4 ms·

Here's a good (but long) video that goes in depth into why they believe this happens https://www.youtube.com/watch?v=oVfHeWTKjag https://www.youtube.com/watch?v

Here's a good (but long) video that goes in depth into why they believe this happens https://www.youtube.com/watch?v=oVfHeWTKjag https://www.youtube.com/watch?v=oVfHeWTKjag The theory is pay-for-likes click farms click everything they see, including legitimate "like" ads to avoid detection. This increases the number of fake profiles liking your page, which decreases the percentage of users liking your posts, which decreases the number of users posts are shown to (unless you also pay to promote your post...)

johnvschmitt 13y ago
Yes, this is a much better description of the problem. FB is a really bad ad platform. Yes, you can use it & get good results, but that takes more luck & skill than other platforms. From the article: "FB estimates that ~1% of their users are fake generating fake likes". But, if those accounts have 10-100 times more likes than real accounts, you can see that the polluting effect can dominate the overall activity.
- georgemcbay 13y ago
  Also 1% of 1.23 billion[1] is 12.3 million. Which might only be 1% of the total users but is still a very, very large number of fake people. [1] http://www.theguardian.com/technology/2014/feb/04/facebook-10-years-mark-zuckerberg http://www.theguardian.com/technology/2014/feb/04/facebook-1...
  - alextingle 13y ago
    Yeah, and the 1.23 billion is going to be mostly ghost accounts anyway. People who check their FB once a week, if ever.
- rikf 13y ago
  It should be possible to determine fake from real accounts using various statistics like number of likes per account average number of time between likes etc etc, with all the brainpower at Facebook I am surprised they haven't significantly mitigated this problem. This is not a problem that can ever be completely solved as its a typical arms race between good and evil but i am surprised at how badly Facebook appears to be doing as others have noted perhaps its because it in there best interest or maybe its just more visible on Facebook then it would be on google addwords or similar.
  - jonknee 13y ago
    It could be devastating to their business. Facebook is in a tough spot, they are done or nearly done growing in the US and much of the developed world. Most of this fraud seems to happen in emerging markets, the exact same ones much of their "growth" is coming from. If growth numbers go negative or even flat Wall Street would run away from FB faster than you can cancel an advertising campaign.
  - fblp 13y ago
    Yes, we wrote about this at http://fbk.io/stopfakelikestoday http://fbk.io/stopfakelikestoday. In short Facebook could compare behaviour as you suggested (they've got a whole bunch of data scientists). An even more robust way would be making their phone number verification more thorough.
- trothoun 13y ago
  Presumably that 1% is continuously depleted by facebook's detection and deletion of fake accounts and continuously replenished by the creation or subversion of new accounts. A high turnover rate may drive the proportion of fraudulent activity on facebook still higher.
- sjg007 13y ago
  Retargeting seems like the only real viable option.
nhebb 13y ago
That video explains one aspect of the issue (the fake clicks), but what I found most troubling from the article was the five-fold discrepancy between the FB link count and the Bitly and Google Analytics visitor counts. If anything, you would expect Facebook to be able to detect at least some of the fake ad clicks, so their count would be marginally less than Bitly and GA detected.
nyrulez 13y ago
This is a great video. What I find very confusing is the fact that Facebook has no spam detection to eliminate fake clicks/likes ? Given the obvious nature of fake likes, some of the heuristics seem very straight forward i.e. - just filter profiles with an insane number of likes - have very few friends but relatively a huge number of likes, - certain regions in the world where the brands that the users have liked have no market. It looks like a straightforward matching problem. - Suspicious Like activity and patterns from an account (like/click arrival rates, diversity in liked campaigns etc). - De-duplication of profiles that may have liked a certain entity by clustering on some obvious signals e.g. if too many disproportionate likes all fall in egypt relative to other primary markets, that would be an obvious red flag. All of the above is not difficult to figure out for the quality of talent Facebook has. Here is my real confusion: Is it a lack of inclination for improving their ad network quality or a fear of exposing the low ROI of ad spend that's stopping Facebook from implementing quality spam/fraud detection ?
- amaks 13y ago
  I'm pretty sure they do, but they are also torn by the apparent dilemma between generating more revenue through fake clicks and maintaining an appearance of the trusted ad platform. Hard to be them;-)

johnvschmitt 13y ago

Yes, this is a much better description of the problem. FB is a really bad ad platform. Yes, you can use it & get good results, but that takes more luck & skill than other platforms. From the article: "FB estimates that ~1% of their users are fake generating fake likes". But, if those accounts have 10-100 times more likes than real accounts, you can see that the polluting effect can dominate the overall activity.

georgemcbay 13y ago
Also 1% of 1.23 billion[1] is 12.3 million. Which might only be 1% of the total users but is still a very, very large number of fake people. [1] http://www.theguardian.com/technology/2014/feb/04/facebook-10-years-mark-zuckerberg http://www.theguardian.com/technology/2014/feb/04/facebook-1...
- alextingle 13y ago
  Yeah, and the 1.23 billion is going to be mostly ghost accounts anyway. People who check their FB once a week, if ever.
rikf 13y ago
It should be possible to determine fake from real accounts using various statistics like number of likes per account average number of time between likes etc etc, with all the brainpower at Facebook I am surprised they haven't significantly mitigated this problem. This is not a problem that can ever be completely solved as its a typical arms race between good and evil but i am surprised at how badly Facebook appears to be doing as others have noted perhaps its because it in there best interest or maybe its just more visible on Facebook then it would be on google addwords or similar.
- jonknee 13y ago
  It could be devastating to their business. Facebook is in a tough spot, they are done or nearly done growing in the US and much of the developed world. Most of this fraud seems to happen in emerging markets, the exact same ones much of their "growth" is coming from. If growth numbers go negative or even flat Wall Street would run away from FB faster than you can cancel an advertising campaign.
- fblp 13y ago
  Yes, we wrote about this at http://fbk.io/stopfakelikestoday http://fbk.io/stopfakelikestoday. In short Facebook could compare behaviour as you suggested (they've got a whole bunch of data scientists). An even more robust way would be making their phone number verification more thorough.
trothoun 13y ago
Presumably that 1% is continuously depleted by facebook's detection and deletion of fake accounts and continuously replenished by the creation or subversion of new accounts. A high turnover rate may drive the proportion of fraudulent activity on facebook still higher.
sjg007 13y ago
Retargeting seems like the only real viable option.

nhebb 13y ago

That video explains one aspect of the issue (the fake clicks), but what I found most troubling from the article was the five-fold discrepancy between the FB link count and the Bitly and Google Analytics visitor counts. If anything, you would expect Facebook to be able to detect at least some of the fake ad clicks, so their count would be marginally less than Bitly and GA detected.

nyrulez 13y ago

This is a great video. What I find very confusing is the fact that Facebook has no spam detection to eliminate fake clicks/likes ? Given the obvious nature of fake likes, some of the heuristics seem very straight forward i.e. - just filter profiles with an insane number of likes - have very few friends but relatively a huge number of likes, - certain regions in the world where the brands that the users have liked have no market. It looks like a straightforward matching problem. - Suspicious Like activity and patterns from an account (like/click arrival rates, diversity in liked campaigns etc). - De-duplication of profiles that may have liked a certain entity by clustering on some obvious signals e.g. if too many disproportionate likes all fall in egypt relative to other primary markets, that would be an obvious red flag. All of the above is not difficult to figure out for the quality of talent Facebook has. Here is my real confusion: Is it a lack of inclination for improving their ad network quality or a fear of exposing the low ROI of ad spend that's stopping Facebook from implementing quality spam/fraud detection ?

amaks 13y ago
I'm pretty sure they do, but they are also torn by the apparent dilemma between generating more revenue through fake clicks and maintaining an appearance of the trusted ad platform. Hard to be them;-)