As a follow up to my previous article on the discrepancy between Google and MyBlogLog in regards to click counting I present the following information:
| Month | MyBlogLog Click Data | Google Click Data | Over/Under |
| March | 492 | 375 | +117/131% |
| April | 637 | 637 | 0/100% |
| May | 756 | 692 | +64/110% |
| June | 1302 | 1349 | -47/97% |
| July | 1041 | 1132 | -91/92% |
| August | 632 | 294 | +338/215% |
| September | 288 | 40 | +248/720% |
I'm not sure why March numbers are off by so much but most of the time my numbers are within 10% of each other. Then something happened in August.
Look at September! If MBL click data is correct and Google is discounting clicks it considers to be fraudulent then why have I not yet recieved one of those "you have been banned from AdSense" emails? With a click-fraud rate up near 620% I should be banned from AdSense!
I can understand Google blocking it's "10% of all clicks are click-fraud" clicks but this is far worse than that. To make matters even more interesting I am finding that the discounted clicks look to be from articles that have gained attention on social media sites.
Is Google looking at how a reader came to my site and discounting a click if they came from digg, netscape, stumbleupon, etc? It looks that way to me.
I have been in contact with some of my fellow bloggers who also use MBL on their sites and some of them are also noticing big differences in the numbers being reported.
I realize that in August I split off the Pokemon content over to PokeFarm and that may account for some of the drop off in clicks but if that were the case why did MBL track so many more clicks than Google did during that same month?
In May the Pokemon content was very minor and August should have reverted to the same click levels as May. I base this on the fact that overall traffic levels and patterns returned to my pre-Pokemon levels.
Like I said before, the tracking for PokeFarm clicks is still very very close between both services. PokeFarm also gets zero traffic from social media sites.
The only thing I can see (since Google does not provide any real information to its AdSense and AdWords members) is that the articles that were getting serious attention via the social media networks have had their clicks discounted in the 95% range. I still can not believe that MBL is mis-reporting on clicks because I have other sites (with no social media attention) where the click data is still within 10% across both services.
If you are a blogger who runs AdSense and MyBlogLog please let me know if you are also noticing this social media discounting. I would like to think it is only me and not something Google is blindly doing in an attempt to take care of its click-fraud problems.
Update
I just found this article on click counting with Shuman Ghosemajumder. I will provide the interesting bits below:
We monitor signals like Internet protocol addresses and the kinds of browsers people use--for traffic to look real, it has to have the right proportion of visitors using the Firefox browser, Internet Explorer, et. cetera.
Does this mean that if you get a high number of Firefox users coming to your site you will also be assumed to have a high number of fraudulent clicks?
When they land on the advertiser's site, they click on products, hitting the "back" button to go back to the landing page. Many browsers reload the landing page each time. We don't count those as clicks, but third-party auditors actually register each click on the "back" button as another click on an ad, which grossly overestimates the number of ad clicks.
That would not apply here because MBL only tracks the click when they leave and does not follow through to the site they go to next.
A very simplistic fraudster might just click on ads, over and over. But of course, we've learned to recognize spikes in the click-through rate, the number of times the ad is clicked on for every time it appears.
Translation: If you have a traffic spike we will discount those clicks. (Social media anyone?)
The number of clicks that we proactively throw out is less than 10%. So then the question is really: How much are advertisers getting for free thanks to our detection methods?
No, the question is, "how many of your AdSense publishers are you screwing over by not paying them for legitimate traffic?"
The problem as I see it
AdSense and AdWords are two halves of the same program, you can not have one without the other. Yet Google continualy treats their AdSense side of the house as a bunch of criminals by "assuming" that 10% (or more) of clicks are fraudulent.
Shuman Ghosemajumder talks a good game about providing a great ROI for AdWords users but the simple fact is that when someone clicks on an AdSense ad and arrives on the AdWords member's site it is up to them to make the conversion and earn their ROI not me. My job is to send them the traffic. If someone arrives on your landing page and bounces away there are a few things that may be happening:
1. Click-fraud
2. Your site sucks
3. Your ad did not properly convey what your site is and the clicker did not find what they were looking for.
Of those three only the first one should result in the click being discounted and the AdSense member not being paid. I would further postulate that someone who is attempting to do click-fraud would tell his/her friends to be sure to click through a few pages on the destination site to make it harder for the system to tell if it were true click-fraud. Why do I say this? Because I read an article a while back about a "click-fraud ring" that does exactly that!
I think it is high time that Google acknowledges the fact that if there were no AdSense program there would also be no AdWords. The two can not survive without each other.
- Google Assumes 10% click fraud rate
- Google Analytics
- If Google can’t count how can I count on Google?
- Saying Good Bye to AdSense
- Why Ad-Block Users Are Good For Your Blog
Our numbers have always been drastically different between the two systems.
August 2007:
- MBL 2374
- Adsense 1576
Sept 2007 (so far):
- MBL 1588
- Adsense 1244
August was way, way off between the two. I’d love to get credit for those missing 800 clicks.
Your August numbers are not as far off as mine were!
The thing is, out of all of the sites I run both MBL and AdSense on this one (hmtk) is the only one showing this huge difference. It is also the only one with “spikey” social media traffic.
I am seriously beginning to think that Google is looking at social media (as a whole) as click-fraud and discounting clicks that come from soneone who arrived via social media.
Question for you Matt. Look at your incoming numbers, is your click data off every day or only on days where you get a large amount of social media traffic?
I just looked at the month as a whole. I’ll have to go back and spend time with all the data and look for patterns.
Steve, check out getclicky.com as another stat tracker. I’m setting it up tonight to give it a try (1,000 pageviews/day for free, you can pay to raise that limit).
I’m now in the middle of replacing my AdSense blocks with AdBrite blocks. I’m going to do this for a week and see if my CTR jumps back up or not. If it stays low than odds are MBL is at fault. If my CTR jumps back up to the normal level that I will take that as indicator of something fishy going on over at Google.
I have already dropped the Google Ads from rotation on PokeFarm (may take time for AdBrite to adjust) except for the very bottom one on site. I may not be a huge AdSense publisher but I do provide Google with over 400K AdSense impressions every month when you factor in all the various sites I own.
If there was some way I could test the tracking without resorting to click-fraud I would. The only thing I could think of would be to open up an AdWords account and pay for CPM on my own site and ask people to click ONLY on my own CPM ads but that is just too much work and a pain. Instead I will just replace the Google AdSense blocks with AdBrite blocks and see how the CTR goes.
It will not be foolproof because it also depends on how relevant the resulting advertising is. I’m trying to get into Yahoo Publisher but no response from them yet.
Very nice research and great findings. Sorry I couldn’t help with my MBL stats, but I’ve been really busy and didn’t have time to pull all the data in time. So will there be a follow up article on AdBrite? Seems like a cool alternative.
Granted, your job is to send traffic to advertisers, but you need to be realistic about the bind that the advertisers (and engines) are in. There is no foolproof way for anyone to determine that a publisher has not clicked on his or her own ads. Furthermore, even if the clicks are not fraudulent, they’re not converting (at least not to the advertisers’ expectations), so a rational response is for them to spend less money. In order to keep their business, Google must somehow make it worthwhile for them to continue advertising, and discounting ads is one way of doing that. But the unearned money has to be accounted for somehow, such as what is paid out to publishers, employees, or shareholders.
If click fraud became a more serious problem, and Google (or someone else) could not contain it, they would be forced to drop the sources of traffic that are the most difficult to determine whether fraudulent activity occurred. This means they would need to continue to drop AdSense sites that are questionable sources of traffic, or even close down the program altogether.
If you sold space on your sites directly to advertisers, who were getting low ROI, what would you do? How would you convince them to continue to advertise?
In general, I have expressed my doubts about Google’s claims, but I can’t fault their decisions to issue refunds (if the alternative was that they would drop out of the program altogether).
WRT browsers, my guess is Google keeps trends of various types of usage on sites, and they look for overall correlation with other sources. So if the Firefox (for example) usage exceeds some threshhold on some sites, it raises a warning flag. I wonder how they will handle mobile AdSense, however. For example, there are people who use the browsers that come with phones, but others who use browsers from tablets and PDAs. Some of these are from well-known companies, but others are of the homebrew open-source type. So a spike in that usage could be fraudulent, but could also constitute broader adoption of one of the newer browsers. This is just an example of how difficult it is to detect fraud, because there are other possible scenarios. But the fraudsters know this, so they can exploit the uncertainty. But there is no win-win situation, because either advertisers are overcharged (for poor ROI), they drop out of content networks, or the engine has to issue refunds.
I have no problem with Google refunding advertisers but they are doing it at the expense of publishers and bragging about it as if they were the only ones taking the financial hit!
I am currently phasing out AdSense. I’ll get paid at the end of the month so I will turn off all of my ads before then. The ads on here are already gone.
As for direct publishers, I charge a flat monthly rate and offer no guarantees. the only thing I can point to is past performance which can not be used as an indicator of future results (just like the stock market). What I also offer many advertisers is a 15 day free trial. If after 15 days the ad is not making you happy than I remove it and you are out nothing.
Click fraud is likely to be very hard to detect but in my particular case if all of my clicks are fraudulent i would expect to be banned from Google. Look at the numbers because the discount rate applied to me is far higher than 10%!
I opined some time ago that the quickest way to kill CPC advertising (and AdSense) would be for all of the diggers in the world to turn off their AdSense and be sure to click an ad on every site they visit via digg. Perhaps they have already started doing this and that is why Google appears to be discounting all of my social media traffic that generates clicks?
[...] into other areas. I thought everything was going good until about a month ago when I noticed that my tracking data was no longer even close to Google’s tracking data in regards to clicks. Around this same [...]