may find hard to get.

I wonder if it's looking at categories of friend groups and trying to separate them out by influence. If it was, it might have another set of advertising options based around that the same way that real-life ad agencies try to get certain people in the real environment (I don't mean celebrities to be first-movers).