By now you have probably seen the “Facebook Fraud” video from Veritasium that blasts Facebook for selling fake likes. You have probably also seen Jon Loomer’s great response on his blog. Vertasium does make some valid points and if you repeat his steps, you will get the same results with fake likes and low engagement. Targeting is the path to ROI, and there is no way around a taking the time to create a well-targeted campaign.
A client of ours is working on an indie film about baseball in the late 1800s. It is a biopic film due out in October, and we are just launching their social media campaign. We decided to retest the theories above and go one step deeper by validating likes one by one, checking if the profiles clicking like were authentic. It was tedious work and produced some interesting insights.
The experiment: are these likes real people?
Targeting is king on Facebook. We tested 2 cohorts, one with US males over 25 interested in vintage baseball players (Cy Young, Shoeless Joe Jackson, etc.). The second test targeted US males over 30 interested in baseball history and comedy movies. We also created Custom and Lookalike audiences for the campaign; however, for this test, we went old school and used only interest and demographic targeting. We paid an average of .25 cents per lead.
Here is the million-dollar question: how to identify a fake profile? This was actually tough to answer. We decided on a few easy-to-see metrics and compared them to averages across Facebook. There was also a “gut feeling” factor; did the profile feel like a click farm bot?
-number of friends (FB total average:130)
-number of page likes (FB average: 70)
-photo content (were there multiple photos of the same person?).
First the big numbers: profiles for about 1 in 4 new likes were deemed fake or at least “questionable.” For example, a person claiming to live in Murry, Utah, had 3,094 friends, liked 13,674 pages, had no pictures of themselves on their profile, and the profile wasn’t in English. A classic click farm profile.
Others fake profiles were tougher to identify: A profile from Milwaukee had 254 friends, a few pictures of the same person, several baseball-related interests, and 3,504 likes. That is a lot of likes; is this a well-designed click bot? Tough to tell; I gave this one the benefit of the doubt.
I was surprised by the results of this test. The targeting was fairly specific and while hardly scientific, 1 in 4 profiles being questionable or clearly fake is something to consider – especially since this campaign was targeted to US users. This could indicate the click farmers are getting more advanced. Innovation in the click farm industry could be creating better hidden bots. In Facebook’s defense, every form of advertising has its share of wasted impressions, and likes appear to be no different. This further reinforces the need for an ROI-driven approach to your campaigns (which shouldn’t be news to anyone).
Have you seen similar results when checking on who likes your page?