Home > Paid Traffic Sources > Facebook & Instagram

---

I'm trying to make sense of this but I just Can't...ANY IDEAS? (11)

---

Facebook has been really weird for me this week. I am running identical ad copies to identical target audiences with identical bids. THe only thing different is the images. Yet, for some reason, they are performing WILDLY differently when it comes to the relative CPC I'm paying for similar if not BETTER CTR's. Here's an example of one of my accounts:

http://awesomescreenshot.com/0993ft78f5

The ads with a $250 budget are one image type. The $50 budgets are 3 other images. You can see that I have MASSIVE CTR's on some of these images (all ads running to a pretty big general audience) but yet, Facebook is charging me MASSIVE amounts more per click despite having over 5x the CTR on some of these ads. The image that is performing the most horribly from a CTR basis is actually my cheapest one. You can see that my CPM's are absolutely outrageous too.

This is the third day in a row I've tried relaunching the ads to see if anything different happens. Every day, these same ads have awesome CTR's and pay a ridiculous amount per click (today was the first day i bumped the budgets on those 250 ads). I've had the same problem in multiple accounts across multiple niches using the same images: Facebook seems to just almost be "punishing" a certain image by giving charging it MASSIVE CPMs and CPCs in spite of the fact that it's actually being clicked on a ton. Facebook wont even give these ads impressions unless I'm bidding over 80 cents in the US, which is pretty damn high for mobile.

Has anybody experienced issues like this? Is there any reason for it?

I almost feel like Facebook is literally just fucking with me rather than banning the accounts.

---

Given the large CPM variation, I'd say it's because the better ads are getting served higher in the news feed and you are getting charged more for it.

Check the average news feed position values. Higher in the news feed (lower position number) >> higher CPM.

Outrageous CPMs tho!

My advice: go to oCPM or CPM and throttle the CPM bid, which should lock down your news feed position better.

Definitely split-test it - can't guarantee CPM/oCPM will work better.

---

I've also had crazy occurrences the last week

On Thursday at 10 am account stopped spending except 5 dollars and hour until 8pm. At 8pm it slammed 1k in 20 minutes before I turned off.

The account spends 4k a day day in and day out and never broke 1500 even w the late slam of traffic

Friday the cpc was double and where normally 70 percent of my clicks turn into website clicks this day it was under 50 percent. Now testimg some things on Saturday and I'm getting clicks, seeing ctr in reports but can't see reach or website clicks. Only the ctr and spend

I'm pretty much bugged out now. I'm not sure if it's the new ad set / campaign bid objective changes or not liking my niche but not banning.

Any help or communique is appreciated.

---

I tried oCPM with the $5 default bid. I got absolutely raped on clicks. Charged me like $2 a click with CPMs ranging from $30-$40

Something really fishy is going on here. It's almost like they're utilizing their face-detection technology to selectively jack up the click costs on images of particular people. I've tried this now with 5-6 images of this particular person and they all behave exactly the same. If I just throw up an image of a random guy taking a selfie though, the bidding seems to behave completely normally.

I've tested across multiple accounts, multiple bid types, multiple images of the same person in multiple countries, targets, etc. I get the same behavior every time. Seems pretty likely to me they're doing something sketchy and screwing with things

---

I think it's pretty clear to me at this point that Facebook just screws with you when they figure out what you're doing rather than outright banning you at first. I've also noticed weird stuff with the relative % of website clicks. That's such a shady metric to me. I don't think there's any data you can pull in the reports that shows what all these "other" clicks are that aren't likes, page views, shares, etc, so it seems like they have free reign to just selectively "tax" you in that area (maybe I'm wrong about this but I've never seen it in the reports).

---

Weird that you got smashed with oCPM - for me at least the bids I put seem to throttle the CPM quite well (I would never pay $40 CPM if I was bidding $5 CPM toward something). But then, I use a PMD, so maybe there was something different about the bids cf. the power editor - though that would be odd.

Regarding your website % clicks... there are a lot of actions that constitute a click and you can get to most of them through the reporting interface.

E.g.
Post link clicks
Post shares
Post likes
Post comments
Inline likes
Inline comments
Inline shares
Page likes

All of these generally should add up to give you X actions, which you paid for.

However, sharing of the ad - if directed to people not in your target audience, can give you actions that you don't pay for... and I think this makes the stats get a bit convoluted, e.g. having 100 paid actions but 110 website clicks!

As for your observations on the particular image... that's really interesting.

So are you saying image X >> retarded CPMs but any other image, all else being equal, gives you more sane CPMs?

---

I've added up all of the metrics that you outlined there many times in the past. Along with website link clicks, those items you listed typically add up to give me the total number of ACTIONS but they do not even come close to adding up to the total number of CLICKS on the ad. Do those items actually sum up to total clicks for you? I would love to see a screenshot of that because that has never been the case for me, especially on mobile. Typically on mobile, only like 60% of my clicks are actually website clicks. Usually for every 100 "actions" i have I'll see like maybe 85 link clicks and 15 shares, likes, etc... So the items you outlined typically only make up like, 5-15% of total "clicks" on my ad, which comes nowhere near filling the 40% gap between clicks and website clicks.

Occasionally, I run ads though where Facebook tells me I have more website clicks than clicks, and my cost per website click is then actually LESS than my CPC. It just makes absolutely no sense and only adds to my belief that they are doing something screwy. There's just no way that some ads get 60% link clicks in one instance but then 110% link clicks in a slightly different situation with the sample size I'm running at.

And as for your last question... yes, I'm saying that if I run like 4 or 5 images with otherwise completely identical ads, that particular image will get INSANE CPM costs while the other images will be be normal costs. The really strange thing about it though is that it is not specific to one image. It is specific to any image that contains this particular person's face! I've created my own custom sort of "mashup" images so that there's no way there's some identical image floating around and that image will still be hit with the high CPMs. The only conclusion I can reach is that they are utilizing face-detection or something along those lines to specifically penalize ads that are running with this particular person (and it's certainly a high-converting person who they might want to penalize the use of).

I have started these ads images with known low CPM images and set them both to bid say, 41 cents for mobile clicks. The safe image will get normal impressions and the other image will get literally zero impressions. In this instance, I slowly incremented the bid up and that image did not receive any impressions until I bid over 70 cents, which was about 3x the recommended max bid at the time and probably close to 4x the number where I would normally see at least some impressions for the ad. The ad then proceeded to have insanely high (almost 10%) CTRs but even after almost $200 in spend, never had it's CPC drop to a level anywhere neard the safe ad, which was running at about 2.5% cpc and getting 28 cent clicks, as opposed to 80 cents or so for the other image.

---

Wow that's some really interesting data.

Facebook using image interpretation or facial recognition in advert pricing would be amazing, and equally scandalous!

Is it a celebrity or just some attractive person who gave high CTR in the past? Have ads with that face ever been disapproved?

I'll have a look at some example data and see if I can make better sense of the website clicks vs actions vs clicks etc. disparities.

There should be some logic to it!

---

If they're using facial recognition software, it'll be defeatable using the normal methods that work against facial recognition.

It might be worth trying a test using an image of this person Photoshopped to make facial recognition much harder, and see what happens then.

Here's some ideas: http://www.thedailysheeple.com/how-t...-robber_112012

---

It's a celebrity, and def one that might have a lot of heat on them. I've run them in the past and they convert extremely well. I tested like I said along side some badly-converting celebs and the bad celebs had terrible CTR performance but much lower click costs. Was getting like .30 EPC's with the lame celebs but like .90 cpc's with the hot celeb, although I couldn't profit at that EPC, where I'd normally be at like 200-300% ROI

Would be interesting in hearing your take on the click vs link click stuff

---

Ahhh... celebrities. I wouldn't be surprised if Facebook had implemented a system to reduce advertisers infringing on their IP. Quite cool!

Unfortunately since you're using a celeb you're not going to get any sympathy, from anyone.

I'd need to have a look at your odd data to say more - since in that one screenshot you posted there were no problems other than the crazy CPMs.

---

Home > Paid Traffic Sources > Facebook & Instagram