Wednesday, March 08, 2006
For some time, we’ve provided information for advertisers about invalid clicks on the AdWords site here. We thought the notice of today’s invalid click settlement might prompt some additional questions about this issue, so we had Shuman Ghosemajumder, Business Product Manager for Trust & Safety, address them.
What is click fraud? I often hear the term “invalid clicks,” too. What’s the difference -- or are they the same?
The term "fraud" implies deliberate deception. Our aim in fighting invalid clicks is broader and includes clicks that we suspect may have been deceptive or malicious, as well as clicks that we deem invalid for other reasons, such as accidental double clicking on an ad. The usage of the word "fraud" in this context has caused a great deal of confusion, as it's practically impossible to "prove" that an impression or click was caused by deliberate deception. Our servers can accurately count clicks on ads, but we cannot know what the intent of a clicking user was when they made that click. When we identify a click as invalid, it simply means a click we won't charge for, in order to deliver the best ROI to advertisers.
Why not say more about the specific methods used to identify invalid clicks?
There are many things we do to detect invalid clicks, including looking at duplicate IP addresses, user session information, network information, geo-targeting and browser information. These are all important signals for detecting invalid clicks.
The technology we use to detect invalid clicks is highly sophisticated and was developed by some of the world's leading experts -- PhDs in artificial intelligence, machine learning, and statistics. We’re reluctant to share more about our technology and methods, however, because doing so would make it easier for fraudsters to try to defeat our systems.
How big a problem is invalid click activity?
We take it very seriously and have devoted significant resources and some of our best talent to this. By far, most of the invalid clicks we see are detected and discarded by our automatic filters even before they reach advertisers accounts. If an advertiser is monitoring click activity, these automatically filtered clicks may show up in an advertiser's logs, but not in their bills. When invalid clicks are detected after an advertiser is charged, we reimburse for them. Because of our detection efforts, losses to advertisers from invalid clicks are very small.
But don't advertisers report invalid clicks to Google which weren't detected in advance?
We do receive reports from advertisers and we look at them very closely. When we believe those clicks are invalid, we reimburse advertisers for them. Some invalid clicks do make it through our filters, but we believe the amount is very small.
Also, we often find that the clicks are legitimate, but from unexpected sources such as broader targets the advertiser has set up for their campaign. And, as I mentioned earlier, some of the invalid clicks advertisers see in logs are clicks we've already caught, discarded and not charged. Most investigations we conduct concerning invalid clicks are cleared up with the advertiser after explaining the source of the traffic increase or showing them that the clicks were never charged.
Does Google have an incentive to allow some amount of fraud because it means more revenue?
Actually, it is the opposite of that. We have much more of an incentive to do a better job of handling invalid clicks than our competitors -- and we believe we do. Fighting invalid clicks aggressively is in Google's best interest and essential for us to maintain a viable business. In addition, we offer free tools to advertisers so they can monitor their return on investment -- which is a helpful way to determine whether too many clicks coming through are not resulting in sales. Those free tools help advertisers manage to a bottom line value of their ads.
Some people suggest that click fraud may account for as much as 30 percent of traffic -- what do you say to that?
We believe the methodology behind that particular estimate is flawed -- and that many who have cited the figure have done so irresponsibly by using it differently than it is characterized in the report.
Here's a link to a .pdf version of the study where that figure originated. We encourage you to read the report and evaluate it yourself. Some things we think you will see that undercut those who use this estimate carelessly:
Even the report does not say that click fraud is 30% of all clicks. What it does say is that of three ad campaigns (only three ad campaigns were examined in this study), evaluated over a ten day period, one had questionable clicks of 8%, another 10% and a third 30%.
So, the 30% figure comes from analysis of a *single* ad campaign, not a study of many. This means that the figure of 30% that is used to characterize click fraud for the whole search and advertising industry comes from the analysis of *one* ad campaign looked at for ten days. Even in that campaign, it is not clear what methodology they used to determine which clicks are "bad" and it is possible that they marked legitimate clicks as "duplicate."
Moreover, the study does not indicate whether the advertiser was actually charged for any of the clicks, only that the traffic analysis suggested that the clicks may have been invalid. As I mentioned above, it’s very possible that clicks recorded in an advertiser's logs have already been caught by Google’s detection systems and not charged. (I should also note that when invalid clicks are detected and discarded before they are charged to an advertiser they are also not recorded as revenue.)
When considering the validity of this exaggerated 30% figure, you should also consider who is most aggressively using it: it is those who have the most to gain from hyping the problem. Those who are throwing around this figure are doing so as part of their marketing efforts to sell products they claim detect click fraud. The more they can convince others that click fraud is a problem, the more they hope to see increased sales. In other words, these companies have a huge financial incentive to make people believe invalid clicks are a larger problem than they really are.
What else on this topic should we be aware of?
Whether it’s online or offline, advertising should be about ROI and results. It is in our interest to serve the interests of our advertisers, which means delivering superior ROI. Our efforts to combat invalid clicks are an essential part of that.
What we have seen so far is that advertisers continually increase the amount of money they spend, which suggests that they are pleased with the return delivered by their ads. We encourage advertisers to track ROI and contact us when they see something that doesn't appear to add up. If you see suspicious activity on your AdWords account, please contact our click quality team.