Marketing & SEO Discussion List - LED Digest

 
Home arrow Featured Posts arrow Click Fraud - Searching for a Signal
Click Fraud - Searching for a Signal Print E-mail
Written by Michael Motherwell
November 6, 2006

Searching for a Signal Amid the Click Fraud Noise

Click Fraud What defines a valid click? Further to this: What defines an invalid click?

All the talk of click fraud reminds me of the mess in Iraq. Part of the problem with the war was that there was no plan for the peace. Click fraud complaints calling for a "solution" is putting the horse before the cart, IMHO, because we still don't know what it is we are solving, and we don't have any criteria for success.

So what would success look like? What is our plan for the peace? Do we want every click to be attributable to a real person? If so, I refuse to give you my details, so what do my clicks count as? If we want anyone that clicks twice in a day to be charged once, is that really realistic? What defines the "anyone" from the previous sentence? A cookie (I turn mine of)? An IP address (think AOL)? What?

And what % of clicks as fraud will success be? 30%? 20%? 10%? 0%? -10% (not as unrealistic as it sounds, as Search Engines could, in a gesture of goodwill, only ever charge for only 90% of valid clicks)? -20%? (Stop when I get there).

If we want a "solution" to click fraud, when need to know what that looks like, and then go get it. So lets hear some proactive ideas and means. Otherwise, click fraud is just, IMHO, more noise in a world full of dangers the TV won't stop barking at me (what, you mean to tell me that drinking red wine may not be good for me? But I am up to a bottle a day to fight heart disease. Oh man..  Stuff it, poor me another glass).

> Stockholders in Google and the other companies whose
> revenues and earnings are tainted by click fraud scams
> may take a hit, along with the companies.
    - David Yancey, LED Digest 2279

There was a study I once saw cited that showed that, as a result of the fear post-September 11, many more people died on the road because they were frightened to fly. Despite the fear, flying was and still is the safest way to get between two places significantly apart. And that was true even on September 14. In the same vein, the greatest risk to the profits of Google et al is not click fraud, but the perception of click fraud. Ironically (not really irony, more irony in the sports writer sense), it is also the greatest risk to advertisers, because generating and citing the fear of click fraud causes two things:

1. It stops people from advertising profitably which, lets face it, is the goal of all advertising. I know people don't like to hear it, but if you gave me a way to spend $100 and make $110, I would do it all day long. I don't really care if I should really only have paid $90, I am sitting over here with my Latte and Friand living it up on the profits I am generating.

2. It stops all of us from getting a good solution. I am sure people are going to doubt this, but hear (OK read) me out.

When an issue becomes an important publicity piece, what we invariably get is not a good solution, but a solution that makes the noise die down. The two are not one and the same. A solution that pleases people is usually one that sounds convincing, but has very little real value. All examples I could have would be political, so I won't go there, but ask yourself the how often an issue becomes a big news item and was solved well. My guess would be rarely at best. That is because the goals are not aligned. Party X wants the noise to go away, and party Y wants the issue solved. Party X (cool name that) probably will settle for any idea that sounds good, and most likely the FIRST idea that sounds good. I would personally rather have a good solution, not a hasty appeasement.

I firmly believe in and want click fraud to be handled transparently and fairly, and I hope everyone tries to keep that goal in sight, and not concern themselves with fears that are irrelevant and cloud my muddy mixed metaphor.

Michael Motherwell


Comments (3)add comment

Barry Mills said:

  There are 2 types of click fraud. There's organised, mass clicking by humans or robots. Google ought to be able to detect that, and if it does, it's important the perpetrators are prosecuted, not just banned from Google. Much harder to detect, and much more prevalent I expect, is site owners manually clicking ads on their own sites from time to time. I doubt that can ever be stamped out, but I don't really think it matters too much.

What surprises me is all the talk of advertisers losing and Google gaining. I don't think so, not in most cases. PPC is a "perfect market" and if fraudulent clicks were reduced, the value of clicks would go up, bids would follow, and equilibrium would be restored. Google get paid for generating sales, not clicks, because advertisers determine their bids according to expected ROI. So the victims of click fraud are most probably not the advertisers, but the honest publishers, who get a bit less for their genuine clicks than they should.

Technologically, if there's an answer, it might be the use of affiliate style tracking to sales, with low-converting sites being weeded out. What matters (or should) to an advertiser is the cost per acquisition or sale, rather than the proportion of dishonest clicks.

As an advertiser, I'm more worried about the PPC ads on irrelevant parked domain names than I am about bogus clicks. On domain parking sites I'm sure most visitors click a link or two just because they are bewildered as to why they've typed in a domain or followed a search link to generic_specific_term.com and found a page full of random links about anything and everything apart from the subject they'd expected. Gut feel tells me that those sites generate lots of clicks (I know this because I know how much people running such sites pay for domain names) but as near as doesn't matter no sales at all (that is a guess - I would love to hear from anyone who knows for sure).

Barry S Mills
November 06, 2006 | url

Shaun Johnston said:

  I agree, click fraud is just one more source of noise. One way to deal with it is the same as any other source of noise: finding a way to identify signals and distinguishing between sources of visits on the ratio of signal to noise.

I am part of the problem since I run a directory for which I charge PPC. Even after eliminating any non-Mozilla browsers and any browser with "bot" in the name, some long sessions have many intervals between page-accesses of less than 5 seconds. Clearly not a human being surfing. But this is a source of noise that I let through and charge for.

My standard test for "signal" is average number of pages visited, or ratio of one-page visits to 3-to-8-page visits, something like that. It should be a standard part of web metrics services to rank sources by such tests, maybe also for number of less-than-5-seconds intervals between page-views. This would be a rough ROI test. Maybe some services already do, the one I use doesn't (Web Ceo). Click fraud could be expected to be all single-visits to home pages. Of course, as fraud evolves to mimic signals, web metrics tests would have to evolve too. No problem.

When I analyze visits to customer's site I host, visits from my directory come out as averagely qualified, a little better than Google, not quite as good as MSN, better than most national directories, as good as most other local directories. Then I publish these figures on my directory. Of course, I'm not the best person to do this, since I'm one of the directories being tested, but at least that data gets published.

Cheers,
Shaun Johnston
November 06, 2006 | url

Brad Waller said:

  Actually, Google is very active in detecting this type of activity. When we first signed up for AdSense years ago, I saw an ad for a service I was interested in on our site so I clicked. I got an email within a day informing me that they had detected this click and that I was not to click on any ad on my site, even if I was interested.

I'm sure they use cookies and IP addresses to look at where the clicks are coming from, and the onky ones who can get away with fraud are the ones using sophisticated programs to spoof IP addresses and click at random intervals while also not having a steady CTR for the site. I have seen Google catch fraudulent clicks from sites that looked fine and did not show signs of fraud that I could detect (and we saw their site stats too). These sites had consistent (within a tenth of a percent) CTR rates, but the traffic seemingly came from all over and behaved pretty normally.

As for Google fraud, I have done spot checks and I have never found a day where they charged me for more clicks than I got. Just yesterday I got 10% of my clicks for free. But looking at one keyword for October, I was charged for 391 clicks, but my stats show I got 681 visits. Another had charges for 427 clicks and I recorded 578 visits. This implies Google is not charging for all clicks, and I am assuming that these are the ones being screened for fraud.

Brad Waller
November 06, 2006 | url

Write comment

security image
Write the displayed characters


busy
 
[ Back ]