More On ClickFraud

Wired News throws another log on the click fraud fire here. This is the story of BlowSearch, a meta search engine which has discovered that click fraud is rampant and is creating a tool to deal with it. Let's pull back to 50,000 feet here for a second. Advertisers…

Wired News throws another log on the click fraud fire here. This is the story of BlowSearch, a meta search engine which has discovered that click fraud is rampant and is creating a tool to deal with it.

Let’s pull back to 50,000 feet here for a second. Advertisers and SEMs are saying that clickfraud, in aggregate, accounts for 15-20% of all spending (I had two major agencies confirm this figure to me again this week). Why do they say this? Because they are on the front lines and see the non-performing clicks, the domain parking lots filled with AdSense ads, the odd patterns of clicks in the middle of the night from what have to be robots or armies of third-world paid-to-clickers. It’s hard to substantiate these claims, but let’s think about this for a minute: 20%! That’s more than a billion dollars in fraud in an industry growing faster than anything else in the media business.

Now, it’s in the advertisers’ best interest to claim this kind of fraud, for the more they can say some clicks are fraudulent, the more refunds they get from Google et al, and the higher their “real” conversion rates are. So, caveat number one, take advertisers claims with a grain of salt.

Now, on the other end of the spectrum, folks at both Yahoo and Google have sworn up and down to me that click fraud is not a very big deal, that the level of fraud is, by their estimation, in the low single digits. Why would they say this? Why, because they see everything, of course, and are deeply motivated to not only find fraud, but to ferret it out. After all, if their networks are seen as scamming the advertiser, everyone loses.

Except, let’s examine the motivations on this side of the equation. Clearly, it’s in the engines’ best interests to say that clickfraud is not a major issue, and that they have it under control. It’s also in their best interests, at least in the short term, to allow a bit of fraud to happen. After all, it’s just a tax on what is still an incredibly efficient method of marketing. Clearly advertisers are not stopping their spending because of what they see as 20% clickfraud. Search marketing is so efficient that a 20% “clickfraud markup” is justified in the advertiser mind. The market will bear it, so why not let it happen?

The truth is probably, as always, somewhere in the middle. There is probably more fraud happening than the engines will admit, or find, and less than the advertisers claim. But it’s maddening that we don’t have a way of knowing for sure, because Google and Yahoo will not engage with the advertisers in a high level dialogue of trust, one where the two sides can compare notes. I hope that changes soon.

UDPATE: After some feedback, I’ve realized that my post may have led some to think that I believe that the major engines are in fact actively allowing fraud to occur. I don’t think that is the case. In fact, it’s in their long term interest to eradicate it, which is why I think it’d be great if they worked with major agencies/advertisers to combat it in a unified fashion. I sense that for every ten smart engineers Yahoo and Google have working on detecting and beating click fraud, there are 10,000 folks trying to game the system. Unless both advertisers and engines work together in a trusting relationship to compare notes and figure out how to fix this, the Force Of Many will outpace the engine’s best efforts (and the advertisers will always claim that there is more fraud than the engine’s claim there is).

13 thoughts on “More On ClickFraud”

  1. John,

    In an otherwise fair piece, you repeat the canard that the SEs “see everything”.

    That is false. The SEs see the clicks come through and get re-directed to the advertisers. They have no visibility into what happens to the click once it gets to the advertiser’s site.

    This is an important distinction as the more the SE can understand about post-click activity, the better it can tailor its protection systems to detect fraudulent activity.

  2. Isn’t it possible both that some ad agencies are seeing 20% fraud rates and that the overall fraud rate is in the low single digits? Perhaps some agencies are getting hit worse than others. The ones seeing high fraud rates will be the ones making the most noise about it.

    Also, there is bias on both sides. You mention that Google wants you to think that click fraud is low so you’ll keep spending, but agencies may want you to think it is high so that you’ll hire a pro who can help you deal with it. I agree that the truth is probably somewhere in the middle.

  3. There is also a question of what is click fraud. For example is a visitor clicking on ad, backing out and then deciding that was the right site and clicking on the ad again committing click fraud? Technically no, it was not malicious, and he had no other way to get back to the site, but it’s two clicks by the same visitor so should the advertiser have to pay twice?

    There are a few other similar examples in the grey area and I think if those are clearly classified as click fraud or not, the estimates of how much click fraud actually exists, will be much closer between advertisers and SEs. However, if they will ever agree on one number …

  4. John,

    I used to work for a search engine and this was a problem we thought about a lot. Here’s a way to approximate how much click fraud there is.

    Go back and look at U.S. search traffic levels three years ago. Look at the organic growth rate of search traffic for 1999-2004 or so. Carry that same growth rate forward on top of 2002 traffic levels and you get roughly 70% of the reported traffic levels of today. So unless everybody is searching far more than they were a couple years ago (unlikely) roughly 20% of all traffic is fake. Now do the same exercise with paid clicks only (based on reports from public search engines + estimates from MM). You’ll notice that traffic levels are through the roof, doubling or more every year. Even when you account for the rise of content search as a contributor, you still get a figure closer to 50% of all traffic being fake than 20%. You’ll especially notice that exponential paid click growth has occurred among the smaller search engines, not just Google and Yahoo. I don’t know about you, but I haven’t been doing 10x more product searches on Search Froggy than I used to, so this seems suspicious. I’d encourage you to do the exercise and see what you come up with.

  5. Almost every form of advertising has some kind of independent auditing. But the SEM/SEO world is so secretive on both the buy and sell side, that no one is even talking about independent auditing today.

    Furthermore, my experience is that most advertisers are satisfied with the ROI today, so the 5-10% that could be fraudulent is in the noise – cost of doing business.

    As inventory gets more expensive and ROAS goes lower, you have to imagine that there will be more pressure on Google and Yahoo to share info…

    Until then… plausible denialability…

  6. sure, that may be 20% for the entire marketplace but what if you are a small individual on a small budget in a competitive keyword and it makes up 80% of your online ad budget. hurt’s don’t it.

  7. I work in an online advertising agency, and obviously follow this extremely closely. What I’ve noticed is that the definition of clickfraud that you use is probably the largest determining factor in how much clickfraud an individual or firm thinks is happening.

    The three most widely used definitions are:

    a. any click where the person clicking did not end up buying anything at the store.

    b. any click where the person clicking had no *intention* of ever buying anything at the store.

    c. any click where a competitor intentionally clicked on an ad in order to spend their competitors’ money

    d. any click where an adsense publisher has someone click on his/her ad in order to make money for the publisher.

    Nearly every study defines click fraud differently. From doing this everyday for a very long time and with *very* high stakes for our clients, I’ve become a believer in the lower Google/Overture produced numbers.

    C and D are extremely easy for the major search engines to stop. We’ve been what we thought was a victim to this many times, but every time we did real research on it, it came out that there was no fraud.

    D, itself, is shaky ground as a measurement, too, because of how powerful even subtle references to visitors can increase click-through-rate. If a blogger mentions to his/her visitors to, “I’ve added text links on the right that are sponsored by Google,” then the blogger will possibly raise click-through-rates by whole percentages. A percent can mean a great deal of the page gets 10,000 hits per day. That 1% can be an extra 100 clicks. 100 clicks on a 30cent word is 30$ more everday. Over the course of a year, that adds up ($10,950). Is that unethical? Not to me because the advertiser still got what he/she paid for- a person to visit their site/store. If Best Buy could pay only 30 cents to drop someone in the middle of their brick and mortar store for five seconds before that person decided to leave, then Best Buy would put 10 million dollars into that form of advertising, which is what they do.

    A and B are just ridiculous measurements because your job as a web designer/content creator is to do something well enough that people want to buy. That’s why the word, “conversion rate,” exists.

    The biggest “fraud” are companies who are too lazy to do anything but complain about what is essentially the fact that they haven’t put enough effort into optimizing their store’s conversion rate or writing ads and choosing keywords well enough that they are getting buyers to click instead of browsers. Maybe I’m wrong, but I can’t imagine a world in which we could cut our already low advertising expenditures by 20%-50% and get the exact same profits as we do now for our clients.

  8. To “observer,”

    Google can see after the click insofar as a good number of their advertisers have Google Conversion Tracker installed. In fact Google acknowledges that this data in part helped them with the formula for “smart pricing” on content clicks. They could, and probably do, also use this data to detect patterns of fraud.

  9. I’m not sure debating the scale or scope of the problem is relevant any longer. The industry needs some degree of surety to be reached within a year or so. As with other security and fraud problems on the ‘Net, the path to a solution starts with creating an open source interface at which multiple parties can plug in tools to attack the problem.

    At Feedster, we say “Run all your ads, including AdSense, through a copy of phpadsnew — the leading opensource ad server — whether you host it, we host it, or your ISP does.” If we can get phpadsnew over the hump to being a self-sustaining open source movement, the clickfraud services will have a fighting chance of helping us.”

    Six guys in 6 cubicles at each of Google, Yahoo, and MSN won’t solve the problem.

  10. John – The cost per lead (whether it be a sale or a demo download) is so substantially lower when generated via search that the elimination of some fraud would all be upside; a good thing for my clients. But, I don’t want to see a fraud-elimination system put in place that is half baked.

    As an agency I feel a responsibility to my clients to watch out for clickfraud. If you see spikes in traffic that don’t match up against other online or offline marketing campaigns then contact your SE sales rep. The engines are really, really anxious to give money back if an advertiser even hints at clickfraud.

  11. ok, this is going to get me yelled at in here, but why not: Does it really matter? I agree that click fraud is bad for the industry as a whole (especially image wise), but really, avgerege amounts of click fraud get priced in by the market. If you look at the economics of it, click fraud will not be a major factor unless the cost to acquire a purchase = marginal cost to advertise. If click fraud is driven way down then the key word price will go up proportionally (and total spend will remain the same, or increase until advertisment cost = profit from advertising). The market bid driven pricing model will keep this inline.

    That said, the major problem of click fraud will be targeting one company, rather then across an industry or grouping of key words. And if one companies clicks are out of sorts then that is noticable.

    I know there are a lot of other issues behind click fraud that are preventing a perfectly effecient market but the rate at which information is gathered and distributed within this system will price in click fraud until the engineers can firure it out, if it is possible to fix.

  12. Click Fraud is an intersting topic, one which will cost you an estimated 20%. This is an interesting article with valuable information. What we and many other webmasters are starting to do is invest some of our marketing dollars into a click fraud prevention/protection software. If you are looking for the best one for your company i recommend you take a look at:

    Mike Baker

Leave a Reply

Your email address will not be published. Required fields are marked *