| 6:24 pm on Dec 11, 2006 (gmt 0)|
Interesting, if that's accurate.
Thanks for sharing.
| 6:27 pm on Dec 11, 2006 (gmt 0)|
Good artice, some observations:
"more than [u]two dozen[/u] Google employees" for manual reviews is too vague, for their number of publishers and advertisers I would have hoped it would be much more than that.
we'll have to take their word for the "Google is becoming more transparent" bit.
| 7:41 pm on Dec 11, 2006 (gmt 0)|
Good article. Worth signing up here just to find and read stuff like that
| 11:14 pm on Dec 11, 2006 (gmt 0)|
Of course, the only proof we have is their word.
| 12:46 am on Dec 12, 2006 (gmt 0)|
I'm guessing it is accurate. The referenced number is for 'user detected' click fraud and I'd be willing to bet that the number of users who can actually detect click fraud is very small.
| 4:24 am on Dec 12, 2006 (gmt 0)|
So, advertisers are detecting a little less than 2% invalid clicks and telling Google whom is verifying their claims.
How many invalid clicks is Google and advertisers not detecting?
I don't see the validity of this threads title "Google’s Click Fraud Rate is Less than 2%".
| 4:29 am on Dec 12, 2006 (gmt 0)|
If no one detects an invalid click, is it indeed an invalid click?
| 5:08 am on Dec 12, 2006 (gmt 0)|
That article smells just like when a White House story is "leaked" to the press just to give out some information in a non-formal way. That way, you can steer the discussion but can't be pinned down on any details.
Plus, what the hell was the that circle diagram all about. Without any meaningfull explanation of the way the areas where determined it might as well of been a drawing of puppy dogs.
| 7:37 am on Dec 12, 2006 (gmt 0)|
The author of the article seemed to be basing his numbers on the relative sizes of the circles in that diagram. I'm not sure that's a valid analytical technique, LOL.
Correct me if I'm wrong, but the diagram seems more like an illustration of a simple concept, not a drawn-to-scale chart of the distribution of invalid clicks.
| 9:19 am on Dec 12, 2006 (gmt 0)|
>Correct me if I'm wrong
Why would Google publish an accurate (to scale) diagram of the amount of click fraud? It's just there to give a general idea.
The less than 2% invalid clicks is an estimate by Shuman Ghosemajumder "Google’s business product manager for trust and safety" not a derivative of the diagram.
At least that's what I understood.
| 12:55 pm on Dec 12, 2006 (gmt 0)|
So what we all got for xmas was google fud ..via an outside blog ..and a very tacky powerpoint explanation designed to look logical to 3 year olds ..then again anyone who beleives a word of what we know as the "xmas story" might go for it ;-)..
surprising thing is that the blog didn't mention the cow and the ass looking on and the star above the plex to guide the webmasters.
| 2:32 pm on Dec 12, 2006 (gmt 0)|
Very un-informative, and sometimes plain ridiculous.
|Another common example of how click-fraud detection companies get it wrong is when counting reloads of an advertisers landing page. He described how a customer could click through to the landing page, view a product page, and then hit their “back” button, returning to the same landing page. Without correct tagging, Ghosemajumder said that one click and five page re-loads could easily be mislabeled as six clicks from the same visitor. |
Huh? Is this guy serious?
Is that an example of G's top-secret techniques to detect invalid clicks?
The simplest tracking software must handle this correctly, or it would be useless.
| 3:56 pm on Dec 12, 2006 (gmt 0)|
|The less than 2% invalid clicks is an estimate by Shuman Ghosemajumder |
I don't think so. I just took another look and the only specific figure I saw attributed to Ghosemajumder was that overall percentage of invalid clicks (including clicks filtered automatically by Google) was "in the single digits". Everything else appeared to be suppositions and guesswork.
The fawning fanboyism of the article is also a little hard to take. This forum has a number of people who usually line up on Google's side (I'm one of them), but that kind of support is nothing compared to the gee-whiz breathlessness of that article.
| 4:34 pm on Dec 12, 2006 (gmt 0)|
2% has to be somebody's estimate. If Google really knew which clicks were invalid, the fraud rate would drop to zero. :)
| 6:37 pm on Dec 12, 2006 (gmt 0)|
>I don't think so.
|For those of you short on patience or time, here’s the revealing information gleaned from my conversation with Ghosemajumder: |
The click fraud rate - as discovered by most AdWords advertisers - is on average, less than 2%of all clicks through Google’s system
|Ghosemajumder then proceeded to share another graphic that more clearly describes Google’s filters and appears to confirm that undetected click fraud is less than 2% |
|They have no reason to keep quiet for much longer, as Ghosemajumder explained. Google is already filtering more than 98% of invalid clicks, before they show-up in the AdWords console |
| 7:05 pm on Dec 12, 2006 (gmt 0)|
I'm frustrated by the vague phrases such as "appears to confirm", "by deduction, we can already estimate", "looked to be", etc. throughout the article.
Anyway this seems to be the salient passage. Google won't make any official assertion, but does appear to be confirming the author's interpretation with a wink and a nod:
|...the amount of user-detected “requested investigations” looked to be a number in the 1-2% mark, at most. While Ghosemajumder would not reveal the exact number, he did confirm that my estimate was on track, “if not less”. |
| 7:27 pm on Dec 12, 2006 (gmt 0)|
I know, and "more than two dozen Google employees" manually checking for click fraud, how many is that, 25? 250? very vague.
[edited by: Hobbs at 7:27 pm (utc) on Dec. 12, 2006]
| 10:21 pm on Dec 12, 2006 (gmt 0)|
The Tuzhilin report mentioned a "Spam Operations" component of the Click Quality team whose size was about two dozen.
| 11:23 pm on Dec 12, 2006 (gmt 0)|
>>>"more than two dozen Google employees" manually checking for click fraud, how many is that, 25? 250?<<<
i'd guess that it's less than three dozen, because they would have wanted to take credit for every dozen they could, instead of just two dozen :-/
| 1:53 am on Dec 13, 2006 (gmt 0)|
Once upon a time, Europe had an extremely low rate of mad cow disease. What caused that rate to shoot up? They did way more and better testing for mad cow disease. Nowadays, if you want that real low rate of mad cow disease, you have to go to the U.S.. How do we achieve such low rates? Why, by testing very, very few cows. In fact, the U.S. forbids cattle ranchers from testing their own cows for mad cow disease (I am not making this up).
The point is, the title should not be "click fraud rate is less than X%", but should be "click fraud rate is virtually guaranteed to be greater than X%", where X is whatever number Google thinks they are catching. When you catch 2 mice in your mousetraps, only a fool thinks "huh, I guess there were just 2 mice in the house."
If Google wants to inspire confidence in their detection methods, they should hire an independent black hat team to perpretrate and document a series of click fraud attacks, and then hand over all their detected fraudulent clicks and let the independent black hat team reveal how many were detected correctly.
Ain't gonna happen. Unless and until fraud fear (or lawsuits demanding fraud refunds) start to seriously impact the bottom line, Google has no motivation to allow a serious independent audit of their fraud rate.
Instead, it's in their best interests to do what they are doing here: try to show bits and pieces of information that cast their secret (and ultimately, unknown, even to Google) fraud rate in the best possible light.
|Wow! A user-detected click fraud rate of less than 2% is a world away from the 20% number given by some click fraud detection companies. |
This author does not exhibit much ability for critical thinking. What does Google's claim for X% of fraud, distributed across the entire Google population of advertisers have to do with the fraud rate claimed by a click fraud detection company? Can the author really not grasp how unlikely it is that click fraud is distributed uniformly across all AdWords advertisers?
And how random a sample do the click fraud agencies represent? Let's see, if you pay a click fraud detector, you likely a) spend more than a few bucks on advertising and b) already suspect fraud. Hardly a random sample of the AdWords population.
Although it was a year or two ago, I myself have seen a near 100% fraud rate on a relatively obscure keyword on the Search network that Google absolutely did not auto-detect. After some weeks, I was able to get refunded, so apparently they agreed that this absurdly visible attack (that they did not auto-detect) was fraud. Even if Google were God, and therefore able to state with 100% certainty what their overall fraud rate was, that has little bearing on what an individual publisher may see.
It's sad to see Google willing to be intellectually dishonest like this. I see they are also still trying to claim that their internal employees who use AdWords and Adsense themselves are competing on a level playing field with non-Google employees. Sad. The "do no evil" pledge is clearly not a superset of "speak honestly".
|Google is already filtering more than 98% of invalid clicks |
What a gullible interviewer! When you don't know how many invalid clicks there are, you cannot know that you've filtered 98% of them!
|So why did Google break its silence to MarketingPilgrim.com on this touchy subject? How did we manage to get the real click fraud number out of Ghosemajumder? |
Ummm, I'm guessing it's because you established yourself as someone who is easily impressed and can't reason very well. If your current website doesn't work out, you might want to look into becoming a movie reviewer!
| 2:24 am on Dec 13, 2006 (gmt 0)|
Count me in the minority, but I believe their figure of 2%. The only people who can reasonably profit from click fraud are publishers, and we're always complaining, so how many can there be actually cheating?
As a publisher, I want to endorse the lowest possible fraud rate and anything higher than 5% actually seems pretty much absurd, so 2% is no stratch at all. G has a reputation at stake, but besides that, the ROI for advertisers has been historically solid and that is indicative of a low fraud rate.
My feeling is that this whole issue has been blown far out of proportion. Gaming adsense would seem to be a pretty serious undertaking and difficult from which to profit.
Again, maybe I just don't see it, but until somebody explains to me how anybody other than a publisher could profit from click fraud, I don't see it as a real problem.
| 5:33 am on Dec 13, 2006 (gmt 0)|
|The only people who can reasonably profit from click fraud are publishers, and we're always complaining, so how many can there be actually cheating? |
Advertisers can also commit click fraud to deplete competitors ad budgets, thereby pushing their own ads to the top.
| 5:57 am on Dec 13, 2006 (gmt 0)|
|What a gullible interviewer! When you don't know how many invalid clicks there are, you cannot know that you've filtered 98% of them! |
When you want an estimate, you go in with few assumptions:
- It is always an estimate, nothing is 100% accurate
- The more samples you take, the more accurate your estimate will be
- If you reach a conclusive estimate, and take more samples and the same figure comes out, you are in the ballpark of an accurate estimate
This is how the 2% can be as accurate as possible, the only gripe one may have with the report could be in the research not being verified by an independent authority, which raises the question, can Google be trusted enough to generate unbiased reports?
My money is on yes they can be trusted, they are a public company, have demonstrated mostly ethical behavior, the only vice I see in them is not a fault of their own, but in the highly paranoid attitude needed to survive a ridiculously litigious business environment.
The report is missing only one thing, the word "estimate".
| 4:33 pm on Dec 13, 2006 (gmt 0)|
hobbs, you are missing the entire point of what's been said:
"The click fraud rate - as discovered by most AdWords advertisers - is on average, less than 2% of all clicks through Google’s system"
adwords advertisers are not capable of determining how much click fraud there is, because only google has access to the data that's needed to determine an accurate number.
| 6:11 pm on Dec 13, 2006 (gmt 0)|
The disadvantage of building upon what's being discussed not what the report says!
| 9:06 pm on Dec 13, 2006 (gmt 0)|
Actually there are different ways of defining what at 1-2% number is supposed to refer to, and I don't think the article makes it clear. Ghosemajumder could be talking about: (in decreasing order)
all invalid clicks not discovered by Google, whether discovered by anyone or not
invalid clicks "discovered by advertisers"
invalid clicks discovered by advertisers and reported to Google
invalid clicks discovered by advertisers and reported to Google and reimbursed by Google
The first two numbers are speculative; even Google would have no way of knowing the exact figures. The final two numbers are known by Google, but they would of course be a significant understatement of the true number of invalid/fraudulent clicks.
| 10:56 pm on Dec 13, 2006 (gmt 0)|
the entire article is based on the concept that google will always catch ~98% of all the possible click fraud that could ever happen... which is impossible to prove.
from the article: "...Wow! A user-detected click fraud rate of less than 2% is a world away from the 20% number given by some click fraud detection companies."
that worthless blog post is comparing apples and oranges, because all click fraud companies detect click fraud themselves, that's what they get paid to do... their numbers are never based on the rate of "user-detected" click fraud.
| 6:28 am on Dec 14, 2006 (gmt 0)|
|Again, maybe I just don't see it, but until somebody explains to me how anybody other than a publisher could profit from click fraud, I don't see it as a real problem. |
Competitors of advertisers and publishers can benefit from click fraud (to kick someone out of a position or out of the program).
Stockholders can benefit from click fraud to raise the stock price via clicking to generate more revenue.
Ex-employees of an engine or network can benefit by driving up the costs of processing potentially invalid or fraudulent clicks, etc.
These are just a few examples.