Gigaom are missing the point. It's unlikely that Facebook targeted ads on third party sites are going to convert at a substantially better basis than Google Ads where Google has a huge depth of expertise and data on optimising for site content.
The biggest problem in ad networks isn't cost-per-click, it's fraud. And that's where Facebook has a huge advantage, Facebook has a much deeper level of expertise of telling a real user from a fake user. And that's Facebook's edge if they decide to roll out an ad network.
Is Facebook doing anything about these fake users though? I just started a page like ad campaign on FB (I got a free $50 voucher so I figured what the heck..), but of the ~120 likes I got it seems like 98% of them are fake, or at least people that like 1000's of different pages and don't seem to convert to my actual site at all.
Generally the above happens when you don't use targeting on your Facebook ads. Different cultures and groups treat "likes" differently, so you shouldn't be surprised if some groups are a bit gun-ho when it comes to liking pages. Generally I've found that if you're using targeting the US or UK using keyword targets that it isn't generally a problem.
"The biggest problem in ad networks isn't cost-per-click, it's fraud. And that's where Facebook has a huge advantage"
wat? In what world do you live in where Facebook has a huge advantage, let alone any advantage, over Google who has been tackling the problem of click fraud for 10+ years. FB doesn't even do a very good job of detecting fake accounts (I have a few active ones of my own).
They have access to a huge amount of data that Google don't.
The problem of detecting fake accounts is different from that of detecting fraud. Let say you have a probabilistic model that assigns a score to an account based upon how likely it is to be fake, what do you do with this information ?
To detect fake accounts you'd set a threshold which if you drop below a certain score you'd start to suspend accounts Lets say you're right 99% of the time, that still means you're going to suspend hundreds of thousands of legitimate users. Suspending a real account is far worse than letting a hundred fake accounts continue to exist, so Facebook errs on the side of letting fake accounts continue to exist (but they do for example delete likes from what they suspect to be fake accounts as that has a much smaller false-positive downnside).
However to detect click-fraud you can look at the scores across the range of accounts that are clicking on an ad. If you detect a high amount of activity from a group of users who are at your threshold level for being "fake users" you can apply a lot more sophistication to your click fraud model.
Because click fraud means higher customer acquisition cost for advertisers, which means advertisers stop advertising.
It also drains advertising dollars away from legitimate content producers to fraudsters meaning that they switch away to other ad platforms.
To maintain a two-sided market you need to ensure both sides of the market are healthy, click-fraud damages both sides of the market and so is very dangerous.
Which is a nice theory, but there's apparently already a huge amount of click fraud in Facebook advertising and Facebook haven't done anything about it so far.
The biggest problem in ad networks isn't cost-per-click, it's fraud. And that's where Facebook has a huge advantage, Facebook has a much deeper level of expertise of telling a real user from a fake user. And that's Facebook's edge if they decide to roll out an ad network.