Study: Half Of Ad Impressions, 95 Percent Of Clicks Fraudulent
- by Laurie Sullivan @lauriesullivan, September 17, 2009
Perhaps that's because data released Thursday in a study conducted in July reveals that more than half of ad impressions and 95% of clicks in online ad buys were fraudulent. Gluck compiled and analyzed the findings published by ad optimization company Mpire, Seattle, Wash., with help from its AdXpose technology.
Nearly all the fraudulent traffic was hidden behind numerous layers of nested I-Frames -- ad units pulling ad content from other sources that can hide URLs and in-view data. The test shows significant impression fraud and URL padding in standard run-of-network (RON) online ad buys.
"There's an enormous amount of budgetary and impression waste taking place because of click fraud," Gluck says. "Ad networks, which have traditionally been seen as suspect, need to eliminate click fraud if they want to be taken seriously by advertisers."
Many of the sites in these exchanges use multiple layers of I-frames, which further complicated efforts to track campaigns. Sites are able to hide fraudulent traffic behind numerous layers of nested I-frames, leaving advertisers blind to in-view data, according to the study.
Mpire believes the default trafficking behavior of many RON buys is to include fraudulent and well-known botted sites. Gluck writes in the study that not all marketplace or exchange traffic is bad, but rather simply includes nefarious inventory that ad networks could block, but for some reason do not.
The report states that most campaigns are automatically optimized based on CTR and/or eCPM -- so this fraudulent traffic, if left unchecked, will dominate contracted volume. Gluck believes it would only take a little bit of effort to exclude these sites to reduce or eliminate click fraud.
The tests were conducted with two types of campaigns. The first was a managed campaign where ads were served to reputable sites, pre-approved by the advertiser to reduce concerns surrounding fraud. The second was a run-of-network (RON) campaign.
The buys were delivered in July on more than 20 million impressions to ads from 53 advertisers. These impressions were filled by the initial nine ad networks through downstream daisy-chaining on at least 45 additional ad networks on more than 100,000 sites.
Mpire also conducted a test directly on a top-ten ad network, with the goal of proving the power of referrer and fraud data to unlock the hidden value of horizontal networks. The impressions were bought on a CPM-basis, rather than on a CPC or CPA basis. The revealed that while 50% of the impressions were never within view of a user, the click and impression fraud volume was significantly lower than on the exchange-based buys.
Before optimization based on Mpire's data, there was an in-view-to-impression ratio of just 53%. After optimization, traffic generated by bots nearly disappeared and user engagement improved dramatically. Overall, the in-view-to-impressions ratio rose to more than 92%, post-analysis and implementation, according to the study.
Thanks Laurie - in case your readers are interested, here is a link to the study:
http://servedby.adxpose.com/Mpire_whitepaper_09-09.pdf
Direct download no longer available - please use this page: http://www.mpire.com/adxposerevealed
Please help me clarify...does that mean if I do an ad through Google or Yahoo, that the sites on which they put my ad will generate significant click fraud? Also, not sure how to take the steps to eliminate such.
When we designed a Behavioral Targeting Ad system, it was quite obvious that RON Ad networks always exhibited a higher click thru rate but very low click to buy rates.
Our suspicion was that these sites had some mechanism that 'simulated actions'. When we analyzed the IP addresses, the first node was always the same, clue to either bot controlled or site controlled robot clicks.
As for the Iframe issue, search engine vendors should start to provide some indexes on what html methods are used. I bet a good Phd in computer science could come up with a ranking index of sites that are "suspicious".
Search engines can then publish that against a site domain name so advertising metrics can take such measures into account when compiling eCPM statistics.
and the link is now
http://www.mpire.com/adxposerevealed
Laurie, wow that was a great headline. after 5+ years with Ad Networks I was intrigued. I knew things were bad but 95%? As many will I assumed it was based on all traffic. That was until I read further and found it was qualified as Run of Network or RON traffic. With this in mind the statistic was still surprising but not as much. Consider this, most advertisers pay MAYBE a penny a click for RON traffic. Meaning each party involved is talking about seeing sub penny bids. Yet they are going to complain about the quality of the click they are getting for one penny? I would love to sell a $40 pair of shoes and only pay $0.01 to generate that sale. Wouldn't we all? I think there is a problem of expectation setting on the networks side or expectations of the advertisers that is the problem here. Try getting any clicks on Google or Yahoo for a penny or less. There is a reason you get that click for a penny else where. If you want cleaner traffic....pay more. I remember when I worked for MIVA and the minimum bid was $0.05. Things got competitive from there and advertisers complained they could get traffic for a penny somewhere else so they dropped the minimum bid. Then they began complaining about the quality of the traffic...go figure. You can't get quality converting traffic for a penny or less. Set your expectations now If you want a 3-5% conversion rate be prepared to pay $0.25 or more for that click. Then you will see your valid rates go up.
Kudos for the sensationalism on your part and Mpire's. I am sure you accomplished your goal of sending even more customers running scared to Googles waiting arms. No matter how bad Google's customer service is.
Sheds new light on John Wanamaker's famous quote, except that in this case it isn't a matter of not knowing which half of advertising doesn't work, but rather which half is really there!
Sorry my comments are a bit late, but here you go: http://bit.ly/9iSAX for any interested, methodology is a bit flawed in my personal opinion.