Exclusive: Google’s Click Fraud Rate is Less than 2%
UPDATE: Ghosemajumder has clarified that my assumptions of less than 2% should be based on “invalid clicks”, which means the actual number is more likely just a fraction of one percent!
Back in November, Google’s business product manager for trust and safety, Shuman Ghosemajumder, declared that
click fraud invalid clicks at Google was “on average is in the single digits, quarter over quarter.” I recently sat down with Ghosemajumder to discuss exactly how large of a problem click fraud is at Google. Whether intentional or not, he gave me access to information and shared data never before seen outside of the walls of the Googleplex.
For those of you short on patience or time, here’s the revealing information gleaned from my conversation with Ghosemajumder:
The click fraud rate – as discovered by most AdWords advertisers – is on average, less than 2% of all clicks through Google’s system.
So why did Google break its silence to MarketingPilgrim.com on this touchy subject? How did we manage to get the real click fraud number out of Ghosemajumder? Read on.
Proof Google’s Click Fraud is Less than 2%
Ghosemajumder sat down with me during the recent Search Engine Strategies conference and went thru a PowerPoint presentation that he confirmed was previously shown only to employees of the world’s largest search engine. He explains how Google has a four-stage process in identifying and filtering what it calls “invalid clicks”. Google’s definition of invalid clicks includes non-fraudulent clicks (such as a visitor genuinely clicking an AdWords ad more than once) and “click fraud” (those clicks that are obviously not legitimate).
Ghosemajumder shared with me the following diagram* which explains the makeup of Google’s AdWords clicks.
As you can see, out of all clicks on the AdWords network, Google is able to filter out the majority of invalid clicks before reports a served to the advertiser. You’ll notice a small amount of “click fraud” that falls outside of Google’s invalid clicks filters. That small amount represents the click fraud often discovered by advertisers, through the AdWords console, and click fraud detection companies.
Here’s where it gets interesting. The blue “invalid clicks” circle you see above, represents the “single digits” number Ghosemajumder previously shared in November. By deduction, we can already estimate the actual percentage of click fraud that escapes Google’s rigorous filters and detection systems is barely 1-2%. However, Ghosemajumder then proceeded to share another graphic that more clearly describes Google’s filters and appears to confirm that undetected click fraud is less than 2%.
Google’s Four Layers of Click Fraud Filters
Google utilizes four layers of click fraud detection. The first layer is purely automatic and is used to filter clicks from both “search” and AdSense partners (contextual ads). This filter is able to detect invalid clicks in real-time, with the goal of removing them before they ever show up in the AdWords console.
The second and third layers are aimed at filtering only AdSense clicks. The second layer is what Google calls its “flagging system” and is an automatic process to remove invalid clicks from the AdWords system. The third layer of filtering is a “manual review” process with more than two dozen Google employees manually reviewing and removing any suspicious clicks.
Google’s goal is to have the first three layers of filtering identify 100% of all invalid and fraudulent clicks. Those clicks that manage to escape Google’s filters are what causes many advertisers to raise concerns and has spawned the growth of many so-called click fraud detection companies. The fourth layer of click fraud detection falls to these advertisers and detection companies and is what Google calls “requested investigations”.
Here’s a representation* of the slide shared by Ghosemajumder.
Ghosemajumder confirmed that the blocks representing each layer of filtering were to scale and that the internal version of the graphic includes the actual percentages. After studying the graphic, I asked Ghosemajumder to comment on my perception of the numbers each block represented.
If Ghosemajumder had already confirmed the total amount of invalid clicks represented a number in the “single digits”, which I assumed would be around the 8-9% mark, the amount of user-detected “requested investigations” looked to be a number in the 1-2% mark, at most. While Ghosemajumder would not reveal the exact number, he did confirm that my estimate was on track, “if not less”. Wow! A user-detected click fraud rate of less than 2% is a world away from the 20% number given by some click fraud detection companies.
This is an amazing revelation and clearly shows that Google is getting tired of speculation and rumor filling the void left by the lack of transparency Google has with regards to their click fraud numbers. Ghosemajumder confirmed that internal discussion are ongoing as to whether Google should reveal exact numbers and put an end to the inflated estimates often quoted, when discussing click fraud. The biggest reason Google has for being hesitant about revealing the exact numbers, is fear that Yahoo and Microsoft will be able to leverage the numbers to deduce more information about AdWords.
Not All Click Fraud is Click Fraud
So how does Google explain the 15-30% click fraud numbers that have surfaced over the past several months? Ghosemajumder described how they monitor hundreds and hundreds of different signals in its efforts to detect click fraud. These identifiers are a strictly guarded secret with only the “click quality” teams having access to the information. Many advertisers – and click fraud detection companies – are looking at the wrong signals and often class valid clicks as fraudulent, or request refunds for clicks that Google had already discounted.
He gave a number of examples of how valid clicks could be misidentified. One included an advertiser seeing many clicks from the same IP address and surmising they must be fraudulent. Ghosemajumder described how many of these types of clicks are indeed valid, with so many people using corporate computers sharing the same IP, or ISPs assigning the same IP to more than one customer. Another common example of how click-fraud detection companies get it wrong is when counting reloads of an advertisers landing page. He described how a customer could click through to the landing page, view a product page, and then hit their “back” button, returning to the same landing page. Without correct tagging, Ghosemajumder said that one click and five page re-loads could easily be mislabeled as six clicks from the same visitor.
Google Becoming More Transparent
There is no doubt that we’ll see Google becoming more and more transparent in its efforts to share click fraud information. They have no reason to keep quiet for much longer, as Ghosemajumder explained. Google is already filtering more than 98% of invalid clicks, before they show-up in the AdWords console, their goal is to filter 100% and suggestions they are not doing enough are misguiding in their eyes.
* The graphics shown above were not supplied by Google, but we believe they are accurate representations of what was shared with us.