How real are your followers?

With elections coming up later this year, every political party in Belgium tries to win the hearts of the voter through social media. Some parties succeed very well in doing so, other parties are extremely clumsy. Earlier this week, CD&V, the Christian Democratic party in Flanders had to explain why they had over 10,000 followers from India, on a total of about 36,000 followers. Moreover, these followers were all acquired in a very short timespan, hence the suspicion that they were acquired in return for money. If that's true, then who paid for those followers? It's very hard to think of a reason why a political party in Belgium would benefit from having followers in India, unless that reason is stupidity.

This morning, the Belgian financial newspaper De Tijd published some more numbers about fake followers on Twitter. Studies indicate that only about 40 to 60% of the Twitter accounts are "real people"; the other accounts being automated bots. The newspaper mentioned a couple of tools (such asTwitterAudit) that can be used to count how many of your followers are real, and how many are fake. For instance 15,966 of the 68,250 followers of the newspaper De Tijd are fake. De Tijd also checked the accounts of some Flemish politicians, and as it turns out, about half of their followers are fake.

After reading this article, I was very curious about the followers of my own Twitter accounts. I have three of them:

  1. bruno1970 (6009 followers) is my personal account. I'm a real person. I don't post (re)tweets as often as I used to, but it's still an account I actively use.
  2. Then there is our company account willow9040 (3928 followers), not a real person, but a real company. This is a semi-automated account, in the sense that it automatically retweets my personal tweets as well as tweets from my third account. Once in a while, I tweet something that is specific to Wil-Low, but not very often.
  3. Finally, I also have directmediatips (5385 followers). This is a fully automated account that automatically sends out a Tweet from a database that I created twice a week, and that automatically retweets three tweets from selected third party accounts on a daily basis.

I have a script running in the background of these three accounts to keep them "healthy." I won't explain the "secret sauce" I use, but the larger part of the automation was written to avoid fake accounts. However, up until now, I've never done the test to see if my scripts were efficient.

Let's ask TwitterAudit:

TwitterAudit results: bruno1970, willow9040, directmediatips

As you can see, my personal Twitter account only has 24 fake followers, resulting in an audit score of a whopping 99%. My (semi-)automated account have an even better score: they don't have a single fake fake follower! Granted, there's probably an error margin on those numbers, but I think that they are proof that my script works.

Now let's take a look at the detail provided by TwitterAudit:

TwitterAudit detail: bruno1970, willow9040, directmediatips

TwitterAudit considers my three accounts to be real, which is surprising since one of those three is fully automated. If I interpret the graphics correctly, the quality score of the automated account is even the highest of them all. In general, the percentage of followers with a "real points" score of 4 is above 50% for the three accounts (even 60% for my fully automated account), and the percentage of followers with a "real points" score of 5 is above 25% (30% for my fully-automated account).

Isn't that surprising? I managed to create a fully-automated account of which more than 90% of the followers have a "real points" score of 4 or higher. Stop hiring people in marketing if you want to be successful on Twitter, hire an engineer instead! ;-)


Topics: technology


Buy Bruno's book