Funky Facebook Insights: Facebook Suspicious Account Removals

Portent Team Sep 25 2013

Hi. My name is Sara, and I’m a Facebook Insights junkie.

I was getting my latest fix yesterday, while trying to figure out an explanation for a client’s relatively high rate of Facebook page unlikes that we detected in August. Yes, we use tools for this stuff; but we also still dive deep into the buried tabs in the reports you can export from Facebook, since tools and Facebook’s shiny new Insights dashboards can’t always explain exactly what’s happening. And when you’re really lucky, Facebook’s shiny new Insights reports don’t match their detailed insight exports, and you get to scratch your head, look at your spreadsheets sideways with an eyebrow raised, and go, huh.

This just doesn’t add up.

So this post has more questions than answers, but I’d love to get a conversation going among other like-minded Facebook Insights junkies, so here’s a peek behind the curtain at what I’m seeing. And if you read this, and nod along, and get excited to chime in, then please do — I’d like to make a section in my little black book for fellow Facebook Insights junkies and add you to it.

Let the Head Scratching Commence

One of our clients saw a precipitous drop in Facebook fans during August, and have seen some smaller drops during September so far. We counted a loss of over 800 fans on a total following of less than 7,000 fans during August, for example.

And here’s where the head scratching starts.

According to their Facebook Insights Exported data, they’ve only had a handful of unlikes: one or two here and there, at the Page level and the Post level.

But when you pull up the fancy new web-based Facebook Insights reports for Total Page Likes, you see the problem our manual tracking picked up clearly:

August 23rd wasn't a happy day in Facebook-land.

August 23rd wasn't a happy day in Facebook-land.

That report shows a precipitous drop in fans: no question.  And digging deeper, the Net Likes: What Changed graph shows without question: the few days starting August 18th were not great for this client. And then, WHOA. Holy CRAP. August 23rd? That day SUCKED. To the tune of -853 net likes.

We in the biz call this a bad week.

We in the biz call this a bad week.

Facebook has added a nice detail to this interactive report — you can click to get a detail about the Likes and Unlikes on a given day.  We started working chronologically, looking at the detail on each day that there was a net loss in fans, and made some interesting findings.  In the days leading up to Precipitous Drop Day, the client saw a higher than usual number of unlikes as reported by the web-interface to Insights:

  • 50 unlikes on 8/18
  • 61 unlikes on 8/19
  • 44 unlikes on 8/20

even though the detailed Page Level Insights download shows significantly fewer Page Unlikes for those days (like, 1/10 or less of what Facebook’s graph is showing).

There’s no paid reach at the page or post level, so (from what we can see) advertising didn’t play a part.  And there were no new posts made during those days (although the client had a high-post day on 8/17,  with four updates going live (more on that in a minute).

Then, things get even more interesting.

On 8/21, the client saw 13 unlikes AND 17 “suspicious account removals.”  On 8/22 it was back to just unlikes (24 of them).  And then August 23rd happened.

Here’s the detail on Unlike Sources from August 23rd:

That's a whole lot of suspicious accounts removed.

That's a whole lot of suspicious accounts removed.

829 accounts removed in one day.  That’s a whole lot of suspicious accounts, scooped up and cleaned out in one day.

We’re used to seeing this on Twitter:  periodically, Twitter purges inactive accounts and / or accounts that are otherwise not in compliance with their terms of use.  Facebook has done clean-ups in the past, too, but we haven’t seen an example that affected this large of a percentage of a page following before.  And, a suspicious account clean-up wouldn’t explain the discrepancy between Facebook’s detailed negative feedback reports and the numbers of Page Unlikes shown in the Insights interface.

And remember that high-post-volume day I mentioned?  August 17th?

Well, that’s one of the other head scratchers.  One of the posts that went live that day experienced higher than average viral impressions according to the detailed Post Level Insights we downloaded.  Cool, right?  That’s what we’re shooting for, yes?

Well, yes.  But the post only received 4 comments and 7 likes.  It received zero shares.

The same thing happened with posts on August 13th and 23rd: significantly higher than average viral impressions, on posts that received zero shares. We’d typically expect to see viral impressions go up on posts that received shares, and to be lower on posts that receive no shares.

Did the viral impressions have something to do with the uptick in unlikes and suspicious account removals?  Or did the unlikes and suspicious account removals somehow affect viral impressions?  Hard (or impossible) to say.  The high viral impression post on the 13th came first (before the unlikes and and suspicious account removals) but so far I can’t make a connection between that post and what transpired next.

Where does that leave us?

This client saw higher than average viral impressions from organic (not paid) reach, right before they saw a significant drop in page fans. The drop consisted of a small number of reported unlikes according to the Page and Post level detail Insights export; but a substantial number of unlikes reported according to the Unlike Sources Insights graph; and a breathtaking number of suspicious account removals.

We’ve reviewed the content posted during the month and there are definitely some tweaks and recommendations we’d make around creating more user-friendly content; but usually when we see unlikes we can look at the content and BAM. We know what went wrong. Content that wasn’t well-targeted to the audience. Content that was either unintentionally or intentionally offensive. Content that was too heavy on the overt selling. But this time, there was none of that — just relatively average content that could potentially have been crafted in a more user-friendly way, but wasn’t the kind of content we’d expect to see substantial rates of negative feedback on.

So Who’s On First?

We have more questions than answers this time around, so we’d love to hear from others who are seeing interesting things happening with their Facebook page followings.

Have you seen a drop in fans? Are you seeing suspicious account removals from your business page? Are you seeing a discrepancy between the negative feedback reported in Post-Level and Page-Level Insights reports exported by Facebook, and the graphical information displayed via the Insights dashboard? Are you seeing higher than average viral impressions on posts before and/or during any recent “fan loss” episodes? If so, I’d love to hear about it. Are these episodes more common during or after paid campaigns? Are they occurring for pages that boost posts? Chime in, and if there are other folks interested in digging deep into this stuff, I’ll put together a community for us to take these types of conversations offline.

Have you noticed anything funky in Facebook Insights land lately?


  1. Lara


    Really interesting post! I’ve never considered Facebook to have such an analytic approach to its data. Thanks!

  2. Johnna


    I found this article after a google search. Earlier today I had nearly 1000 likes on my Facebook page. I just looked at the page again and I’m down to 960. The graphs don’t show a significant number of unlikes, and the graph for total page likes doesn’t show a significant drop in page likes either. 38 likes just disappeared out of thin air!

Comments are closed.