• by Augustine Fou , April 10, 2014

Mediapost’s Video Insider Summit, held earlier this week, tackled the topic of video ad fraud. I was part of the panel that was well-represented by people on the front lines of the massive and growing fraud problem -- from the agency side, security technology side, and video ad serving platform side.

The audience was also savvy on the topic -- the majority of them correctly estimated the problem to be in the 25 - 50% range.

What was eye-opening were the variety of things the bad guys were doing and how sophisticated the technology employed by the bad guys has become.

We can break viewability fraud into three strata:

1.Those ad views that are deliberately fraudulent because they are generated by bots on sites where no humans would ever go;

2. Ads that are not in view (e.g. below the fold) or not visible to humans (1x1 pixel in dimension, or transparent),

3. Sub-optimally targeted ads.

Because the topic is so broad, our panel mainly focusing on deliberate fraud.  If the advertisers can cut out that sizable chunk of waste, the productivity of their ad dollars would rise dramatically.

So what are some of the techniques used by the bad guys to rip off advertisers’ dollars?

Bad guys use millions of malware-compromised PCs (collectively their botnet) to create hidden browser sessions in the background, unbeknownst to the human user.

These browsers visit various Websites, click on ads, start videos in order to load video ads, add things to shopping carts, etc. In this way, these “bots” generate ad views that advertisers pay for.

Furthermore, bad guys autoplay ads, stick many ads on the same page, stack ads on top of each other, play in-stream ads in-banner instead, among other techniques.

This creates a large amount of “inventory” which is sold into ad exchanges, to be resold often several times, ultimately to unsuspecting big advertisers and their media agencies. (See Video Ad Views Exploded in 2013).

The panelists cited a few more examples of just how sophisticated and advanced the bots are.

Jay Benach from WhiteOps cited bots that are designed to visit a variety of mainstream sites for the purpose of building up a desirable “cookie profile” -- e.g. they visited these fashion sites, added a few items into a shopping cart and abandoned it after searching for specific items. The next time that cookie shows up in a real-time bidding environment, big advertisers would “bid up” what they are willing to pay to show an ad to that user because it has all the desirable characteristics that match a set of targeting parameters.

Matt McLaughlin COO of DoubleVerify, talked about bots which selectively delete cookies to avoid frequency caps. Bots can create “dwell time” to trick video completion metrics. They can also stay on a video ad long enough for it to be counted as viewed.

Now that advertisers are becoming aware of the large and still growing fraud problem, the next stop is to start to mitigate the waste due to this kind of fraud.

That takes changing our way of thinking. One of the basic conclusions by our panel was that as long as brand advertisers and their agencies defined ROI as “more ad impressions for lower cost” they will continue to buy cheaper and cheaper ad inventory. That inevitably means there will be “dirty” or fraudulent inventory.

But if the definition of ROI shifted to become more “performance” oriented -- like actions that a real human customer would enact to make a purchase on a site-- then there wouldn’t be the gross incentives to just get more impressions. Instead advertisers would focus more on “productive” impressions, which mean those that are viewed by humans.

We asked the audience for a show of hands, and most of the “demand side” in the audience said they already maintained blacklists and whitelists. They also had various technology platforms like WhiteOps, DoubleVerify, BrandAds, Integral Ad Science, Spider.io, etc. deployed to help detect bot activity.

But only some had internal benchmarks they used to detect anomalies that would be indicative of fraudulent activity. In other words, they tracked the usual pattern of activity and when certain events appear to be out of pattern they could further investigate them for fraud.

They all admitted it was hard to keep up. That echoed a panelist’s sentiment that it was very unfair to have brand managers and media planners go up against the world’s most advanced hackers.

Clearly, though, the panel and the audience agreed that ad fraud has severe long term implications for the entire digital ad industry if left unchecked. And we all agreed that more concerted effort amongst the various constituents -- buy side (advertisers), sell side (publishers), industry groups and technology providers -- was needed to truly make an impact.