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.

[shydevil]

Normally, I just read these forums, trying to learn more about Adsense and how to improve it, but this topic was really helpful.

I am getting less from ads each month. My blog used to earn around 200 euros a month, then it went to 142 in January. I made all changes I could, following Adsense suggestions. Revenue optimization went from 2 stars to 5, yet there is less and less money coming in.

I have improved SEO and boosted visits fro 12k a month to over 20k, there is no sign of invalid clicks, but revenue is going down all the time.

March is the worst, so far only 18 euros and I have started looking for alternatives.

[kireb]

Adsense reporting is now completely out of control. I think that at least 80% of all clicks are being removed within 20 minutes. The stats don't make any sense at all and RPM is record breaking low these days. This is one roller coaster ride and doesn't improve but get worse by the day!

[kireb]

Adsense reporting is now completely out of control. I think that at least 80% of all clicks are being removed within 20 minutes. The stats don't make any sense at all and RPM is record breaking low these days. This is one roller coaster ride and doesn't improve but get worse by the day!

[netmeg]

The interesting thing about all this foofaraw is that bottom line, I'm still up every month, after all the ups and downs.

[webcentric]

The interesting thing about all this foofaraw is that bottom line, I'm still up every month, after all the ups and downs.

Same here as far as an upward trend is concerned. Just not seeing the rollercoaster some are seeing. One thing is I'm using a new Adsense account (had to set up a new one for business reorganization reasons). So while my main site is fairly old, the publisher id on it is fairly new. Makes me wonder if there are bot involved in any of this if they're targeting publisher id's. That's my theoretical question for the morning.

[netmeg]

Makes me wonder if there are bot involved in any of this if they're targeting publisher id's.

I dunno. Only about 4% of my sites are affected (all using the same pub-id), but they're the highest trafficked ones.

[webcentric]

Does anyone run a website without G analytics on it, and if so, have you been affected?

Yes, I removed Google Analytics from my site about six months ago and see very little of what others are reporting (just a very occasional take-back which I chalk up to things like double clicks mainly -- and no click bombs). This is a line of question that's been in my head as well. Could there be some sort of conflict between the Analytics script and the Adsense script causing this? You can expand that question to cover synchronous vs asynchronous versions of those scripts in various combinations.


Per the Google link on invalid clicks above.

The vast majority of all invalid clicks on AdWords ads are caught by our online filters. These filters are constantly being updated and react to a wide variety of traffic patterns and indications of click fraud attacks.

This statement from Google may very well explain the "red bar" situation and point to why it's possible that Google is introducing bugs into its own software.

But if you have a niche website for which there is demand for information and little supply, why should you continually change things?

If your revenue is plummeting but traffic is staying the same or growing and your ad network changes the format of all your ads (like by removing hyperlink underlines from the titles) then it's probably time to find some way to get more actionability back into your ads (such as by using a more pronounced color scheme). Hey if nothing has changed, including click volume and revenue, then your argument holds but in a state of sustained decline, very seldom is doing nothing the best strategy. In fact, maintaining the status quo in times where things are going good is called "resting on your laurels" and that's not a good strategy for preparing for future change either (something that's inevitable per the axiom that says "change is the only constant").

You're right, bickering is not the best way to get to the root of a matter and if you compare this monthly thread to previous months, you'll see a very distinct difference in the level of sharing and detail presented. Whatever the impetus, the shared sense of a common foe, the feeling that we're all in this boat together in some way, whatever it is, it's refreshing.

Are the answers here and/or will we find them here? Don't know but the level of the dialog is at least encouraging. Thanks to everyone who's making that happen.

[wa desert rat]

Does anyone run a website without G analytics on it, and if so, have you been affected?

Webcentrics does and he is reporting less of this yo-yo, take-back-clicks than many of us. I have two analytics systems running: Analytics and Piwik. But I always thought that Analytics simply analyses the data for the Analytics reports and did not affect the Performance Reports.

I could be wrong. Does anyone who is being hit by whatever this is want to disable Analytics on their site and report back what happens?

[wa desert rat]

Jaideemaak

When this happens, one or both clicks are taken back. He spent so little time on the first site that the Adwords goal couldn't have been successful so the click doesn't count. It's not the publisher's fault, but the publisher is punished.

Almost any "burst" of clicks will be taken back. Especially by one user... and most especially if it's one user who has arrived multiple times on your site and made one or two clicks and left.

Nomis5

Regarding "lost" clicks. Is there a growing consensus that it's G which is under attack and not individual publishers / websites? It's the interpretation I put on the past few months but with no solid evidence.

There are several theories:

1) Google's Adsense reporting system or its Analytics system (or both) are out of control;

2) It is botnets under control of middlemen "ad agencies" whose goal is to executetargeted clicks on specific ads in a manner as close to human as possible and thus evade the filters Google has set up to eliminate them but that Google eventually does eliminate most of them;

3) Click-bomb attacks may actually be attacks on competitive web sites; or,

4) Click-bomb attacks are just older click-bots still loose on the Interwebz.

Any one of these - or any combination of these - may be true. Or not.

Read this: [icsi.berkeley.edu...]

And this: [infowar-monitor.net...]

[wa desert rat]

webcentric

Makes me wonder if there are bot involved in any of this if they're targeting publisher id's. That's my theoretical question for the morning.

I don't know how targeted ads work from the advertiser's side of things. Do they use the publisher ID when they target sites for specific ad placements or do they just use the URL?

Maybe netmeg can answer this since she's on both sides of this equation.

[netmeg]

No, advertisers don't target by publisher ID, but I think webcentric means he thought the bots were targeting by publisher ID, not the advertisers.

The advertisers can target by theme (determined by a handful of broad keywords, and the name of the ad group or campaign), by category (Google has a list) or by specific site and/or URL.

[webcentric]

thought the bots were targeting by publisher ID

Yes, that was the idea. Id's are part of your page's source code so easily readable.

[nomis5]

I could be wrong. Does anyone who is being hit by whatever this is want to disable Analytics on their site and report back what happens?

I'd be really interested to know if someone did this. I've no problem doing it on one of my sites but I've already disabled / replaced the worst G ad unit and although others are affected it's not so noticeable.

The problem as well is that the clicks are added and taken away so quickly that it's not possible to be sure what the result of any test actually is. To tell the truth I've given up believing in the earnings figures until I'm paid.

Nowadays I'm more addicted to watching the Analytics real time reporting! It'a already pinpointed an embarrassingly stupid coding error on one of my sites. And, as far as I can judge, the figures on real time appear to be correct. They don't include earnings or click of course but they do show when users enter, give some idea of flow etc. etc.

Also, we appear to have two threads about almost the same subject - this one and another on click-bombing. They simply have to be related so make sure to read both rather than just get side-tracked into one of them as I was.

[nomis5]

Just talking about click bombing and invalid clicks with with my totally computer illiterate partner and she asked an interesting question.

As I said earlier I have removed the worst (for invalid clicks) ad unit (for size read long and slim) ad unit from my pages and replaced it with Criteo / affiliate ads. The question she asked was how was the ad block doing on Criteo (now referred to as C)? The answer is that it's now performing exactly as I would expect on C. That I find interesting.

C are one of the bigger players in display ads and especially so as far as bidding for ad slots is concerned. How come they can can display ads on my site with no signs of click-bombing /invalid clicks but G can't?

Others will know know far better than me but C bid against G and others for slots real time so the thought is that they should all be affected. But from my perspective they aren't.

The sample size is admittedly small for a comparison because tall and slim ads just don't work well for the site in question, but it is into the low hundreds of ad impressions per day.

Any thoughts or am I just hopelessly misguided?

[wa desert rat]

netmeg

The advertisers can target by theme (determined by a handful of broad keywords, and the name of the ad group or campaign), by category (Google has a list) or by specific site and/or URL.

That fits the C&C botnet scenario quite nicely. With the numbers of Internet advertising agencies out there looking for business it's certainly likely that some percentage of them might not be completely legitimate. One click-fraud scenario includes ad agents who are not legitimate.

An advertiser contracts with an agent to place ads for him/her. Payment to the agent is a commissiion based upon results in the form of some variety of click-throughs, goals, page impressions, or conversions.

The agent then uses the broad goals of the advertiser to determine specific ad placements down to the specific site. So a company selling tents (for instance) would want their ads placed on web sites where people who buy tents might be expected to visit. Like www.I_like_to_camp_a_whole_lot.com.

But what if the agent decides that he/she can make more money by directing click-fraud bots to those specific ads on those specific sites? So one of the sites the bots are sent to would be www.I_like_to_camp_a_whole_lot.com.

At that point the only issue would be evading detection.

In this scenario the bots and the advertiser (the agent placing the ads) are indistinguishable. The same entity that targeted that site for ad placement also targeted that site for click-fraud.

The bots click the ads the advertiser sends them to click, the advertisers budget is debited and the agent makes a little more money.

This is not the way all click-fraud bots act, of course. But if Google has figured out a way to detect these bots and their clicks it could, at least partly, explain some things. Including:

1) Constant shaving of 2 or 3 clicks at a time throughout the day;

2) Why Webcentric's site - which has ads that are not targetable - is not as seriously affected: and,

3) Why mobile ads seem not to be so heavily affected.

This is all speculation, of course, but I think it hangs together. Any critique?

[webcentric]

@wa desert rat -- seems possible. You know to target an ad the bot has to be able to read the link target for the ad or some other ad feature like the title which means it has to be getting the info from G's internal script I think. Would mean a bot capable of targeting specific ads is getting it's info from something other than what's visible in the page source (your Adsense code). Rather it would be deriving it's insight from what the Adsense code actually generates (and I'm not claiming to understand how that works). I know I can get link targets in many cases by hovering over the ad but that's the extent of my knowledge on that subject. Another tool that does something along those lines is the Adsense Publisher's Toolbar or whatever they're calling it these days. I don't know where this thought is headed but thought I'd throw it out there.

[wa desert rat]

The bots that do this are called 7cy bots (after a part of the domain name that the bots use for Command and Control. The bots are all hosted on exploited windows boxes. The C&C servers tell them when and where to go and when to execute a click. They also click on unrelated ads in order to confuse detection algorithms.

If a new player in the 7cy scheme entered the arena around the first of the year it would also explain the sudden uptick of problems

WDR

[wa desert rat]

Here is a very interesting account of a clickfraud system that used a DNS changer to misdirect clicks from legitimate sites to fraud sites. Earned the operators $14 million in 4 years. The FBI mounted an operation against them and took them down:

[cs.indiana.edu...]

EDIT: One of the interesting aspects of this particular clickfraud scheme was that the ad size played a significant role in whether the fraudulent ad was served up to the victim or whether the original (real) ad was. If the ad size was not one that was supported by the ad network contracted by the perpetrators, then the real ad was served to the victim.

WDR

[webcentric]

OK, having read the initial study posted by wa desert rat (albeit on a phone with numerous interruptions) and from subsequent bits of information coming out here, something seems to have just clicked into place in my slowly moving imagination.

First, we were on the subject of impacted ad sizes and these issues do seem to be more prevalent with the most popular ad sizes (most of which are ad sizes Google itself recommends).

Now, let's say a botnet is in collusion with an agent set up to offer only those size ads or at least specializing in those size ads (maybe running specials on 728 x 90's for example). This would mean that the bot could eliminate a lot of noise in its search by specifically looking for a particular ad size. That piece of information is in the Adsense script itself (at least on fixed-size units). So, if I'm a bot looking for a particular ad and can filter out a bunch of possibilities, just by size, then my job just got easier.

Now, we've also had discussion about fixed sized units vs responsive units. While responsive size units don't display an ad size, they are very capable of displaying an ad size that the bot is targeting so the bot could still theoretically be targeting a 728 x 90 ad and find it in a responsive unit.

So, I wonder if people who have removed an affected unit and seen the problem migrate to another unit are seeing the migration between similar sized units or at least to units of a size in the more popular ad size set which also includes 160 x 600 skyscrapers and a couple others mentioned by netmeg earlier.

Here'a a radical test. Rather than remove an impacted leaderboard. Replace it temporarily with some smaller, less popular size unit (like a 250 x 250 or 200 x 200 for example. See if the problem migrates away from new unit to a more popular-sized unit on the page or latches on to the new unit. Seems logical to me that if an agent is targeting ads it sold, it would know what size they are and it would be most easy to hide my activities in the largest pool possible, e.g. the pool of leaderboards and skyscrapers, etc. that a great many publishers favor, the sizes Google says will make you the most money.

[wa desert rat]

Webcentric

Here'a a radical test.

I agree that it's the sort of thing that could give us all a better clue as to what is going on. But it is not without ramifications. Let's say you replace a targeted ad of a common size with an untargeted ad of an oddball size; because you're first ad is no longer on you will miss out on some targeted ads (and clicks).

It would be nice if several of us tried this at a time. I'm going to have to mine out time to re-learn how to do it. Normally I am in servers in a terminal emulator all day long. But it's been so long since I edited this specific html code that I'll have to take the time to make sure I know how to do it properly.

Maybe I should learn how people who don't know how to use emacs and vi do it.

Nahhhhh.....

WDR

[webcentric]

because you're first ad is no longer on you will miss out on some targeted ads (and clicks)

Targeted ads, perhaps. Clicks, perhaps not. Some of my best ad slots are oddball sized units. The issue that would concern me more would be how to make them work on the page visually without getting into a whole site redesign (just to run a test). Replacing a rectangle banner at the top of a page with a square could call for some rearrangement and for sites with responsive design, a fixed unit like that can be a pain in the neck. Trust me, I've been there. Anyway, if it's a bot looking for particular sized ads, the change could be telling, if it's a bug in G's code, could also be a bug that's related to certain sized units so could also be telling. Then again, it wouldn't be the first wild goose chase I've initiated in my day either. Process of elimination is the goal but it might just eliminate nothing. My best producing unit is a 250 x 250 and I use leaderboards sparingly (and never above the fold). Does that mean anything? Only if someone else sees a similarity in what they're doing and can relate similar results. Otherwise it's all guesswork on this end (or dumb luck). ;)

[jbayabas]

Earning is almost half than it used to be. Ive never imagined this would happen to me. So depressing.

[extremist]

same here.
earning is almost half than it used to be since March.
and Google constantly reduces the number of clicks and my estimated earnings since the beginning of March.

[Mentat]

For the first time in many days, my eCPM is rising quickily, but my impression are low, so not to much revenue, but is a start...

[RedBar]

Yet another extremely poor performance from the USA yesterday with both clicks and EPC at less than half of US average resulting in the UK out-earning the US by 397% ... That's another new record for me!

[kireb]

Help my ads have disappeared! My pages load normally but the ads are 'hanging', I can only see the small triangle. What's going on G?

[jbayabas]

Kireb, clear your cache/cookies.

[webcentric]

OK, I just has another "duh" moment.

I know we discussed bouncing stats and where people are seeing them e.g. GA linked to Adsense interface vs the actual Adsense interface. What wasn't discussed though is what version of the Adsense interface you're using if that's where you're seeing the bouncing.

Someone just mentioned that they're using the old Adsense interface in another thread and it reminded me that I'm using the newer version of the interface which doesn't show things like CTR and actual clicks on the main screen. I'm just wondering if the newer interface is better at hiding the bouncing stats so I'm not seeing them the way I would if I was using the older interface? I don't use GA anymore so there's another place I can't possibly be seeing this issue,
obviously.

Question here is, are you seeing the clicks come and go in the old Adsense interface or the new one? I could see G putting more effort into the new interface than the old as my guess is that the old one will eventually be phased out. This question leaves anything to do with GA out of the mix for the moment.

[wa desert rat]

I'm seeing them in a version of Performance Reports, under "Quick Reprots", "Custom Channels Today" that show "Custom Channel", "Product", "Page Views", "Clicks", "CTR", "CPC", "Page RPM", and "Estimated Earnings".

I really don't use GA all that much anymore.

WDR

[jrstark]

I'm seeing the bouncing in the new version, my earnings are bouncing. Since I realized it, I am seeing the channel where it is happening on Top Channels. If I go to View reports I can see the clicks.