Barracuda Labs (BLabs) has posted a statistical summary about Twitter accounts that pay money for followers. They call it an underground economy.
In short, they setup a Twitter account and paid for followers to find others who did the same. They then recorded this correlation as data on “fake” accounts.
- There were 72,212 unique fake accounts identified
- 61% of these fake accounts are less than 3 months old (since April 16th, 2012)
- Average age of these fake accounts is 19 weeks or about 5 months
- 55% of fake accounts have ~2000 followings
- The average number of following for a fake account is 1,799
One issue I see with this summary is the lack of a thorough definition of what constitutes a fake account. They only say the following:
(created by dealers for selling followings or tweets business)
Is an account fake because it is not a one-to-one ratio with a user? Is it fake because it is created for business purposes? Their brief sentence, which we are forced to use as a definition, seems to contradict a BLabs Facebook fake profile infographic one-liner.
…the sole purpose of this fake profile is to entice you into into befriending them.
Sole purpose? What if there is also a business purpose, like with Twitter accounts?
A definition of fake needs to be clarified. And on that note an interesting question is whether BLabs could define as fake the account they setup for research.
Why am I reminded of the Three Laws of Robotics?
There also is no statement of why BLabs consider this an underground economy other than to say in their conclusion that it is a violation of Twitter’s ToS.
Finally, creating fake Twitter accounts and buying/selling followers is against Twitter’s ToS, and gradually erodes the overall value of the social network. Twitter keeps on detecting fake accounts and followings, and suspending them in last few years. However, if they do not move faster and smarter, these fake accounts will continue to be created, blended into the massive Twitter population, bringing bigger and bigger impact.
And the impact is? Eroded value of a social network? At the start of the BLabs post it seemed to say that they were out to protect their customers. Yet their analysis of impact suggests only a weakened Twitter.
They are right about the Twitter ToS, which includes the Twitter Help Center Rules. The section on Spam and Abuse is quite clear.
Username Squatting: You may not engage in username squatting…creating accounts for the purpose of selling those accounts
[…]
Selling user names: You may not buy or sell Twitter usernames
[…]
Your account may be suspended for Terms of Service violations if any of the above is true.
The Twitter profile economy thus could be a grey market activity (legal but an unauthorized/unintended use of goods) unless BLabs investigators are able to prove that all those selling accounts are in violation of actual legal statutes.
The California EDD offers this definition, which emphasizes a lack of government oversight and regulation.
“Underground economy” is a term that refers to those individuals and businesses that deal in cash and/or use other schemes to conceal their activities and their true tax liability from government licensing, regulatory, and taxing agencies. Underground economy is also referred to as tax evasion, tax fraud, cash pay, tax gap, payments under-the-table, and off-the-books.
The BLabs data is very informative and interesting but lacks thorough analysis. A clear definition and more detailed report on the economics would help support their final conclusion. I can agree with their accusation against Romney but only if I agree with some open and uncomfortable assumptions.
…we believe most of these recent followers of Romney are not from a general Twitter population but most likely from a paid Twitter follower service.
Romney is probably buying his popularity and Twitter is taking their time to shutdown accounts that violate their ToS. We have confirmation of what we would already know to be highly likely. We do not have proof of illegal activity or an underground economy.
Ultimately it is easy to find fraud and tell Twitter there is a problem without carefully defining it. It is much harder to profile a threat in order to tell them exactly how to detect and prevent fraud without alienating real users.
Update 8/14: Alex Hutton tweeted about Status People’s faker tool. I used it on Romney’s account; it samples up to 500 accounts and offers the following results:
@MittRomney Faker Scores
Fake 12%
Inactive 30%
Good 58%
To be fair, although this is like Gawker’s report last year (based on PeekYou analysis) that 92% of Gingrich’s followers are fake, PeekYou’s CEO put it like this:
Using algorithms to determine whether an online presence is real or fake is obviously more art than science