This is the 11th in a series of interviews with vendors combating the problem of ad fraud. Other companies participating in this series include Sizmek. Read previous interviews with comScore, DoubleVerify, Dstillery, Forensiq, Integral Ad Science, PubChecker, Telemetry,Videology, White Ops and RTB Asia.
The meta-problem with ad fraud, according to Moat CEO Jonah Goodhart, is that it’s a tremendous industry concern, yet it doesn’t have a definition. Bots surely constitute ad fraud, but what about a 1×1 display banner, or ads stacked up in a single iFrame?
Goodhart acknowledges the latter examples are problems, but do they technically represent fraud or just bad media?
“Fraud” is such a blunt instrument of a term that advertisers can be careless wielding it. Its broadness creates a lack of transparency, which from Moat’s point of view means it can be difficult for vendors, publishers and advertisers to reconcile the issue among themselves. Goodhart rejects the notion that fraud is whatever the advertiser believes it is.
Moat’s solution is to specify exactly what it’s seeing, a discipline that overlaps with its core analytics expertise.
“The solution is to automate clear, defined metrics that roll up to fraud as a theme, without actually saying it’s fraud,” Goodhart said.
An advertiser should be able to know what percentage of traffic came from hijacked devices, for instance, and if she doesn’t want to pay for that, she shouldn’t be obligated to do so.
“We really tripled our efforts in the fraud detection arena,” said Goodhart.
Fraud detection had traditionally been part of the company’s internal tool set, but the company is making plans to expose it to clients as part of its analytics platform.
“We’ve seen huge demand from marketers, who want transparency into what’s happening on the viewability side as well as the fraud side,” Goodhart said, adding that this has come up in every client conversation Moat has had.
Goodhart and Dan Fichter, Moat’s VP of engineering, spoke with AdExchanger.
AdExchanger: How is your anti-fraud solution – for lack of a better term – different from the others?
JONAH GOODHART: The [industry] theme is: “We can’t expose what we’re tracking because all of the fraudsters will understand how to game the system and beat us all.”
We don’t fully buy into that. We believe being transparent with what we’re tracking and allowing the marketer to make that decision based on their data is the way to go. When they see fraud detection in the nonhuman space, it’s opaque. The ad is blocked and there’s no indication what happened or no reason why something is being classified as fraud.
What exactly is fraud?
JG: When people talk about fraud, they talk about bots or automated traffic. But another is autoplaying a video that’s muted off-screen in a way that’s almost guaranteed to not be seen by a person. Those aren’t bots, but the marketer feels they’ve been defrauded by that.
Unlike viewability, there’s no definition of what fraud is. Some people say that if you’re showing ads on places that are not viewable, that’s fraud. Bad media doesn’t necessarily mean it’s fraud. If you’re generating fake impressions into an iFrame, that’s fraud.
How do you act on that distinction?
JG: Through transparency. We’re not in the judgment business. We’re in the business of analytics, and that’s a pretty big difference in how Moat approaches the market. “This ad was served in a way where none of the impressions were viewable. This ad was served in a way where it all went into robotic iFrames.”
Marketers might say, “Don’t run my ads where there’s fraud.” But they see viewability is 10%. That’s not fraud, but performance is awful. The way forward is being very clear about what you’re doing and to measure your ad at a very granular level and deliver the data judgment-free.
We don’t think the approach of a black box calling something fraud will work at scale.
DAN FICHTER: Our goal is to be very transparent about what we’re seeing, because reconciliation with vendors is impossible unless there’s transparency.
Break down how Moat approaches fraud from a technical standpoint.
JG: The way Moat analytics works today for nonfraud, nonhuman traffic, is we list out 40-50 different metrics people can choose from.
DF: Internally, we have 40-plus metrics indicative of suspicious traffic, which began with work we’re doing on viewability optimization and exposure optimization. We have attention segments and audience segments we built that have higher viewable completion rates in online video. The idea is to pick out the users who are most attentive.
As part of that development, we’ve started to notice different types of fraud. There’s adware, nonbrowser agents, tag hijacking.
How do attention segments relate to fraud?
JG: For instance, you come to AdExchanger and spend more time on the site than usual, you’re more likely to be attentive. We build an audience segment [around that], which enables marketers and publishers to target true attentive audiences.
We combine that with looking for signals that tell us it’s a human being or signals someone is more likely to be attentive. We’re beginning to do this at some scale and the results we’re seeing are pretty interesting. We see higher rates of viewability, lower rates of nonhuman traffic.
You mentioned your fraud detection capabilities were mostly internal and you were rolling it out for external deployments. What’s the status?
JG: It’s mostly internal today, but we’ve been going client-by-client and starting to talk to them about the data they’re seeing. It’s not something we’ve rolled out at scale yet.
DF: From an ad ops and integration perspective, our goal is to be easily pluggable into existing systems. There’s no integration we’re asking people to do.
When will you roll out fully?
JG: We’re collecting the data today but not exposing it at scale. In terms of exposing it as an always-on set of metrics, Q4. Our view is we want to make sure that we’re doing the right thing, exposing details about the metrics.
Back in July, we thought of leaving it as a rolled-up number, but our view is in order to get over this hump of everyone having different opinions on what fraud is, we needed to be transparent. Let’s be clear about what we’re tracking and expose it, and see where we go from there.
Why is transparency a controversy in the industry?
JG: A lot of talk in the market is that fraudsters will [win] if you expose what you’re tracking. They’ll suddenly figure it out and adapt. But if they’re true fraudsters, they’ll adapt to whatever you’re going to do anyway.