Maine Munchies Ad

Friday, March 02, 2007

Google AdWords Invalid Clicks Numbers are Invalid

With all due respect to Shuman Ghosemajumder, Google's invalid clicks numbers are misleading. They claim that less than 10% of clicks are invalid and that those are identified proactively by Google's systems. Beyond that, they claim only a tiny 0.02% of clicks are invalid and that those clicks are reactively identified by Google after advertisers complain. I think that number is misleading. What this number really tells me is that Google dismisses all but 0.02% of the complaints that advertisers send in.

The real problem is with Google's definition of invalid clicks. I recently requested a click quality investigation which was summarily dismissed. When I filled out the form I was told to expect a response after a few business days. I included detailed information from server log files to support my claim that the clicks were invalid, of a low quality, and not from normal search activity. Did I get a response after a few days? No! The response came after only a few hours:
Our team received your report regarding suspicious clicks on your AdWords ads. Thank you for your patience while we researched this issue.

After thoroughly reviewing your ... campaign from January 1st through the 10th, we were unable to find any conclusive evidence of invalid clicks charged to your account. The clicks your ads received appear to fit a pattern of normal user behavior.

Our monitoring system is designed to protect advertisers' ads from unethical or automated activity. Multiple data points are automatically analyzed for each click, as our system aims to discard potentially invalid activity before it is charged to your account.

...

If you aren't satisfied with the value of the traffic from searchportal.information.com, please reply to this email. We can then assist you in preventing your ad from appearing on these pages.
With the amount of information I submitted when filling out the form, there is no way Google performed a thorough investigation. I expected them to request more information or ask for clarification on some of the details I submitted. They could have at least waited a full business day or two to respond to make it look like they actually investigated the matter. Clearly, the 0.02% number means that Google ignores and then dismisses click fraud complaints 99.98% of the time. ;-)

I contend that any click from searchportal.information.com for an AdWords account opted out of the Content network is an invalid click. This is distribution fraud (others are calling this practice syndication fraud) and this could be a bigger problem for advertisers than Google's definition of click fraud. Now, why do I think these clicks are invalid? Perhaps excerpts of the information I provided Google will shed light on the matter (I'm replacing my client's keywords with "Foo Bar" so as not to divulge my client's important keywords):
Trends: Anywhere from 10%-40% of the traffic was coming from searchportal.information.com. That's clearly artificial. More traffic from this domain than from google.com itself. Plus, all of the referer info had query data that had the first letter capitalized. Here's an example referring URL from the logs:

http://searchportal.information.com/?epl=[snip_long_string]&query=Foo%20Bar

Note that the actual "search" query was: Foo Bar

That's not normal search behavior. When people search, they don't take the time to capitalize the first letter of each keyword. Most searches are either all lowercase or all uppercase. So, these are clearly not searches. These are people clicking on links. That's fraud. Clicks from the Search network should come from searches, not people browsing links.

Data: See data from above. Let me know if you need further information. I request that you identify all clicks for the history of this account that ever originated from searchportal.information.com and you refund my client for those charges. Note that in many cases, these CPC charges were in the $2-$3 range so I'd expect a refund of thousands of dollars. Here's a count of hits from these parked domains from last month, for instance:

$ grep searchportal.information.com ex0612.log |wc -l
164

Figuring on an avg CPC of $2.50, that's $410 for last month alone. Also, what disturbs me the most is that this click fraud seems to be approved by Google. These parked domains should not be on the Search network. My confidence in the Search network is severely shaken. For this client, we have simply turned off the Search network. After you have issued them a clickfraud refund, I'd like to know one of the following:

1) How to block searchportal.information.com on the Search network (I tried site exclusion and that did not work - even though I've heard rumors that site exclusion works for parked domains on the Search network and not just the Content network)
or:

2) That searchportal.information.com (and any other domains owned by the parent company, oversee.net who owns domainsponsor.com) has been kicked off the Search network

If I get a satisfactory answer on one of those points, I'll turn the Search network back on for this client, either with that domain blocked or with confidence that they're not on the Search network.

Dates_Times: I will start with this month. They've hired me recently but I see the click fraud in prior month's log files as well. All of the click fraud is from a single domain: searchportal.information.com which appears to power parked domain sites run by domainsponsor.com. This campaign was opted into the Search network but out of the Content network. IOW, all of the click fraud is on the Search network. Timestamps (GMT I think) + IP addresses for this month (by keyword):

$ grep searchportal.information.com ex0701.log |grep Foo%20Bar |awk '{print $1,$2,$11}'
2007-01-01 07:00:17 68.62.110.159
2007-01-01 15:54:04 67.8.81.151
2007-01-01 17:58:51 70.143.52.149
2007-01-01 18:03:51 24.145.212.157
2007-01-01 18:04:40 70.143.52.149
2007-01-01 18:13:10 70.143.52.149
2007-01-01 19:04:07 64.12.116.11
2007-01-01 20:15:32 71.101.26.230
2007-01-01 21:35:48 199.126.188.14
2007-01-01 22:53:47 24.127.194.77
2007-01-02 03:31:41 74.225.65.47
2007-01-02 04:13:04 76.175.242.247
2007-01-02 04:13:55 68.160.60.229
2007-01-02 22:46:54 65.11.42.133
2007-01-02 22:51:30 65.11.42.133
2007-01-02 22:55:04 74.233.0.201
2007-01-02 23:34:32 66.213.126.226
2007-01-02 23:34:34 71.198.15.32
2007-01-02 23:34:42 71.166.143.94
...
(this goes on and on and on and on and on...)
I think it's worth stressing the point that the ads were not being displayed for actual searches. No one was actively typing keywords in a search box. All of these paid clicks came from people somehow getting to a parked domain and then browsing the navigational links provided by that domain. If these are not invalid clicks, then I don't know what a valid invalid click is. These parked domains do not exist to help people find what they're looking for. No, their goal is to drive traffic to pages with high pay per click fees. Here's an example of the "navigation" from one of these parked domain sites:

parked domain navigation
Notice how the first letter of each word is capitalized? Someone clicking on a link like these and then seeing an ad isn't even close to someone actively typing keywords into a search box and then seeing ads. This is clearly closer to contextual advertising than search engine advertising. Shame on Google for including these sorts of sites in their Search network.

BTW, it's only going to get worse. One of the bigger companies in the direct navigation space, NameMedia, just bought a domain parking company (Smartname) and announced a switch from Yahoo! to Google as their ads feed provider. This company has "quietly amassed more than 650,000 Internet domain names." If you use AdWords, watch your traffic carefully! Getting back to searchportal.information.com, the source of invalid clicks I identified and Google chose to ignore, if you look at this traffic graph, you should be alarmed. No, I'm not surprised by the 0.02% invalid clicks number reported by Google. That's simply the number of invalid clicks they choose to NOT ignore.

Tags (created with TagBuildr): , , , , , , , , , , , , ,

1 Comments:

Anonymous Anonymous said...

Agree, why would Google provide more information to advertisers unless they're forced to? I'm seeing many clicks are invalid, even turning off the content networks

Mon Jun 30, 01:27:00 PM EDT  

Post a Comment

<< Home