How Google Manages Click Fraud

Finin writes "In February 2005, Google was sued by Lane's Gifts & Collectibles in a class-action lawsuit over click fraud. The company alleged that Google had been improperly billing for pay-per-click ads that were not viewed by legitimate potential customers. As part of a settlement earlier this year, Google agreed to have an independent expert examine their click fraud detection methods, policies, and procedures and make a determination of whether or not they were reasonable measures to protect advertisers. The report of the expert, NYU Information Systems Professor Alexander Tuzhilin (a Professor of Information Systems at NYU), is now available." Update 07/26/2006 at 12:52 GMT by SM: Fixed the link to Tuzhilin's report.

real link to report

this is the correct link, the other one is just legal blahblah:

http://googleblog.blogspot.com/pdf/Tuzhilin_Report .pdf [blogspot.com]

Re:real link to report

For anyone interested in reading the findings without having to wade through it all, then go to page 36 and start with section 9.3 where a little further on it also refers to account terminations and how this occurs, section 9.6 is the bit that I guess most may be interested in...you never know, you may then decide to read it all

- OptiRex [webmasterworld.com]

Statistics about click fraud shown to Adwords user

Google is (as of yesterday) now showing statistics about how many invalid clicks an adwords account has recieved. You can read all about it in the adwords blog [blogspot.com]

Re:Statistics about click fraud shown to Adwords u

Now this seems like a damn stupid idea to me. Say I'm trying to discover methods to click fraud my competitors or perhaps come up with automated software to sell. I can now use a dummy account with Google using search terms no one would hit and test different methods of fraud while getting feedback on which methods trip their detection.

Jonah HEX

Enron

If ad-sense is its major source of money, and it keeps the underlying numbers pretty well buried, could we be looking at another Enron? Imagine it comes out that 90% of all clicks are fraudulent. How many advertizers leave? How badlu does the stock drop? This is one of the things that makes me nervous about Google as an investment. Remember, Enron was loved by Wall Street too. Enron did not produce anything physical either. Enron reported great numbers. Underlying numbers were hidden away.

How is Google diferent that the big "E"?

Re:Enron

How is Google diferent that the big "E"?

Branding. For example, Google's famous motto is: "don't be evil". If memory serves, Enron's less-famous motto was: "Ken Lay needs a new boat".

Re:Enron

Google make pots of money because AdWords does result in a good return on investment relative to other forms of advertising.

Consider this. Imagine a fictional company that spends $10,000/month with Google on advertising, and is completely happy with the service because it results in lots of leads and sales. So it is actually profitable to spend this money. There are many companies in exactly this situation. Does it matter if 90% of clicks are fraudulent (which is probably not going to happen unless you are being deliberately targetted by a competitor)? If you are still getting a great ROI then no it doesn't matter. The invalid clicks are just noise.

Google does produce something valuable if not physical - it produces a ton of people viewing websites who then go onto buy things. Same as any form of advertising. Except the relevance of AdWords makes it more valuable than most.

There is no report

The PFG linked to in the summary does not contain the report of the expert.

Rather it is the answer to the judge and mentions (2 or 3 times, shortly) the report of the expert. All the meat that is to be found in the PDF is that the report is conclusive that Google does all it can reasonably to combat click fraud.

The PDF is interesting only if you're interested in legal stuff...

My 0.02

Link Fraud

If you don't have time to read the full 47 page report, Search Engine Watch [searchenginewatch.com] has summarized some of the most interesting findings.

404 - File Not Found

From the objections_response

Nice quote:

"The California attorneys take the position that the damages are 200 times $90 million, or 18 billion, which is more revenuse than google has received in its entire existance"

You just have to hand it to lawyers, they'll try anything.

Re:From the objections_response

Let's say your advertising budget is $100. In our example, each $1 you spent on advertising (historically) returns you $20 in sales and $5 in profits. You spent $20 of your budget on Google Adwords, 25% of which went to click fraud. So that's not only $10 wasted, but $200 less in sales and $50 less in profit that you could've had if there was no fraud.

That's the basic logic, I believe.

Billboards

Advertisers will pay for a billboard without any guarantee from the advertising company about how many people will drive past the sign, how many of those will read it, how many will take the information in and act on it. The client is assumed to be taking a risk in that regard.

Over time people decide for themselves whether a particular type of advertising is working for them. If the business keeps coming in why should there be a need for this type of analysis?

Re:Billboards

Advertisers will pay for a billboard without any guarantee from the advertising company about how many people will drive past the sign, how many of those will read it, how many will take the information in and act on it. The client is assumed to be taking a risk in that regard.

But "outdoor advertising" firms make no representations that they can measure these things. You know what you're getting when you buy space on a billboard. AdWords is different because Google sells it with claims that they can track these things (indeed, they bill you based on the results of that tracking). If Google's click tracking is inaccurate, you aren't getting what you thought you were getting. That's the difference.

Re:Billboards

That objection can be solved if you add the concept of 'noise' to Google's reported numbers, and I'd say most people already do that.. at least to some degree.

I don't think many people expect every adwords click to result in a sale for the advertiser. Instead, people just check to see whether there's a correlation between 'more adwords clicks' and 'more sales'. If 100 adwords clicks produce 30 additional sales, you can say that the adwords are 30% effective. Then it's up to you to decide whether those 30 additional sales are worth the cost of those 100 clicks. If so, you can write 70% of their adwords fee off as a cost of doing business. If not, you can close your adwords account, or lower the amount you're willing to pay per click.

If Google was padding its numbers, people would see lower return rates from their adwords accounts. Those people would then be less willing to pay for the service. They wouldn't expand their accounts to cover other words.

In other words, there's already a feedback mechanism that punishes Google for any loss of efficiency in its adwords program, and rewards Google for providing the best results it can.

Re:Billboards

But "outdoor advertising" firms make no representations that they can measure these things. You know what you're getting when you buy space on a billboard.

I worked in advertising for over a decade and your statement is totally false. These days EVERY advertising medium being pitched to a client MUST contain all kinds of analytical data backing it up, and this includes billboards. Locations, lighting, traffic patterns, etc, etc.

And I'm sure that some outdoor media companies have their own internal research demonstrating that some locations feature superior demographics (ie, this road is between the corporate HQs and the wealthy suburbs and gets seen by all the wealthy commuters).

Thinking about stuff

I'm writing a piece of ad-blocking software myself, and I was actually thinking of incorporating a few features. Specifically, the option of whether not to download the advert at all; to download the advert without displaying it; or to download the advert without displaying it and download the linked page without displaying it. Is this last option an example of "click fraud"?

i want my reparations

where is my money for being tricked into clicking on ads? I want $0.05 deposited into my account everytime I see a "Search for on EBAY!". Kramer, Jackie Chiles, and I will be in your office in the morning to sign the paperwork.

Ah..

So this is why I am asked to approve their new terms and conditions?

Academic community

It is interesting to see Google asked someone from the academic
community to do this study.

The most the study says is that the algorithms the Google
uses are good to detect click fraud. It does not give any clue on
how much fraud is not detected by the algorithms.

do not think much of click fraud detection system

Just a comment on their detection process. As a publisher who was terminated (and appeal denied) I do not believe in their process. I have a very active dinosaur website (approaches 1 million hits per month in school year -- probably lots of kids/teachers). it went well for one year then without notice (must be their autormated removal process) I was terminated for me or someother person associated with me generating what they classified as invalid clicks. Well I can state clearly i did not generate one invalid click, and I am only person doing website. So some other process was generating invalid clicks in their checking process. I am not sure what, whether with lots of activity I was getting those repeat 2 clicks that they filter out as invalid? Was there some spider clicking these (a competitor as I have heard about). All I maintain as a publisher I was terminated for nothing I did but was unfaily accused of doing invalid things. Does not make me very favorable to Google and their monolithic, unfair giant system. And I am happy to tell any one who asks what I think of their crummy system!!! Russ Jacobson Illinois State Geological Survey Champaign, IL

A good read actually - Googles Omnibus Response...

The last link is actually very good, and an easy read. Surprising for a legal document. It is "Googles Omnibus Response to Objections". I suggest giving it a read (PDF) http://googleblog.blogspot.com/pdf/objections_resp onse.pdf [blogspot.com]

It is basically a response to the objections of a grand total of 51 people in "the class". An incredibly small number of objections.

From the document:
"The assertion that Google has done nothing wrong was echoed by advertisers that opted out of the settlement."
"Unlike Retailers, Pay per click advertisers can limit the money risked for each click and for each day...Businesses should treat pay-per-click advertising like any other advertising...If it's costing more to advertise than your resulting profit, STOP ADVERTISING."

And, regarding the "click fraud detection", there is only a small portion of this document that mentions the review process by Dr. Tuzhilin. It does mention that the click fraud detection methods by Google were confirmed to be reasonable.

And finally, it was interesting to see read the jabs taken at the lawyers who brought the class action lawsuit to begin with...and the copy-cat cases from California, obviously a bunch of ambulance chasers.

Google Screws p2pnet.net out of its Share

http://wwww.p2pnet.net/ [p2pnet.net] is a website deicated to filesharing news. The owner of the website, Jon Newton, runs the website and barely breaks even. He subscibed to Google's adsense in order to generate some revenue. When a story about a filesharing lawsuit broke in the lamescream news, an article in p2pnet was referenced. This article generate a huge number of visits and therefore a much larger than usual number of adsense clicks. Rather than pay what was owed to Jon, Google accused Jon of click fraud and even showed information implying his guilt. Google continues to ignore Jon's request for information relating to this accusation and refuses to communicate with him to clear things up.

You can read about it at http://www.p2pnet.net/story/9086 [p2pnet.net] . This has happened not only to Jon Newton but also to many other small website owners. I am a Geek who used to love using Google, but now that Google has become big, it is doing what most other big companies do - screw the small guy and just walk away. Needless to say, I use alternative search engines instead.

Headline redundant!

"...NYU Information Systems Professor Alexander Tuzhilin (a Professor of Information Systems at NYU)..."
----------
Redundancy Police,
Department of Redundancy Department

Speaking as someone who has worked on anti-fraud

techniques... I know what they're doing for ad sense. They look for anomolies. You have a campaign c(1) running on sites s(1)...s(n). You are site s(j) where 1=j=n. if the ctr for c(1) across all sites s(1)...s(n) deviates by a certain amount compared to the ctr of c(1) on your site s(j), you are out of there. You can improve this by evaluating more campaigns per site. They also do this in reverse, or if they don't, they should. It's simple but effective. When you complicate it like this paper suggests it will get worse IMO. In fact I think that google makes it more complicated than what I have described hence their problem. The only way around this anti-fraud technique is to inflate impressions to keep CTR reasonable. But you can crack impression inflation since it requires a huge number of impressions thus more opportunities to detect fraud (wow, 10% of the impressions are from the same IP / cookied user).

Looks like I gave out the secret sauce.

So, a professor huh? He forgot to spellcheck.

Either that, or I need to catch up on my knowledge of the Internet.

He claims that "Computing devises attached to the Internet can exchange data of various types..."

Computing devises???

I am sorry, I just had to comment on this. It is not every day that Computer Science professors misspell devices :)

The conclusion

From the article: "Based on my evaluation, I conclude that Google's efforts to combat click fraud are reasonable."

Click Validation will be required soon

Gentlemen:

When it comes Truly validating Paid Clicks from any search engine, there are some critical points that you are missing and need to pay attention to.
This form of advertising is in its infancy and as we progress to an industry standard of Validating this space please remember the following:

1. Do you really think that it is in Google's Best Interest or in the Business plan to actively seek out invalid clicks or click fraud?

2. Every other advertising form in the Media has a 3rd party Validator, Checking the data on behalf of the advertiser. Television has Nielsen, Radio has Arbitron, etc. Even Accounting has SOX. Presently, only the Search Engine is the final arbitrator of credits or refunds. The industry needs to progress to this model eventually to give PPC advertisers peace of mind.

3. The search engine can only see click information up to the point where the ad is clicked- there are other attributes once the click goes into your sight (behavior wise) that the search engine cannot see but the Advertiser has data for. This gap needs to be reconciled by both parties.

Another alternative if this is not reasonable is to continue to let the bank balance your checkbook without checking it yourself.

Re:What Microsoft should learn from Google...

Yes, well the main difference here is that Google probably expects to be exonerated by this. They obviously think their fraud detection is going to be proven at least sufficient.
Whereas MS (and all of us forced to use it's OS) knows that if any 3rd party expert devs looked at their code in any detail, they'd probably be horrified. I'm sure they already know that their code is spaghetti, why prove it to the world?