Why video ad fraud remains a persistent problem for ad buyers

Video ad fraud is about twice as common as display ad fraud, according to a June study by DoubleVerify. The reason: there’s more money in video, and fraudsters go where the money is. Here’s a breakdown of some of the wonkier tactics that fraudsters use to reap video ad dollars.

Misleading pre-rolls
One of the most common forms of video ad fraud is simple and long-running: videos ads are mislabeled as pre-roll in the exchange. Instead of being shown in a video player (which are usually at least 400 pixels wide) before a piece of editorial video content, these ads get squeezed into 300 x 250 pixel display ad slots. Although this type of video fraud has existed for years, it has become acutely irritating for advertisers as the demand for video continues to soar. A study Integral Ad Science and demand-side platform Rocket Fuel conducted in April found that up to 70 percent of video ad inventory is misrepresented in this manner.

This type of fraud is difficult to trace because the ads are indeed video, they are shown to humans, they appear on the correct website the advertiser is trying to reach and they usually have high viewability, said Travis Lusk, vp of analytics at IAS. What makes the ads fraudulent is that they’re much smaller than what the advertiser thinks it paid for, and they don’t appear against the type of video content the advertiser anticipates. An autoplay video ad in a banner isn’t exactly a Hulu pre-roll.

Here is a peek into the ad tech abattoir where the sausage gets made. An ad network will buy $5 display CPMs in an exchange, repackage the inventory and sell it as video inventory for a $20 CPM in another exchange. Ad buyers purchase $20 video inventory thinking it is pre-roll against editorial content, but instead their video ad gets shrunk down and shown against video footage the ad network supplies. Sometimes the networks will license, and perpetually reuse, videos from news services like Reuters or The Associated Press. The networks also pirate video from the publishers’ sites that the ads will show up on to give an appearance of legitimacy, Lusk said.

To accommodate for the difference in aspect ratios between the video ad and the display slot, the network will add black bars to the sides of the ad, similar to how DVD players adjust between widescreen and full-screen formats. When reselling the display inventory, the ad network attaches its video content to the display ad impression so that when the exchange it’s reselling to checks the inventory, it will appear as video.

“The video ad exchange isn’t complicit in this,” said John Murphy, vp of marketplace quality at ad exchange OpenX. “They think they are getting a request for video. They won’t be aware that the video is being stuffed into a display unit.”

Static video ads
Video ad-serving template and video player ad-serving interface definition tags have recently become popular because these tags provide performance data to advertisers, and they help video players and ad servers speak the same language, which publishers need to scale their video ads. But the popularity of these tags has been accompanied by a rise in fraudsters mimicking the tags so they can pass display inventory off as video and pocket the difference between the CPMs, said ad fraud researcher Shailin Dhar.

Once the fraudster has sufficiently cloned a VAST or VPAID tag, it doesn’t have to go through the trouble of attaching generic video content to its display ad slots to trick some video exchanges. If the ad unit has VPAID status, the fraudster can send a static image through the exchange even though it is selling video inventory. Exchanges with more sophisticated fraud tools can detect this trick, but exchanges centered around low-quality (and thus cheaper) inventory do not have strict enough ad-quality standards to prevent this practice, Dhar said.

“They aren’t doing this on YouTube,” Dhar said. “They do this on the bottom-of-the-barrel video exchanges.”

How to avoid the fraud
Ad buyers can do a few things to avoid display-to-video fraud. For instance, if a unit that’s labeled as pre-roll is only around 300 pixels wide, don’t buy it. And since there isn’t enough supply of video to meet the demand, buyers should also be careful of hypertargeting with video because it’s unlikely there will be enough supply to meet the demand for a niche demographic, said Chris Wexler, executive director of media and analytics at ad agency Cramer-Krasselt.

“Just because a terminal in front of you says it’s possible doesn’t mean a buyer can turn off their critical thinking and accept it as fact,” Wexler said.

Another indicator is poor completion rates. If pre-roll video ads are almost always stopped within a few seconds, that indicates users weren’t actually trying to watch the video content the ads ran against. “Any DSP worth its salt” should provide this reporting, said Ian Monaghan, special operations consultant at Adobe Advertising Cloud (formerly known as TubeMogul).

If buyers use an ad measurement service, they can also examine if a particular ad slot flips back and forth between display and video. The networks won’t arbitrage the same display slot 100 percent of the time, so if that slot shows a mix of both video and display, that’s an indication video ads are being squeezed into display units, Lusk said.

Using the Interactive Advertising Bureau’s Tech Lab initiative ads.txt, which is a text file hosted on a publisher’s web server that lists all of the companies authorized to sell the publisher’s inventory, could potentially help buyers spot unauthorized sellers flipping display inventory into video, said Ari Levenfeld, chief privacy officer at Rocket Fuel. But ads.txt will only become a reliable check for buyers once publishers and exchanges widely adopt it, which could take awhile.

In the meantime, buyers should be more attuned to the realities of the economics of video advertising. If an exchange is offering $5 CPMs for pre-rolls, it’s probably too good to be true. But since cheap video is alluring, don’t expect buyers to change their habits overnight.

“Buyers continue to let people on their media plans that don’t play by the rules and rarely care about user experience,” said Marc Goldberg, CEO of fraud detection firm Trust Metrics. “Video is, and will always be, a target for the bad guys.”