Slashdot Mirror
Google Releases Analysis of Click-Fraud Detection
fragmentate writes "This morning Google released information about their analysis of the exaggerated click-fraud numbers. Without pointing fingers, they mention that click-fraud analysis companies need to clean up their methods. From the post, 'A rigorous technical analysis by Google engineers has found fundamental flaws in the work of several click fraud consultants - flaws that help explain why widely quoted estimates of the size of the click fraud problem are exaggerated.' They even point out some obvious shortcomings of the methods used. The entire report [PDF] is available with their complete analysis."
In this case, it sounds like some are crying "click fraud" in order to pay less for the clicks their ads generate. If they can claim large amounts of fraudulent clicks, they pay less to Google.
Sure click fraud exists, but I imagine these "consultants" are advertising themselves as a way to pay Google less, while still having a high volume ad campaign. Taken to the logical extreme, any click-through that doesn't result in a sale was a fraudulent one.
Google Search: 3.5%
Google adsense: 0.25%
This is what kills me. Companies are so willing to fork over a ton of money for cost-per-click (CPC) advertising, when so many sites are not friendly to search engine spiders for organic (non-paid) searches. It's one of the biggest, and most overlooked reasons to use standards-based design practices. And it's free to do so (at least, if it's done the first time)!
In many cases, CPC advertising is another example of throwing money at a problem for a band-aid.
It amazes me anyone would pay any attention to them in the first place.
Google has a great solution for that. If the transaction is online, you can embed a small piece of HTML/Javascript code in your 'thank you for purchasing' page that allows Google to check the value of a cookie they placed on a customer's computer when they clicked an ad.
The cookie links the click to the sale. And there is value to the advertiser as well: Google can then help you track which ad resulted in a sale, and which keywords it was linked to. (So you don't have to buy an expensive but poor-return keyword.)
(I may be mis-describing: Check Google's docs to be sure.)
Join Tor today!
If quotas exist, whether set by man or machine, mechanisms will eventually appear to ensure that quotas are met.
If Software X must discover Y amount of fradulent clicks, then there will eventually be a means that makes certain that Y amount of fradulent clicks are discovered.
For Google, how much of the budget depends on discovering X number or Y percent of fraudulent clicks?
For Microsoft, how many pirated copies of Windows must be discovered each day/week/month/whatever?
The hypothesis may apply in other cases. How much of a town's civic budget depends on income from traffic violations? What happens if traffic violations fail to raise that revenue?
Look for quotas. Sometimes the numbers are the answer.
Unfortunately this is not true with pay-per-click, which is where google makes 99% of their income. At the end of a day in reality, you are paying for CLICKS. The question is are those clicks generated by humans, or by scripts, or by people trying to rip you off.
I often hear the same stupid analogy: "Click fraud is no different than getting up to go pee during a commercial break and not watching the commercials, or tivo'ing through them in fast forward."
WRONG. With clickfraud, you can make REAL, ACTUAL, CASH. I defy you to give an example of how someone can make REAL, ACTUAL, CASH by going to the bathroom during a commercial.
Face it- Google will never end the controversy until they have third party auditing. Right now, all they have is "trust us". Guess what- there's no other advertising industry on earth where "trust us" is good enough. They ALL have auditing by independent parties. Cuz guess what? Businesses will try and rip each other off if they know they won't be caught. Google has already proved they are perfectly willing to rip off advertisers since they ADMITTED they have been charging people for "doubleclicks" for YEARS, even though they knew all those clicks were invalid because they came from the same IP/cookie and happened within 1/4 second of each other.
----
At the end of the day, you're paying me for access to my readers' eyeballs. If your product is irrelevant, overpriced or otherwise not useful to my readers, or you lack the marketing skills to gain their interest with your ad, why should I be forced to let you off the hook on paying me? You still ran your ad on my site...
Interesting, but wouldn't it make more sense to have a neutral party do the analysis instead of Google, whose bulk of the revenue comes from those same clicks they analyzed? Having Google do the analysis and reporting is like having Microsoft do Vista benchmarking. That is, if Vista were actually ready.
Simpy
Actually, here's an entertaining idea for privacy concerns. Munge/Obfuscate the IP Address. If I see a bunch of clicks from xxx.xxx.253.99 or from address "Whahoopa" in a day, I can kind of assume that something fishy is going on.
Networks charge advertisers REAL, ACTUAL, CASH for that guy going to the bathroom during the commercial. The ad rates are determined by the size of the audience, so if the audience counts the guys going to the bathroom then the advertiser is being overcharged.
.... for taking these "analysts" to task with some facts, and publically. Many companies would have just deferred to presenting it all in a libel lawsuit.
Is click fraud an issue? Certainly.
However, these companies purporting to provide analysis and actually providing nonsense are just as guilty of defrauding the advertisers as the click fraudsters they purport to guard against.
Suppose...
1) the networks had some way of counting the number of TV's on a certain channel.
2) TV's were cheap enough the Joe Scammer could buy 10000 of them and have them all turned on to a particular channel.
Then the rates for those ads would go up (because your ad is hitting more people, right?), and the networks could give some % of the resulting money to Mr. Scammer.
Bam. Click fraud. Or rather, view fraud.
The differences are:
A) 1) and 2) above don't work for TV - they do for the internet.
and to a lesser extent:
B) Going to the bathroom during ads is not commercializing the process.
C) Google is the middleman in the adclicks, so the proper analogy of the networks is the final website. You, the network, can make money and pass money on to Joe, similar to how you the website can make money and pass it on to the click services.
The particular dynamics, mostly A), make it so view fraud isn't possible and wouldn't be economical. That said, I seem to recall that advertisers were wary of Tivo for exactly this reason - it would decrease the efficiency of their ad dollar.
So the reason we don't see the difference is because you're wrong. Sorry.
Because you get a return that you're happy with. Although Google does not have an independent auditor to go over their click data, advertisers still pay. Some feel they are not getting enough conversions for the money they invested in their ppc (pay-per-click) campaign, some feel they do. Some cancel their accounts and move on to a different kind of web marketing, some stay and use the Google system.
In the end, it's a service, not an obligation, and even though most countries already formalized auditing measures for other types of advertising mediums, it doesn't mean it makes sense. Personal responsibility and decision making are applicable in this case just like always. If all of Google's clients demanded auditing, there would be auditing. There's definitely no need for governmental regulation.
End of the day, I don't understand why people pay for a service, with defined rules of use and known risks only to start crying murder later (well, I do understand why they do that, it's just that I don't understand why we are taking them seriously).
If you don't like AdSense or AdWords, close your account. You can. They have a button or something.
Much like how your celluar phone company prints out the list of numbers that called you, Google should provide a list of IPs that contacted the website with the search term. I don't see how it's anymore private than if you called a company because you found it in a phonebook and starting asking questions. What type of personal information is Google worried about releasing? For instance, if I search for "outlook plugin" and a company that sells the plugin shows up in the AdSense links on the side, and I click on it - how is that private? Ok, let's take it more extreme so I go and search for "russian girls", then I see Adsense links for that on the right side. I don't see how that is any different than if I called a company and asked for the same thing. Google is just an electronic white/yellow pages. If they just released documents with the request records most of this would be a moot point and people could cross-reference their Apache or IIS logs with Google's click logs.
Think about it this way: If some people make a living out of just crying wolf... guess what they'll do? Cry wolf. Lots. Invent gazillions of wolves to scare their customers with. Tell them that 250% of the North American wolves are in their backyard.
So, yes, basically that's what they claimed in that case: that 150% of the clicks were fraudulent. Literally.
Are you surprised?
All these "click fraud consultants" are people making their money by crying wolf. Unlike any other kind of consultant, they don't even have to fix the problem or anything. (Which, in other kinds of consulting is a clear test of whether the problem is real or bogus. If, say a DB consultant tells you that your querries are too complex, there's a very simple test there: then you write faster ones, please. If the app runs faster afterwards, ok, he was right. If not, well, the problem was bogus.)
but for this flavour it's a job that has actually less reality checks than an astrologer. As long as you say what the hapless customer wants to hear, in the form that he wants to hear it, that's all the "data" you need. And you already know from the start what the customer wants to hear. How convenient is that? You get called by someone who already strongly suspects click fraud (or he wouldn't bother paying a consultant), and has no clue how to check it (ditto.) You only need to do the sacred hocus-pocus and cast the holy runes (or the modern equivalents, involving spreadsheets and powerpoint graphs) and finally tell him "wow, you were right."
Plus, think long term. If you tell someone "well, there's a couple of dubious clicks in there, but nothing that would really tilt the statistics by much", that's the end of that relationship right there and then. If you tell him that your secret voodoo found 150% fraud, he'll call you next week too, to see if it gets better or worse.
A polar bear is a cartesian bear after a coordinate transform.
My father clicks on all sorts of things. He falls for the "Button" that looks like a windows button. The "speed up my PC" button, which installs spyware and actually makes it slower. He clicks on things that blink. He clicks on EVERYTHING. He opens spam. Bascially, if you're over 60, and use a PC, you're the one doing the clicking.
I spend a few hours a year cleaning his PC and making it usable again. I installed tinyPersonalFirewall a couple of years ago, and that helps with a lot of stuff.