Click Fraud Buzz Will Grind On

Click Fraud buzz continues in The Marketing Leaders article by Daniel Jupp. The article takes a look at the landscape of click fraud, but it is still not quite the right industry approach that will be needed for advertisers and agencies to control click fraud.

I’ve been running silent recently, with a lot of work for clients and a series of experiments on geotargeting. Unlikely as it may seem at first glance, geotargeting and click fraud are related. Daniel Jupp’s article does bring this out, talking about armies of Indian and Chinese workers who click on adverts, and botnets that defraud advertisers.

So, lets look at Daniels’ article and find the gold and the dross and, more importantly, the unmentioned bits.

Defining Click Fraud

The article does highlight the basic issue - what does an advertiser, not a search engine company, identify as fraudulent activity? Unfortunately the article does focus on measures that the search engines find useful for their business purposes. I think that much of what I categorise as fraudulent, is activity that benefits the search engine, rather than the advertiser. There is a fundamentally different business goal for search engines and for advertisers.

The article hits at, but doesn’t state out straight, that search engines want to offer as many clicks as possible, but advertisers are (mostly) looking for the highest revenue potential. This means that search engines don’t mind passing clicks to advertisers that won’t convert, just so long as no-one can take them to court for it, and the dilution of lower quality clicks doesn’t provoke advertiser defection. This leaves a gap, where search engines are prepared to offer low quality clciks that cost advertisers money, from sites that advertisers wouldn’t want, if they could control the sources better. It may be that’s just my reading - the views represented here are, of course, mine, and not those of Daniel Jupp, I’m interpreting what he’s said into my words.

Daniels’ article also discusses the ways that click fraudsters can work, and takes a quick look at anti-competitive clicking.

So, all the right sorts of elements in the article, but I don’t think it pushes far enough or hard enough.

• Whose definition of click fraud? The loss to advertisers, or the loss to the search engine company?
• Lots of ways to get clicks that reduce the ROI - meaning subtle techniques may be needed to work it out
• Use of automation, and low cost human clickers

It should be noted though, that a badly set up paid search campaign also yields a low return on investment. That shouldn’t be called click fraud - it is incompetence by the advertiser, and can be solved by adding some training or expertise. I expect that, as the industry matures and the search engines disintermediate the experts at the advertising agencies, that user incompetence will play an increasing role in allegations of fraud. It simply isn’t search engine or click fraud if you use the wrong keywords and a bad landing page. Its solvable, but probably not historically - and those burned by the experience are going to take a while to return to the industry.

Alien Clickers

Quite a lot of the article is spent on addressing human clickers in China and India. This is, I feel, somewhat of a red herring. Geotargets are important. Some of the lower tier search engines don’t let you choose which countries you target, but that’s clearly not a rational or useful decision - a standard marketing acronym is PESTEL (Political, Economic, Social, Technological, Environmental and Legal), all of which pertains to a country and to even smaller units. Where you cross significant boundaries with a message, you risk a reduction in the value of that message.

For example, being British, I can go out for the evening, get pissed and have a good time. Those of you in the USA will wonder why I could have a good time when I’ve been annoyed. Language meanings are geographically relevant.

If you suffer from low converting traffic from India and China, just don’t target those areas, or pay for a lower position in those countries. With Google, MSN and Overture/Yahoo, that’s easy. It’s not so easy with some of the lower tier search engines, where you get whatever traffic they give, from wherever it comes. Your only recourse with the lower tiers is to examine your traffic and complain when you find odd patterns of use.

There are ways for non-local users to see your AdWords adverts, but it is a bit harder than just typing “google.co.jp” to see, for example, Japanese adverts, and both the IP address (via IP location services) and any referer_info will give clues that this has been done. I can think of ways to disguise even that, but then I can also think of ways to check on that… it’s an arms race.

I’m pretty sure that geographically isolatable click fraud can be controlled, by advertisers and agencies - so long as you don’t want to target those territories. If you have to target those (and other) territories, known for their higher than usual click volumes and lower than usual purchases, then you need to get subtle and use all the information that you can squeeze out of visitor behaviour.

The Subtle Knife

Despite the apparent intention to define click fraud, the article once again falls prey to allowing the search engines to define what it is. The article focuses on what the search engines report as identified click fraud. Frankly, it’s not the search engines that are failing, but the agencies and advertising industry.

Where is the external oversight and control over Google’s use of domain parks, as a result page for keyword search? A domain park is clearly not a search page. It doesn’t look even remotely like the process shown on the AdWords sign in page, which is what advertisers think they are buying.

Where is the external targeting control for keyword match? Google offers a very sophisticated system that will probe to get the highest CTR, but advertisers are not buying clicks - at least, the sophisticated ones aren’t. The smart advertiser is trying to buy interested visitors. If you just want clicks, there’s plenty of techniques to offer undifferentiated and largely unlikely to convert visitors.

That means that it is actually important that advertisers and their agencies complain when they find, for example, a keyword about financial services, with a financial services advert and a financial services landing page, being matched to a search for a “hair loss tonic” or a “weight loss clinic”. That’s clearly not an appropriate match, but very few agencies check for this. That sort of mismatch is unusual on Google, but we’ve found exactly this example on lower tier search engines.

How do I know that the agencies don’t check? Because they don’t use analytic systems that offer the information. You need either a well designed web analytics package (most aren’t able to do this properly) or to use web server log file analysis to compare the keyword with the actual search.

This verification of search match quality isn’t made easier by the search engines. While there are some pieces of information passed in the destination URL, it may miss out the actual search string. Kudos to Overture - their automated advert tagging offers both the user search and the keyword, so you can do a conceptual match. Google, however, persistently recommends users should look at the referer_info (sic - it really is called that in the standards documentation) field, which often isn’t passed. In one account I was looking at, more than a third of the dubious clicks had no referer_info. That’s partly because one of our signatures for a suspicious query is a missing referer_info field, especially on a non-AOL IP address.

There’s undoubtedly more that the search engines can and should do to demonstrate that they are trustworthy. After all, this year, I’m expecting that we’ll be handling well into seven figures of spend on behalf of clients. That’s a minute fraction of the revenue stream of the search engines… they are being trusted with a lot of cash, and very poor systems of control and oversight. And the advertising industry isn’t grasping that they need to be the ones to hold the search engines to account - because if they don’t, then advertisers may defect to other, perceived safer, media.

Where else can you get help? The web analytics vendors do not have reports that check on whether clicks came from the origins you bought. The search engines don’t pass completely harmless (to them, if they are behaving properly) information about the search, which would help advertisers. There are click fraud detection services, but every one that I’ve looked at (a handful - a small sample of the dozens that a casual search will show), does not configure the destination URL properly, to allow for disambiguation of double clicks; nor do they use Google’s own daily data on invalid click rates; nor do they recommend ways to configure the account so as to maximise data collection in such a way that future analyses could be easier.

There certainly are things that you can do to add more tracking data to destination URLs on Google. It is possible to develop software to do a conceptual match on actual searches versus the offered keyword, to see whether the search engines are extending to rational broad matches or just optimising revenue. It is very difficult, without more co-operation from search engines, to identify when they use, for example, content matching sites in nominally keyword search advertising.

Optimising Revenue

The intent of search engines is really at odds with those of advertisers. Search engines want to maximise their revenue, which means offering opportunities for adverts to be clicked. Advertisers want to find visitors who are interested in their products and may buy. Not a casual, passing interest, but a 1 in a 10, or even a 1 in 3 chance of actually buying something.

That the search engines have ther own interests at heart has been obvious. Look at Google’s “Budget Optimiser”. This is a tool designed to make sure that Google extracts every penny from the advertiser. Even if the account has AdWords Conversion Tracking, the optimisation is not for keywords that sell, but for keywords that generate clicks. Why would any sophisticated advertiser care about the volume of clicks? It’s the volume and value of conversions that counts. So even when Google have the data to optimise on behalf of a user, they choose to optimise their own revenue.

Good deeds go unpunished

I think it’s great that Daniel has spoken out about click fraud and the needs of the advertiser. I think it’s a shame that he didn’t push that one step further into probing whether the advertising industry should just be using the search engine’s definition of click fraud, and whether there is additional information that the search engines should publish, at least to the advertiser that has bought a click, to assure them that their money is being properly spent in the hopes and intention of delivering a valid interested visitor, rather than just optimising the revenue stream.

As the title of this article suggests, so long as the advertising industry accepts the search engine definitions of fraud, and fails to hold the search engines accountable for mismatching keywords, for geotargeting abuse, for using “keyword search” to host domain parks, and for “budget optimisation”, there will continue to be complaints that the medium can offer the opportunity for unexpectedly low returns - click fraud.

Although I’ve been too busy to be part of the debate, I do see that Rich@Apogee has found a client who gets a significant fraction of their clicks, on a supposedly search network, from content matched sites. This seems to me to be clearly beyond the line in the sand that search engines should not cross, without offering additional information and control to advertisers.

Advertisers have to wake up to the idea that just because these systems are technologically sophisticated, they can’t forgo the inspection that has been required in, for example, direct mail and magazine subscription circulation. These types of measure are currently lacking in the industry, and it will be up to advertisers to organise - because if they leave it to the search engines, the definitions of click fraud and acceptable behaviour will be left to organisations who fundamentally and legally owe a responsibility to shareholders, not to advertisers. That means making the maximum profit for the SE, not for the advertiser.

This mismatch of advertiser needs and search engine needs is, I think, at the heart of why you’ll hear complaints about click fraud. And it is why, when the search engines investigate, they’ll find no problem or rare problems. Nonetheless, as an agency working on behalf of advertisers, I can clearly identify some activities that search engines do, that I know will reduce my clients ROI. Is that click fraud? Arguably, I’d say it is - that’s the core of the issue for click fraud. Do you, can you, trust that every search engine, acts in the best interests of the advertiser? If they don’t, what mechanisms do they offer to assure advertisers that every effort is made to offer a high quality click stream?

Regrettably, the SE’s seem to have decided that the right course to handle this is a stubborn silence and an assurance that the SE’s protect themselves (not the advertiser) from their own definition of click fraud. It’s not enough. Not with a multi-billion dollar business.