Google's approach to click fraud declared 'reasonable'

Posted on 24 Jul 2006 at 10:56

Google's approach to click fraud has been declared 'reasonable' by an independent expert. Dr. Alexander Tuzhilin, Professor of Information Systems at NYU, was asked to make a detailed examination of Google's detection methods and procedures and he declared the approaches to the problem made by Google's Click Quality team were 'reasonable'.

Click fraud occurs when a site owner in the AdSense programme or some rival service deliberately click on ads on their own pages to generate revenues from Google - although it is the advertiser who ultimately pays. Because of this, click fraud can also be used to damage a competitor by generating spurious clicks on an ad which have no chance of conversion to a sale.

Click fraud is becoming a serious problem as the contextual advertising programmes of the major online portals become more popular and more lucrative for successful advertisers.

According a recent report on click fraud the average click fraud rate was 14.1 per cent across the industry although Google fared a little better with around 12.8 per cent of clicks through its programmes estimated to be fraudulent.

The investigation was ordered as part of the settlement of the Lane's Gifts vs Google click fraud case that was heard earlier this year. To settle the class action suit Google agreed pay out up to $90 million in legal fees and compensation in the form of credits for further advertising to companies who believe they have been affected by click fraud.

The 47 page report reveals that the Google detection unit has four lines of defence for weeding out the invalid clicks. These are pre-filtering, online filtering, automated offline detection and manual offline detection, performed by real people, in that order. Dr Tuzhilin says that Google uses different detection methods in each of these stages. There are rule-based and anomaly-based approaches in the pre-filtering and the filtering stages, the combination of all these approaches in the automated offline detection stage, and the anomaly-based approach is undertaken by human experts.

Dr Tuzhilin interviewed both the engineering and spam control sections of the Click Quality unit and recounts the history of Google's efforts to tackle the problem. In conclusion he says 'that the invalid clicking problem at Google was 'under control' by the end of 2005.'

The report gives some of the clearest indications of Google's methods in its pages and will doubtless be read with interest by search engine marketers and click fraudsters alike.

Author: Steve Malone