March 2014 AdSense Earnings and Observations

This May be my Last Month

[JCKline]

I'm starting the March earnings thread as it may be my last month with Non-Sense.

[avalon37]

First 3 days of the month were very good in terms of CPC and CTR - last three days have declined significantly.

[avalon37]

Contextual relevance of ads today is about as bad I've ever seen it. These are not retargeted ads either. Google has obviously said we are placing the highest paying advertisers over and above relevance. Thanks Google! Way to 9/10 go for the biggest money play, but it's a stupid move and you have years of data on AdSense to prove it. Use your data Google - not what you WANT it to look it.

[jrstark]

Just realized something. While my overall earnings number is insanely high right now, my channel numbers aren't. I'm going to pay attention to those and see if they change as much. Maybe there is just something wrong with the overall calculation.

[wa desert rat]

Had my first clear "click bomb" incident this morning. Coffee in hand I checked Performance for yesterday, saved them as the 2400 report) and then checked Performance to today. Much higher earnings than I expected and once I got over the shock and looked closer I could see that there were 28 clicks for one channel that typical has 1 or 2 by the time I get on line. Saved that and then refreshed and another click appeared.

So I got onto my other analysis tool which gives me a detailed report of how many times a user has visited, when, how many "actions" and exactly which actions for every visit (and records MAC address as well as IP). I was looking for a user who would have several visits for short periods of time but lots of actions (and mostly windows). Nothing jumped out at me but I have no idea if clicks that appear in the PR are delayed or not. So I went back a few hours and still nada.

Filed a report with Adsense. Watched closely but nothing more.

[webcentric]

I really beginning to wonder if, in some cases, "click bombing" isn't something being created by Google itself.

@wa desert rat -- You said one "channel." Does this equate to one "ad unit" in your case? In other words, did the bomb go off on a single ad unit or were multiple units involved?

Reason I'm asking is that I can see a bad loop in G's code going off in certain circumstances that could just be recording the same click, over and over again with no bot involved whatsoever. I'm not saying it isn't a bot but this isn't the first time people have been looking for unusual traffic activity that coincides with a "click bomb" and found nothing unusual. That leads me to wonder if the issue is in Google's click recording or reporting code itself.

[jrstark]

I've got one channel going nuts, not normally a big earner. This is one ad unit on multiple pages. It's either something on Google's side, or someone has figured out a way to target specific code.

[wa desert rat]

@Webcentric: it was one ad. I'm not a big player and have two websites with two ads on each. The blog gets little traffic and not much in the way of clicks. The forum gets a lot more traffic but still not a lot of clicks (lots of ad blockers).

This morning the forum had a ton of clicks on one channel (ad)which I reported... then started to get clicks on the other channel (ad) which I also reported. Google managed to carve away the majority of those bad clicks but I think there are still some suspicious clicks lurking on the Performance Report. But they were remarkably quick getting them taken away after I reported them.

Nothing in my logs to indicate suspicious activity. I'm not sure we know what "suspicious activity" looks like, frankly.

[webcentric]

@jrstark

Next question(s). Is the unit...

Synchronous or Asynchronous?
Text Only, Text and Display or Display Only?
What size is the unit?
Also, fixed size or responsive?

We're looking for predictable patterns here or specific types of troublesome units. The possibility is that the problem is related to a specific ad implementation but it could also be a system-wide problem. Just trying to narrow things down a bit. Thanks.

[wa desert rat]

That leads me to wonder if the issue is in Google's click recording or reporting code itself.

That doesn't bear thinking about. If Google itself is somehow generating clicks, I'm glad I'm not an Adwords member!

[webcentric]

@wa desert rat -- I know the ad size you're referring to then. I'll leave it to you to disclose that or not but I'm beginning to wonder if that particular unit size is a general target. It's why I asked in my previous post to jrstark about ad size, type etc.

[webcentric]

If Google itself is somehow generating clicks, I'm glad I'm not an Adwords member!

It's a tough thought to have for sure. But, if I was Google and my system was somehow getting the jitters from time to time and I couldn't get it fixed quickly, the temporary solution would be to put something in place to undo the problem that my jittery software was creating. Be it bots or G's own software that's causing the problem, the temporary solution would look somewhat similar to what we're seeing now.

[netmeg]

if that particular unit size is a general target

I experience this on four different sizes; 728x90, 468x60, 300x250 and 160x600. Over many sites, and many many impressions.

[wa desert rat]

If *I* were going to engage in click-fraud, the last thing I'd do would be click-bombing because it attracts attention immediately.

What I'd do would be to direct a clickbot to a website, have the bot "read" a few pages, then move to the page with the specific target ad, click... and leave for another target.

Then I would direct another clickbot to do the same thing some time later. Rinse and repeat. Your base are belong to me!

This really looked like the pattern on my site until this morning. Most of the clicks I have seen taken back seem to be engaging in a "burst" of clicks (2 or more). If CPC is low then that looks like the highest criteria Adsense is using to determine invalid clicks.

I think the "click-bombs" are an entirely different species. Not sophisticated at all. What surprises me is that Google doesn't seem to respond to them very quickly. Unless, of course, it's some aberration in Google's own code that's generating it.

Makes me suspect that we're not getting Performance Reports in as timely a manner as we might think we are.

According to the Adsense TOS we are still responsible for click-fraud on our sites. I can understand someone losing an acount for not reporting click-bombs but the new bots are impossible for us to really detect.

[webcentric]

@netmeg and all -- Interesting that three of those four sizes are "Recommended" ad sizes per Google and the fourth (468 x 60) is probably the most used size for tablet display. Also, in line with wa desert rat's observations that most of those sizes aren't practical on a mobile site or are at least very obtrusive in that environment, I'm guessing that most of this is happening on the desktop or tablet.

These are also fixed sizes. I use responsive ads almost exclusively (smart sizing) and don't see the degree of this problem that some do. I ask if this could be a problem with very particular units?

I had to just check this before making this statement because I don't use fixed ad sizes at all. So, I just generated a 728x90 banner to see the actual ad code to make sure. Sure enough, the following is in the code itself.

style="display:inline-block;width:728px;height:90px"

If bots are involved, this would be a key way of identifying and targeting a particular ad unit size. A worthy experiment (if you are able) would be to replace one of these fixed units with a responsive unit. Might take a bit of container div magic to get it to display in exactly the same size as the fixed unit but it can be done. Pick a troublesome unit and give it a try. What have you got to lose?

p.s. You can force smart-sized units to a particular orientation by changing the "data-ad-format" parameter value in the ad code like this...

Original

data-ad-format="auto"

For a square unit

data-ad-format="rectangle"

For a horizontal unit

data-ad-format="horizontal"

for a vertical unit

data-ad-format="vertical"

Add some container div dimensions and you're able to constrain the unit to the size and orientation you're looking for.

[ember]

The yo-yoing has stopped, but now I am seeing very high CTRs on the 336x280 and the 728x90. This is across several sites. The ads are synchronous, text and display and fixed size. Two days ago the CTRs were way below normal, and today they are way above normal.

I'd really like Google to get a handle on this.

[webcentric]

I'd really like Google to get a handle on this.

Red bar is showing. Maybe a new wave of "fixes" is in the works. One can only hope. ;)

[netmeg]

My 728x90 is actually a responsive unit. I just assume it's displaying at the 728x90 because the type (with all the extra clicks) is listed as desktop.

70 to 80% of my traffic is mobile. I only get this activity on desktop.

[webcentric]

netmeg, are you using smart sizing or media query sizing? Not sure it matters but just curious. Media queries contain dimensions, smart-sized ads don't.

Generally I'm also wondering how ad position fits into this equation if at all e.g. above the fold vs below the fold. Some have reported the problem(s) with both so it's less of a suspicion.

Also, I'm still wondering if there's a text vs display ad pattern. A unit can be set to deliver both but the performance reports can identify the type of ad clicked so wondering if there are any patterns there.

Added: This is really starting to look like a desktop-specific issue.

[matrix_jan]

I see a lot of you got click-bombed at around 4-6th of March. Perhaps they integrated Spider.io into their system and something went wrong (right?).

[ember]

As I look deeper, it is desktop and it is the 336x280 and the 728x90 and it is the synchronous ads.

Just talked with chat support. She said Google updated its Sabotage information page this morning, which seems to indicate that that is what is going on.

"If we determine that this recent activity was not the source of genuine user interest, you may notice a larger difference than normal between your estimated and finalized earnings for this month. Any earnings derived from invalid activity are refunded to advertisers. Rest assured that you will be paid for all valid clicks and impressions."

[netmeg]

I have both smart sizing and media query sizing on different sites. Affected ad placements above and below the fold. Almost exactly evenly split between text and image, according to AdSense reports. I tried removing all the async ads and putting the old synchronous code in, and it made no difference. All the steps you mention, I've been through months ago. The "click bombing" phenomenon is different than the increased invalid click activity (and is treated differently by AdSense). And I have several other bot things going on that don't even have anything to do with AdSense. I've had experts look at the log files and traffic, and paid for services like cloud flare and others. I'm not exactly new at this.

[wa desert rat]

@Webcentric... remember that the motive for click-fraud is to make money. If it's not the publisher making money then it's a middle-man (someone who is paid by advertisers to place ads). With the new sophisticated click-fraud bots that are controlled by C&C server it's not the size or type of ads that are targeted but the ads that can make the perpetrator more money.

I'm wondering if someone hasn't discovered a way to make money even with invalidated clicks. So it doesn't matter that Google shaves them off; the middle-man can still make money depending upon where the refunds go.

Edit: If we assume two types of bots... the old fashioned brute-force bots (click-bombs) and the new sophisticated C&C bots then it's also possible that there are two type of targets. And at that point size and position might come into play.

I'm also wondering if a site could be click-bombed as a prelude to more sophisticated attacks; sort of testing the waters to see who is alert, perhaps.

Just FYI: until this morning click-bomb the most targeted ad position was the bottom banner. Although both top and bottom banners got hit, the bottom was hit more frequently; especially with high-bid ads.

[edited by: wa_desert_rat at 7:56 pm (utc) on Mar 7, 2014]

[ember]

Pageviews are static now and the CTRs are coming down. Beginning to look like normal, although I am not sure what normal is now.

[webcentric]

@netmeg -- Didn't mean to imply that you aren't paying attention to the details as I'm fairly certain you have as good a grasp on this as anyone here if having a grasp (for any of us) is the right way to put it. The comments of many though imply that the problem(s) are perceived from the bottom line (e.g. clicks/revenue come and then go). My questions are simply trying to get more specific about the issue. There do appear to be some limiting factors involved with some of this (specifically the desktop environment for example) and I've just been wondering if it's even more specific than that. Can a simple change in usage make the problem go away for some?

Granted, it's not practical to stop serving your pages to desktop clients but would switching to a different ad unit help? 728 x 90 is probably the most commonly used ad size there is. It also, for the most part these days. involves a single ad in the space (not always but I think predominately). So now I'm wondering if panel-type ads are more or less vulnerable to either of these issues (click bombing or invalid clicks)? That's one I'm not sure about how to discern but I do wonder. Perhaps using unit sizes that do or don't accommodate panel ads well could answer that question.

I'm enjoying the insights in this thread this month and the details people are providing are creating a better picture of the issues. And you're right, there are at least two distinct issues on the table at the moment. Not sure if any of this will help anyone or not but the potential increases as the discussion expands from lost clicks to lost clicks in specific situations (for example).

[webcentric]

@wa desert rat -- While it's not a fixed rule, larger ad units do tend to attract some of the highest paying bids. Of course, as I say that I see a six dollar click on a tiny unit in my left sidebar so who knows? Again, try changing a standard unit like a leaderboard to some less popular size for awhile and see if the problem persists. Or try using a link unit for awhile and see if that has the same issues. This feels related to ad types somehow but it's just a gut feeling at this point.

[jrstark]

I have a channel that uses a 160x600 ad (text/image) on a few pages for an annual summer event. That channel is now showing one pageview with 5 clicks for today. It's below the fold on the far right.

[ember]

Not to sound like a complete newbie, but what is the difference between click bombing and invalid clicks? Isn't click bombing the same as invalid clicks but on a large scale?

[netmeg]

The invalid clicks tend to be spread out over a bunch of different ad units over the course of a day - maybe twenty or thirty of them, and they're usually gone out of the account within an hour (or at least by the end of the day)

The clickbomb is a mega dump that lands all at once - yesterday I had one that was 925 clicks on two ad units - and, in my case at least, doesn't get taken out till the end of the month.

YMMV.

[HowYesNo]

I use custom size ads like 800x300 but most of the high paying clicks (10+ cents) from US are removed.

I have unusually high ctr today on one 728x90 ad above fold and the clicks are just 1c lol.

Also on another 728 ad at the top I'm receiving clicks from US but they are all removed, when i check traffic source, it's this site bserpa.com (HACK3D BY ABS3NT) so probably they run bot there which is clicking on the ads.

[wa desert rat]

My mobile ad theory may be wrong. Two clicks with CPC of $0.63 on the mobile ad but one of them was taken back 30 minutes later. Oddly enough, the CPC stayed at $0.63 for the single click that was left and 90 minutes later it's still there. All the click-bomb clicks are now gone and we're back to the yo-yo pattern. I never thought I'd be happy to see it. LOL