I’ve recently finished a 4-day boosted post on Facebook. It was about a class I’ve got coming up, and included an image because posts with images are more likely to be interacted with, and interaction with a post loosens the death-grip of Facebook’s filter. It had a link to more information about the class, which the text of the post lead up to. I put $20 into it–a rather inexpensive promotion, provided it gave decent results.
Here’s what I discovered once it was over.
The First Red Flag
The first thing I noticed, even while it was running, was where the “ad” was being run. It was shown 2,941 times to mobile users, and 27 times to desktop users. However, the trend from the start was that the desktop users were the ones most likely to interact. Of the 56 “engagements,” 11 were desktop users. That makes the “engagement rate” of desktop 40.7%, while mobile was 1.5%. And yet Facebook served the ad to the least likely audience 99.1% of the time.
No points to Facebook for effectiveness on that point.
Waitaminute … this is Seattle, right?
I had targeted the post to people with interest in business (specifically, “Business, Retail, Wholesale, Website, Communication”) within 10 miles of Seattle, to get some of the outlying communities who might still come to the class. That’s roughly going to correspond with King County, WA. The demographics of King County are roughly 71% white, 15% Asian, 9% Hispanic/Latino. About a quarter of King County residents speak a language other than English at home.
I’d expect that the interactions with my post would, very roughly, approximate that demographic…more than half white, for example, and mostly English speakers.
But what I got was very much not that.
I can only see the people who Liked (35 people), Commented (0), or Shared (0) the post. Of those, more than 1/3 had what appeared to be Hispanic/Latino family names, and only one had a name consistent with China, Taiwan, Japan, India, Vietnam, or Cambodia.
That got me thinking, so I looked at profiles, and about half of the people whose posts I could see posted predominantly in non-English languages, with Spanish being by far the most common. Very strange for King County and it’s 9% Latino population if my post is only being shown within 10 miles of Seattle.
So I did a little more digging and looked at where people lived. A third didn’t have any geographic information at all. Just over a quarter came from Seattle and neighboring communities I’d expect to see with the “+10 miles” modifier. And the rest came from other places. Like Ohio, and Florida, and New York. Puerto Rico. Suffice it to say that none of those places, while still United States, are not within 10 miles of Seattle.
Nor are the people from France. Tonga. Ecuador. Mexico. El Salvador.
Well, at least the overabundance of Spanish started making sense. But either lots of people were lying about where they lived, or Facebook was doing a terrible job of actually delivering the audience I was paying for.
Or, I suppose, since they said that 100% of the impressions were US based, maybe the French, Tongan, Ecuadoran, Mexican, and Salvadoran people who saw my post were coincidentally all part of a tour group that was visiting Seattle while my boosted post ran, which they saw on their mobile devices. But I don’t think that’s too probable.
And you’re how old?
While I was looking up where people lived, I noticed another strange pattern. Exactly 2/3 of the profiles had a birth date listed between June and September, 2014. Only 2 people had a birth date prior to 2014. At first I figured this must have been not really their birth but when they joined Facebook. After all, no babies are going to be Liking my posts about business classes. Looking further, though, virtually all of those suspiciously young Facebook users had made posts on Facebook prior to the date of their birth. So unless the most common thing to do is to join Facebook, using your join date as your date of birth, posting and then editing the posts to backdate them to before you joined to look like you’d been there longer (but not editing the birth date), and doing so in Seattle but saying that you’re in Tonga or Mexico, and then going and Liking business ads, I’m not sure how to explain what I was seeing.
Oh, right. The Results.
Really, none of that matters, right? I mean, from a business perspective, what I wanted was for Seattle(ish) people to see that I had a class coming up and clicking the link to learn more about it and, hopefully, to have them sign up. Maybe if there’s a little bit of donked up data and some hinky behavior it doesn’t really matter as long as my business goals actually get met.
Well, here’s what I paid for with my $20. I got 57 interactions total, for a cost-per-interaction of $0.35. A pretty good cost for online marketing…as long as you don’t look further.
That 57 interactions breaks down as follows:
- 31 Post Likes (might be tangentially helpful because friends-of-friends could see that behavior, but not directly a business goal result)
- 24 Photo Clicks (essentially useless–someone looked at the image close up)
- 1 Page Like (Okay, so that one person has a 5% chance of seeing things I post in the future)
- 1 Link Click (my preferred outcome)
In typical online advertising, you count cost-per-click (CPC), with the click resulting in the goal behavior, such as going to your website or lead generation page. Of my 57 interactions, only one of those did that. So if we apply typical online advertising terminology and definitions instead of Facebook’s, I had a CPC of $20 per click. Which really is not that great. If I could be sure that person was really interested in business and lived in Seattle, it might still be worth it…but clearly I can’t make that assumption. For all I know, someone in Puerto Rico has now seen the information about my class in Seattle and will not be attending. $20 wasted.
I could have, I suppose, sent out bulk mail fliers. I could be assured, at the very least, that they would go to Seattle area people and that there were real people at the other end. My Facebook ad supposedly reached 3064 people and 57 interacted, or 1.8%. That’s about the typical response percentage that junk mail gets. Only “touched the piece of email” wouldn’t be part of the count, so the junk mail route might be more effective, if more expensive.
The Silver Lining
So what if my ad was shown to infants, a third of which lived in other countries and didn’t speak my language? At least I have “exposure” right? Well, I suppose that assumes none of those profiles are fake, of course. (One has been deleted already.) Okay, so exposure isn’t even a good silver lining…
The silver lining is, then, is that at least I got a blog post out of it.
And maybe your comments. Have you advertised on Facebook and had actual business come out of it? Or have you done so and found similar weird things? Tell me your experiences. (If I seem to harp on Facebook a lot, it’s because I’ve never known someone personally who showed me data-centered evidence of benefit from Facebook. I’ve heard people making the claims without any evidence, and seen second-hand or third-hand data, but nothing that I’d classify as a “primary source.”)
Of course, if you’d like me to do some analysis of your data to see if your Facebook activities have been effective at reaching your goals, I’d be happy to discuss it with you. Just send me a message and we can get started.
Otherwise, leave your stories below in the comments!
Image credit: http://blogphoto.tv/