Google Shares More Click Fraud Data

Invalid clicks on Google AdWords ads have consistently remained under the 10-percent mark, and are generally in low single-digits, Google revealed today. In addition, the amount of invalid clicks that are not proactively detected and are caught by advertisers is less than 0.02 percent, according to Google’s Shuman Ghosemajumder.

This revelation is the latest in a series of moves designed to offer more transparency to advertisers, said Ghosemajumder, Google’s business product manager for trust & safety. Google has come under fire from lawsuits, click fraud reporting firms, and advertisers in recent months, all looking for more granular details on the level of invalid clicks occurring in the AdWords system.

To understand the significance of the “single-digit” and 0.02-percent figures Google revealed, it’s important to look at how Google detects invalid clicks, including click fraud. Two proactive measures detect and remove clicks from an advertiser’s account before they are billed: real-time filtering and offline analysis by Google’s click quality team.

In the first case, proactive filters catch invalid clicks in real time. This method keeps the clicks from being billed or showing up in advertisers’ metrics. This is the ideal situation, since these clicks in effect never occur, in advertisers’ eyes. Not all of the clicks removed here are fraudulent; some are questionable, or have been deemed invalid by Google, such as the second click of a double-click. Filtered clicks make up a majority of the “less than 10 percent” figure that Ghosemajumder is talking about, a level which he says has been consistent since AdWords was launched.

As a result of removing these clicks, Google is purposely deciding to promote network quality over revenue, Ghosemajumder says.

“When we proactively remove these clicks, we forego a significant amount of revenue,’ he says, noting that every 1 percent of clicks removed would equate to $100 million in annual revenue for Google. “We’re providing hundreds of millions of dollars in proactive protection,” he says.

The second proactive detection method, offline analysis, is done by Google’s click quality team after the click has been recorded. These are the clicks that show up on advertisers’ statements as click quality adjustments. Clicks that are detected at this stage are still not billed to the advertiser, but they do show up in click metrics, so they affect ROI measurements.

Most of Google’s efforts here are focused on clicks from the AdSense network, since publisher partners have the most incentive to commit click fraud. Ghosemajumder did not share a specific breakdown of filtered clicks compared to ones that are found by Google’s team, but he notes that the team finds a “much smaller” proportion of invalid clicks than those that are pre-filtered.

Invalid clicks that are not detected proactively can still be dealt with when an advertiser reports apparent fraud to Google. The number of invalid clicks that are found through these investigations has remained below the 0.02-percent threshold since AdWords began, Ghosemajumder says.

“We’ve appealed to advertisers to over-report anything suspicious, and we still haven’t seen a spike,” he says.

Ghosemajumder admits that many advertisers have said they would like to see Google make the process of reporting easier. To satisfy them, Google is undertaking several more click fraud-related initiatives in coming months. Among these are IP Filtering capabilities for advertisers, enhanced invalid click reports, educational initiatives, and an improved reporting format. The first three initiatives are expected to roll out this month, while the improved reporting format will come later this year, he said.

More details on these efforts, and analysis of the moves can be bound in today’s SearchDay.