Why buying followers on Twitter ? That is not really the question to ask anymore as there are so many other articles, posts and tweets done on this subject that can be read on various blogs and twitter time lines. All of these have been written by all kinds of twitter users, so called ‘specialists’ and other ‘community managers’, all of them arguing on its usefulness while praising it or, on the contrary, decrying it. Called SLEIPNIR*, this Pegasus Data’s operation rather try to decrypt technically the procedure of buying Twitter followers, the ‘how to’ buy and its consequences. This operation only serves the purpose of conducting an experiment that is placed beyond all moral debate.
As we are learning that accounts’ statistics of celebrities are almost without exception swollen by "ghost followers", our experiment is here to break the taboo that exists in the so-called ‘All-is-well-in-the-best-of-Twitter-world.’
Firstly, we recommend you to read this very well sourced and recent article of the New York Times on the matter: Buying Their Way to Twitter Fame.
Table of contents
- The transaction : 30.000 followers for 10$
- The execution of the task : precision’s monitoring
- First results
- A glimpse on the purchased followers
1. The transaction : 30.000 followers for 10$
Our first condition was obviously to find a service that was avoiding heavy costs for our volunteer structure. A lof ot offers exist on various specialized websites, their prices varying substantially, from a range of 0,15ct/follower to 1.6ct/follower !
Avoiding these maybe too well organised websites, two offers of Fiverr caught our attention. Fiverr is a sale platform where one can find services for 5$. The price offered there is unbeatable (between 0,02 and 0,05 ct per follower):
Why 2 offers ? For our first purchase (10.000), we planned a backup of our Twitter account’s stats that should be occurring every hour. It turned out that the ‘delivery’ would be concentrated in an interval of a few minutes, and thus would not allow us to analyze the arrival of new subscribers/followers. Because of this, we proceeded with a second purchase (22.500) using a server being able to record and keep a track of our data every 12 seconds. These two offers would also allow us to compare panels of followers in order to try to understand the different modalities of such transactions.
A reminder : there is no need of creating a Twitter account in order to buy followers. It means that it is possible to buy some for somebody else’s account : a Twitter account that contains amongst its followers some ‘robots’ did not necessarily purchased them ; it may have received some without any action taken on his part.
2. The execution of the task : precision’s monitoring
The infographic below (click on it to enlarge, then use the zoom of your browser) brings out the ‘delivery’s procedure’ of the purchased followers (the @PegasusData account had 2705 followers before this operation). The Graph 1 (1 hour gap) shows clearly the two different marked levels going suddenly up while the Graph 2 (zoom on the second delivery, with the 12 seconds gap) brings out two delivery’s rythms happening during what we called the ‘SLEIPNIR 2’.
While the first delivery, carried out 23h after order, exceeds more than 3% of expected followers (10.374 followers acquired in this purchase, but there has been a quick decrease afterwards), the execution of the second delivery encountered a problem :
For an unknown reason (we’re still investigating on an possible restriction by Twitter that could have led to it), the ‘deliverer’ was unable to add the 22.500 expected follower (only 20.853 were delivered) and thus asked us to deliver the remaining followers to another account !
3. First results
- The @PegasusData account’s Klout score is not experiencing a serious increase, running from 60 to 61. This being said, we must bring some nuance to this observation as Klout modified its website’s structure in the meantime. Furthermore, it is frequent that its algorithm (by the way quite abstruse) may undergo changes without prior notice, even sometimes modifying data a posteriori.
- Some tools such as Twitter Counter are unable to correctly monitor the increase. These tools don’t require any identification through OAuth, this allowing anyone to check other users’ statistics. In fact, it operates on the basis of an average. In our case, it spreads the inscrease over four days !
The tools that require an identification are more accurate. Here, Crowdbooster, Twentyfeet and TwitSprout :
- The Faker.StatusPeople tool reflects the change about the ‘quality’ of the followers, but on a very irregular way. The analysis’ results of this gadget tool are to be taken with caution because they can quickly bring charges against a Twitter user with ‘bad’ results. In our case the results are :
- Before the experiment : 0% Fake 3% Inactive 97% Good
- Following SLEIPNIR 1 : 64% Fake 5% Inactive 31% Good
- Following SLEIPNIR 2 : 95% Fake 3% Inactive 2% Good
- Three days later : 93% Fake 3% Inactive 4% Good
As we have recorded precisely our account’s data, we can put these results in terms of exact data (the ‘inactive’ being merged with the ‘Good’) :
- Before the experiment : 0% Fake (0) 100% Good (2705)
- Following SLEIPNIR 1 : 79% Fake (10.374) 21% Good (2705)
- Following SLEIPNIR 2 : 92% Fake (30.907) 8% Good (2705)
- Three days later : 92% Fake (30.694) 8% Good (2705)
This shows that the assessment tool fakers.statuspeople, though far from the reality after the first purchase, is correct about the final result (even slightly too severe, though).
4. A glimpse on the purchased followers
In a future article, we will detail more precisely a followers’ panel in order to decrypt the identity and, if appropriate, the strategy of creation/follow, etc… A study of the network will eventually provide some interesting elements about the organisation of these accounts, that are not as dead as we may think. We also will write about implemented schemes to lure algorithms generated by Twitter. In the meantime, here are a few observations concerning the panels of followers acquired through our purchase :
Besides the many accounts with no profile picture, one can see that this first panel of 10.000 followers does not only comprise, as we will see for the second panel, inactive accounts. In fact, some of them are pretty active ! These accounts are not only used by their owner(s) to follow other accounts, they also serve the purpose of the Retweet. As the sale of followers, the sale of RT’s is also a lucrative market. For example, the last tweets of three accounts, randomly taken : @ArleneWeaver, @tanisharahaman and @JulieBronson1 :
Among these tweets, all being retweeted many thousands of times, one can find, for instance, a tweet on an account followed by only 8 people, but retweeted 3017 times. One characteristic of these purchased retweets is that they are rarely accompanied by a significant number of "favorites".
We quickly noticed that the followers acquired through this transaction are of lesser quality than the previous ones. Only a few of them have a biography (almost a ratio of ¼), usernames are long and complex, 2/3 have a Twitter egg as a profile picture, and most of them have some typical ratios such as 894 following/12 followers. Their rare followers are not similar of them. It seems they look like some normal accounts that automatically follow back. Another quasi-systematic element : 19 out of 20 accounts have never tweeted, not even one single tweet.
In the coming weeks, Pegasus Data will have a particularly precise record of its followers’ numbers’ evolution to be able to appreciate a clear gradual decrease of them (i.e. accounts that stop following, accounts deleted by their owners, accounts deleted by Twitter…). This ‘decrease’ factor will be an important evaluation criterion in the long run to assess the quality of the delivered ‘product’.
A comprehensive study of ‘ghost followers’
A comprehensive study of ’ghost followers’
With a pannel of 30.000 robots, the opportunity is too good to do a thorough analysis to document this phenomenon of purchased followers, more precisely. Let’s meet again, later, to discuss a second report !
Does this study leave you the bitter taste of breaching the idealised rules of the small (or big) world of Twitter, made of respect, honesty, and accuracy, etc ? That might have been our first reaction. We need though to nuance our own reaction while taking in account the multitude of uses one can do of Twitter : personnal use, professional ones, some for political reasons, for communication, promotion, even some other for selling reasons, etc… To know completely the issues, sometimes being dim that underlies this network in its globality, seems important to us, to reaffirm the precedence of authentic relationships !
Oh, by the way, you have now a full range of gadgets listed above that will help you to verify if the authors’ accounts of this study are full of ghost followers :Follow @grandjeanmartin Follow @yrochat
And do not forget to follow the translator of this post :
*Why SLEIPNIR ? Sleipnir is Odin’s fabled horse in nordic mythology : it is a horse with eight legs that gallops over the seas and in the airs. Basically, it is a supernaturally effective creature, in an « over the top » situation, such as the one we find along with a swollen twitter account composed of followers purchased on the gray market !
Other posts you may like :