The topic of digital advertising fraud gets a lot of play in consumer media but it's increasingly becoming an issue in B2B--not just because it's happening more, but because both advertisers and publishers are increasingly able to detect it. Do you have answers ready for when your advertisers call you out on invalid traffic?
At a recent Connectiv Digital Media Council meeting on "Invalid Traffic and Methods to Prevent Revenue Loss", Justin Hoffman, director of technology at Strategic Insight (formerly Asset International) and Dr. Augustine Fou, an independent ad fraud researcher, showed how ad fraud is growing in B2B but also how publishers can take some relatively straightforward steps to combat it.
First, don't be in denial. "We've seen clients impacted by ad fraud in ways they haven't expected," said Dr. Fou. "Historically, publishers who measured bot, or non-human traffic, found low occurrences and thought that meant it doesn't affect them. One of the key points they're missing is that bots don't come in large quantities. Bots come to a publisher's site to load a cookie, but then goes somewhere else to load the ad impression and that site gets the revenue, not the publisher."
As a percentage of ad spending, fraud is getting lower because companies are detecting and mitigating fraud, but the sheer number is still going up according to Dr. Fou. "In 2010, at the beginnings of ad fraud measurement, estimates of digital advertising fraud ranged from 9 percent to 17 percent. Now estimates range anywhere from 3 percent to 98 percent fraud. Which one is correct? Both are correct in very specific circumstances. 'Zero percent' is when the advertiser and publisher are paying attention and taking steps to mitigate it. Ninety-eight percent is when they weren't paying attention."
Dr. Fou gave an example of one ad network (AppNexus) that cleaned up its act and reduced the number of ad impressions it was generating from 260 billion to 20 billion. "If they can clean up 92 percent of ad impressions, what were advertisers buying before?" said Dr. Fou.
There are four main buckets of digital advertising fraud: impressions, clicks, leads and sales. Since impressions and clicks account for 91 percent of digital ad spend that's where most fraud occurs.Impression fraud features fake sites designed to carry as many display ads as possible. Other sites have no content and are designed to carry search ads for the purpose of driving clickthroughs. Both types of fraud feature two common ingredients: fake websites and fake users (bots).
Fake websites (cash out sites): These typically offer no content and no human visitors. Sometimes they buy traffic that is created by bots. Sometimes real sites with some original content steal ad impressions off other publisher sites. Fraudulent sites are often created by templates, including WordPress.
Fake visitors (bots): Bots are automated browsers. Originally developed as legitimate tools that developers used to test sites or mobile apps before launch, bots can be programmed to hit a web page over and over or tak actions like scroll or click. Malware can live on a real person's device and track and mimic what they do.
Ad fraud is extremely lucrative (no content to create and pay-as-you-go data centers) and scalable. "If someone is cheating and puts five ads on a page, why not 15 ads?" asked Dr. Fou. "Why not put 131 ads on a page? Why not put that entire page into a 1x1 iframe and when they use a bot to the load page once, they generate 13,000 impressions?" (If those numbers sound very specific, that's because it actually happened.)
While much of the industry is focused on detecting bots, the bot lists typically feature legitimate bots that have already identified themselves--Googlebot, Bingbot, etc. "These lists typically have several hundred names, while the universe of known bots is 10,000," said Dr. Fou. "Bad guy bots don't reveal their names and they aren't on any lists. Bad bots go to a site, set a cookie and go to other sites to start looking like a professional in the field. Because they can now pretend to be that professional, they then go into ad networks and cause impressions to load on other fraudulent websites. You may not see a lot of bots on your site, but they are still siphoning revenue away."
Emerging technology will actually cause the problem to get worse before it gets better, according to Dr. Fou. "We're on the verge of huge proliferation of Internet of things, wearables that have processors and Internet connection," he added. "At least a PC can be turned off. These devices will always be connected and running."
How Publishers Can Fight Back
But it's not hopeless. Publishers do have recourse to stem the flow of invalid traffic and secure the trust of their advertisers.
"Ad fraud become a hot topic for us this year," said Hoffman. "The bots are getting smarter, there are more devices out there and advertisers are getting smarter, too. We'd been getting alerts from some paid advertisers that they were seeing an increase in invalid traffic on our sites. Advertisers now have tools that show them what they couldn't see anymore and the dominoes are start to fall--advertisers will demand money back or makegoods. As publishers, we need to understand how it affects our sites. It's been going on for a long time but it's recently come to light because the technology is smart enough to detect it."
Until recently, whenever Strategic Insight saw domains that drove suspicious traffic to its properties, it blocked the domain. But increasingly sophisticated bots call for more sophisticated counters. "There are many different things than we can do," said Hoffman. "We were always building honey pots--putting something on a webpage to trick the bot and identify it doing something malicious. These methods now have hit a brick wall. Bots are getting too smart. The tech on both sides of the fence have hit a stalemate."
Strategic Insight did take a series of actions with its Plan Advisor site that paid of big. The site underwent a redesign in spring 2015 that focused on improved viewability--getting ads in sight for at least one second. "Viewability was in the 80 percent to 90 percent range, but advertisers started telling us that we had
invalid traffic," said Hoffman. "There was no way to disprove it and we heard it from a growing number of advertisers."
Traditional counters like blocking didn't work."When we started looking at the domains driving invalid traffic, they were domains that I traditionally wouldn't want to block, like Amazon or Adobe," said Hoffman. "If it's a shady-sounding domain, it's easy to block. But sites like Adobe, those are things I know I don't want block."
A growing list of services like Moat, Forensic, whiteopps, pixalate and others are offering analytics and insight not previously available to publishers. "For me, a lot of the data coming from these vendors is unique and not things I could get from other sources," said Hoffman. "It proved to us quickly what advertisers were telling us."
One of other things that happened on Plan Advisor were crawlers taking content. "A site called Hardy Read was crawling our site, taking code and putting it in their site," said Hoffman. "They were grabbing the content and the ads and putting it on their websites. In addition to doing cease and desist, we don't have time to wait, we needed to cut it off right away."
Hoffman and his team started inserting "wrappers" (java script) around advertising scripts."Something I battled with is, I didn't want to just stop blocking traffic because I was worried about pageviews," aid Hoffman. "Inserting this java wrapper meant that when anyone hits the page these tools make a decision--is the thing going to this page a human or bot? Based on that decision, they serve a real ad or a house ad. It's a good option--if there's a question about serving you an impression, I just serve you a house ad. The last thing I want is to be responsible for is blocking too much traffic. I still will let you through the door but I'm not offering you impressions."
By implementing this approach, bot and invalid traffic dropped from 30 percent to just below 3 percent on Plan Advisor. "We've done really well making good with advertisers and showing that once they sound the alarm we are responsive to it," said Hoffman. "We were at risk of losing advertisers and this was a quick and easy way to address the issue. It's critical to keep an eye out and important to stay ahead of attacks. Two or three years ago, I would have been forced to try to fix this on my own. It's time to fight back and take control. Bots have been cleaning up for years on our sites but now we have solutions that fixed this."
Connectiv members can see the full presentation here.