The big thing for Facebook was always their gigantic inventory, never so much the information. They have a few advantages here, but i think Age + Location + Sex are still more important than all your likes together.
Perhaps, but a better predictor would still be whether or not the person has searched for anything related to luxury goods in the past, thereby demonstrating not only ability, but actual intent as well.
And unfortunately they refuse to produce this history when asked (as is required by EU privacy law).
In any case it should be interesting to see how they will implement this: Their tracking like button seems to have survived largely based on the fact that people don't know about the tracking. When it comes under scrunity from privacy commissioners (for example in Schleswig-Holstein in Germany), it unsurprisingly seems to be considered to violate privacy laws. Without a similar mechanism it will be hard to serve personalized ads, but general awareness of Facebook's (and other's) privacy practices seems to be growing...
All of this is low-friction data production unlike the review example you use above (most people are terrible at writing original content unless absolutely required to do so).
I'm just this will rattle the tinfoil hate brigade but this certainly isn't "selling your data" as such. Publishers would make a call to FB Ad servers that based on whatever algorithm they choose, would then serve you ads, just like every other display business.
Facebook, in addition to knowing what you do, what you've liked and who you're connected to on Facebook, also has a history of pretty much every Website you've visited since they introduced the "Like" button. The comparisons to Google's intent-based advertising on Google search is somewhat misleading as this business would be closer to, say, Google's DoubleClick, a much smaller business that operates on CPM rather than a CPC basis. If Facebook operated a display business on a CPC basis, that would be noteworthy.
I agree with the revenue estimate of adding as much as $5 billion a year. I came up with the same figure myself at Facebook's IPO to give myself a price target of $12-15 (assuming a gross margin of, say, 20%). That estimate might be low but it's not that far off the (current) mark.
Personally I view these two data sources as:
1. What you say: behaviour on Facebook; and
2. What you do: the sites you visit.
(2) is what display advertising has been built on to this point and (IMHO) can present a very accurate picture of what your interests, etc for ad targeting.
(1) is the big unknown. My personal opinion is that the value of this data is a lot less than many predict. Actions speak louder than words. My own experience with Facebook is that it presents a narrow view of yourself (in most cases), presents the world how you think you're perceived or how you'd like to be perceived or both.
The problem is that so few people are data producers. Most are data consumers. It's the whole review problem with local: your audience is so small that getting any reviews is hard and no matter what you do (IMHO) the majority of people just aren't going to go around recommending or "Liking" their local plumber.
So time will tell on the value of social signals. I for one am moderately bearish.
Disclaimer: I work for Google in display advertising. All of the views expressed are my own and do not represent the views of Google.