Ad networks catch a lot of flak as conduits of bad traffic from shady publishers.
“It is true that some bad players intentionally sell fraudulent traffic and that other players are just turning a blind eye,” said Asaf Greiner, CEO and founder of fraud detection company Protected Media. “But there are also others working very hard to solve this thing.”
They’ve got a good reason to make the effort, because if fraud appears in the post-campaign analysis, advertisers won’t pay.
“That’s the point when the real struggle begins because everyone starts fighting about who is responsible and who is going to pay for the damage, and that leads to a lot of uncomfortable discussions,” said Christian Henschel, CEO of app analytics and measurement company Adjust.
Some ad networks have folded under the financial pressure.
“We’ve seen ad tech companies shut down after advertisers came to them late in the game to tell them they’d found fraud and that they were refusing to pay some very large amount,” Greiner said. “The ad network can’t retrieve the money, so that’s it for them.”
But Henschel is hoping a growing trend around transparency and information sharing will serve the dual purpose of fighting fraud and getting the appropriate parties paid.
Adjust recently formed a coalition of 12 ad networks and ad tech platforms that plan to work together to try and come up with a framework for mobile fraud detection and prevention, a list that thus far includes Criteo, IronSource, Jampp, Startapp, Vungle, Tapjoy, Chartboost, Glispa, Aarki, ClicksMob, Crobo and Remerge.
Through an API integration with ad networks, Adjust alerts its advertiser clients if it detects fraud and simultaneously pings the network explaining what was found and why it was rejected. From there, the network can turn off the fraudulent publisher mid-flight.
It’s a level of control that appeals to Ryan Faber, founder and CEO of Flatiron Collective, FanDuel’s digital marketing agency and an Adjust client.
“While traditional advertising agencies are incentivized to look the other way, we operate on a performance-driven model that aligns our incentives with our partners and pushes us to constantly squeeze maximum efficiency,” said Faber, who noted that Flatiron Collective has relationships with “all the major mobile ad networks.”
So far this year, Flatiron Collective has been able to detect and refute more than $2 million in fraudulent traffic by combing through user behavior looking red flags and anomalies and by tapping into Adjust’s suite of anti-fraud tools to suss out fake traffic and block payment on anything that looks suspicious.
“Typically, advertisers take early funnel conversion data at face value but miss deeper funnel irregularities,” Faber said. “Until recently, this was a manual process, and some fraud always slipped through the cracks.”
Although proactively tossing out bad traffic could cut down on a middleman’s profits in the short term, it’s ultimately worth it, Greiner said.
Smaato, for example, a mobile supply-side platform with hooks into hundreds of ad networks and DSPs and a Protected Media client, has an internal quality team whose primary function is to analyze traffic and get rid of the bad stuff.
“We’re also seeing mediators taking action and it’s a good thing,” Greiner said. “We see them actively throwing out traffic that could otherwise be very lucrative for them because they know in the long run that it won’t be good for business.”
Jampp, which describes itself as a data-driven mobile DSP rather than an ad network, jettisons about 30% of the traffic it sees.
“[We’re] not in the business of driving installs, instead we drive engaged users who, for example, complete purchases, book taxis, order food or make reservations,” said Jampp co-founder Diego Meller. “Fraud hurts us in these performance objectives the same way it hurts our advertisers.”
As more tech platforms make an effort to create better-lit environments, mobile-specific anti-fraud tactics will continue to emerge – best practices that Adjust, which is in discussions with the Media Ratings Council about accreditation, are hoping to promulgate throughout the ecosystem.
When traffic is determined to be coming from a data center IP, for example, it makes sense for the advertiser immediately kill payment, rather than waiting until the campaign is over. Or if an install campaign results in a steady flow of downloads even days after it was deployed – around 90% of downloads happen within the first 30 to 60 minutes – that’s also a clear indication that something fishy is going on.
Henschel acknowledged that creating any sort of mobile fraud standard is a project in its very early stages. The main point is to share information and jump start the conversation around transparency. But the motivation is also a little selfish, he said.
“As a measurement provider, we need to provide accurate data,” Henschel said. “But if there’s fraudulent traffic out there, the whole data set you’re looking at is screwed from the beginning.”
