Research conducted by the US-based Marketing Experiments Journal and Clicks2Customers, together with South African specialist Incubeta, has claimed that fraud on the pay-per-click ads used by Google, Yahoo and many others - and which pepper many blogs - is reaching close to 30 per cent on some campaigns. The trouble is, the higher levels of click fraud tend to be for the ads that attract the higher payment rates every time they get clicked. According to the research, one user of pay-per-click advertising lost out to the tune of more than $15 000 over the period of just 10 days. Joe Holcomb, formerly of
Blowfish Blowsearch, questioned whether the techniques used could identify click fraud but has indicated similar percentages in his own research.
When things such as adwords arrived on the scene, advertisers must have thought: "At last, we have a way of measuring response rates and paying only for ads that work". Now, they are probably wondering whether that many living, breathing human beings actually do the clicking, or whether all those pages they serve at the destinations of the clickthroughs are simply getting hoovered up by script-driven bots.
The two companies behind the report acknowledge they need to do more work to truly identify fraudulent clicks with any certainty. But there were some myths they could help lay to rest with their research. One test, using a bogus Adsense campaign based around the words "duarf kcilc" to avoid disrupting other campaigns pretty much settled the issue of whether a competitor down the road could disrupt someone else's campaign. They could, but they would have to be a lot more sophisticated than just sitting at their desk and hammering on your ad all day. They would have to cloak their identity.
The second part of the study used a tool from Incubeta to study the IP logs from three Adwords users. This is the study that picked up the 29.5% suspected fraudulent clicks. This was for a campaign in the legal field that cost up to $2 per click. The cost to that advertiser for those clicks was calculated to be $15 394.50. Cheaper campaigns netted fewer fraudulent clicks: less than 10 per cent apiece for the two sub-30 cent campaigns tracked by Incubeta.
There is some movement in the middle market over pay-per-click ads, it seems. Holcomb claims that Blowfish decided to take the issue very seriously as it could prove a competitive advantage. However, I can't help feeling that the pay-per-click field is still heading for more trouble. Click fraud may not be easy to automate - the pay-per-click engines are a moving target when it comes to detection. But, assuming that the fraudsters are profiting directly from the operation by diverting money from campaigns to sites they operate, it looks like this one might just turn out like spam. For any protection mechanism that can be automated, an attack can also be automated.
For the moment, the fraudsters are perhaps being too greedy for their own good, siphoning off money at a rate that is readily detectable if not easily demonstrable to the pay-per-click engine companies. As the detection mechanisms improve, I can't help feeling they will just improve their cloaking measures, widen their scope and throttle back the click bots they control to the point where they make enough not to be seen easily, but continue to raise the cost of using pay-per-click ads.
There might be a silver lining to this for traditional publishers, especially in the badly damaged controlled-circulation trade sector, who have been missing out because advertisers have drifted off to use more measurable means of promotion. OK, pay-per-click remains a tiny part of the overall ad market, but it has been growing fast and has made advertisers wonder what they can track. If their measures are no longer working, advertisers and their agencies will have to go back to the old-fashioned methods of doing proper market research before and after campaigns, and taking notice of what people read and watch rather than what gets clicked. But I'm getting dangerously into wishful thinking territory. And that rarely works out to be the way things go.