Analysis Reveals Bots Bolstering Romney on Twitter

Did Mitt Romney get really popular last weekend, or has something unusual happened to radically boost his followers on Twitter?

The Republican presidential nominee’s Twitter account, which had been gaining 3,000 to 4,000 followers a day, appears to have grown by more than 116,000 followers last Friday and another 24,000 on Saturday.

By this Tuesday morning Romney had about 807,000 followers, an increase from about 670,000 just four days prior. The Romney campaign has publicly stated it had nothing to do with the spike.

So what’s at work here?  This is just the kind of interesting question we at Crimson Hexagon love to grapple with. Using several proprietary analysis techniques and technologies, we went to work in an attempt to get to the bottom of this question. Here’s a high-level look at what we discovered.

We analyzed a sample of almost 150,000 Twitter users who started following Mitt Romney in the last several days (part of the spike in followers). From the users in that population we’ve found the following characteristics:

A typical follower in this group is someone like: https://twitter.com/#!/DelenaRollison_ . She’s described in her profile as “Hopeful website reader.” Note that she also follows an obscure Turkish singer: https://twitter.com/Berksanonline. And, she’s not alone in those characteristics. Based on our sample, 15% of the recent Romney followers also follow this very same Turkish singer.

About 20% of the recent Romney followers also follow this account https://twitter.com/PrettyNekMusic, which belongs to a slightly less obscure American rap artist. We believe this is an odd connection, too odd to be observed in such a high percentage of “real” followers.

The profile descriptions of most of the other followers are quite interesting as well: “Certified user.”, “espresso expert”, “Web site trailblazer.”, “business owner”, “Twitter nerd”, “Unapologetic leader.”  These descriptions were not cherry-picked, just copied and pasted from the screen randomly. They look to be automatically generated from a dictionary by using very simple heuristics (like “adverb adjective” patterns).

Bots Bolstering Romney on Twitter Word Cloud

Words used in recent Romney follower Twitter profile descriptions.

As strange as many of the words above are in the context of a Romney follower user profile, what makes many of them truly nonsensical is the way they are assembled in the actual follower profile descriptions. For example: “award-winning bacon organizer.”

We also compared the people who started following Romney on the 17th and 18th (before the surge) to the ones who started to follow him on the 21st, 22nd and 23rd (after the surge). The former set look like fairly reasonable and conceptually aligned folk, who follow the Heritage Foundation (5.4%), William Shatner (4.9%) and John McCain (5.6%). The latter set, who started to follow Romney in bulk, have the strange distinction of also following a Turkish “sexologist” (28.6%), and another Turkish singer, http://www.youtube.com/watch?v=BrhuZ3rDYs0, in large numbers.  Very odd indeed.

Based on the quantitative measurement of these qualitative factors across many thousands of follower accounts, as well as several other quantitative measures, it appears that the recent surge in Mitt Romney Twitter followers is the result of thousands of auto-generated bots.

True Twitter popularity, no.  Bots taking a liking to the Romney Twitter account, yes.

What do you think about this Romney bot burst? Comment below or tweet @crimsonhexagon with your thoughts. (And, as always, please let us know if you’d like a live online demo of the ForSight platform.)


Written by John Ponte

As Director of Marketing at Crimson Hexagon, John manages the company's global marketing lead generation activities. John resides in Rhode Island, is an avid tennis player, enjoys a wide variety of music and sports.

  • http://twitter.com/Jessicakdurham Jessica Durham

    This is so interesting! Are you always able

  • JP Benedict

    I have the same question as Jessica.  Can CH users perform such analysis as well?

  • Wayne

    Hi JP.  This is a special service we can
    perform through our client services team.

English Japan France Usa Australia Slovakia