The Association of National Advertisers (ANA) and cybersecurity company White Ops opened its fourth annual Bot Baseline report with the proclamation that “in some important ways, things are better than they have ever been.” But in spite of that cheery headline, the report left industry leaders feeling cautiously optimistic at best, and downright skeptical at worst.
The report, released last week, predicts that the budget lost to ad fraud this year will be a mere $5.8 billion—down from more than $6 billion just two years ago. White Ops pulled those numbers courtesy of the 50 ANA member participants in its study, who ran 2,400 campaigns across 606,000 domain, with 27 billion impressions between them.
“For the first time, the majority of fraud attempts are getting stymied before they are paid for,” the report reads. “[Either] by DSPs and SSPs filtering fraudulent bid requests, by clawbacks, or by other preventative measures.”
In other words, anti-fraud measures are baked into more layers of the ad-tech stack than ever before, lightening the load that those on the buy side need to lift when buying a publisher’s inventory. Meanwhile, industry heavyweights like Google are pulling its partners to adopt ads.txt in an effort to quash bots and brand-unsafe experiences.
“Thanks to the wonders of capitalism and competition, there’s a whole market for fraud controls that simply didn’t exist when we ran our first report,” said Michael Tiffany, the president and co-founder of White Ops. “Nobody in 2014 could have done any of this.”
When those fraud-fighting tools become more robust, he explained, bad actors struggle to find work. According to Tiffany, while the demand for bot traffic is still booming in 2019, the market for it has largely dried up; either because the bot sellers weren’t sophisticated enough to outfox modern algorithms or were forced further underground, cutting off their supply to buyers on the open web.
The news, while positive, led to little more than some hesitant hoorahs from voices in the industry. John Montgomery, evp of global brand safety at GroupM, said that “It’s encouraging to see the number come down, but the potential financial risk is still high for marketers who do not take robust actions to avoid ad fraud.”
“Measuring, benchmarking and using technology to avoid ad fraud, and optimizing media schedules to stay away from risky environments must be standard protocol for every campaign,” he cautioned. “If marketers and their agencies take these precautions, exposure to ad fraud can be minimized.”
Montgomery wasn’t alone in coupling his optimism with a measurement of caution that agencies and their clients need to keep their guard up.
Johnny Ryan, chief policy and industry relations officer at Brave, said it was “useful that the ANA continues studying the multibillion-dollar ad fraud problem. Every dollar that criminals steal from advertisers is a dollar stolen from publishers who could have used it to fund editorial production.”
He also said that he agrees with White Ops’ “warning to ‘reject the temptation of buying for tonnage,’ adding, “Buying ad views of unknown provenance is a losing proposition.”
Greg March, CEO of independent media agency Noble People, said he wasn’t surprised that the report found forms of ad fraud on the decline, given the attention paid to the issue in recent years.
March added that just accounting for fraud was not enough, arguing that media agencies need to employ higher standards for media buying more broadly.
“By default, if you optimize your buys properly and intensely, that will minimize your exposure to fraud,” he said.
Louis Jones, evp of media and data for 4A’s, stressed that the attention paid to the issue is partially due to measures the association had recently taken to beef up security for its members, like the introduction of the Advertiser Protection Bureau, or APB.
“We’re able to share information, share incidences, alert people to holes in the tapestry more quickly and more easily than ever before,” he explained, adding that recent months have seen the APB take a “more organized approach” to sharing information about fraud among themselves, holding companies and clients.
“I think these moves have made agencies a little more sane,” Jones said. “But they’re still sleeping with one eye open, because you just never know when that dam will break.”
Meanwhile, some critics say that dam has already long broken, and that ad tech is in far worse disarray than the White Ops report would lead agencies to believe.
Given the study’s meager sample size—the 27 billion impressions sampled make up less than 1% of the 70 trillion impressions seen web-wide annually— some industry observers wondered if the ANA could be cherry picking data to show the results it wants to see. One source suggested the ANA “is protecting the status quo instead of helping their members get to the truth.”
Method Media Intelligence co-founder and CEO Shailin Dhar said the ANA’s assessment was “misguided” and based on “flawed logic,” while also pointing out that the group and White Ops did not reveal the sample rate of the study and that participating campaigns were limited to advertisers that agreed to participate in the study and the campaigns they chose to “volunteer into the study.”
“The ANA study is a small sample of overall advertising and to say the overall effective fraud is going down is a mistake,” he said, “because we see evidence to the contrary on a daily basis.”
He emphasized that the report found that bot traffic was “becoming more sophisticated” and yet the ANA extrapolated that fraud was going down, while simultaneously acknowledging that “only about half of impressions were even measurable.”
Dhar claimed that Method Media Intelligence’s own research suggests that automated bot traffic in particular is actually getting cheaper and easier to buy as data centers and cloud-hosted computers are becoming more accessible, in contrast to White Ops and ANA’s claim that “Anti-fraud measures have made traffic sourcing more difficult and expensive.”
He explained that while the finding that botnet traffic in particular is becoming more expensive, that is not the case for bot traffic, more broadly defined.
“How can they say that fraud is going down when they acknowledge that … techniques are getting more sophisticated?” he asked, comparing the assessment to a police officer saying a given drug is “harder to trace and easier to buy on the street” but concluding based on declining arrest rates “that the problem is decreasing.”
Method Media Intelligence co-founder and CTO Praneet Sharma also stressed that bot traffic was only one aspect of ad fraud, pointing to non-rendered ads being a large-scale problem where advertisers are “wasting a lot of money” that doesn’t get the same media attention.
“The botnet storyline makes for good press,” Dhar explained, while Sharma added that the amount of money lost just to bot traffic is still “pretty absurd.”
Dhar also pointed out that White Ops’ fraud tracing techniques left a lot to be desired. White Ops, like many cybersecurity outlets, relies on certain behavioral red flags—like suspiciously formulaic typing or clicking—to tip them off to a bot’s presence.
The issue, Dhar added, is that scammers know these checks exist, and are coming up with increasingly clever ways to mask their traffic as the real deal.
“There’s only so much behavior you can check, so it’s just a matter of trial and error to see what gets through,” he explained. “I’m sure that this [White Ops] study whitelisted bad actors on the basis of their ‘passable’ behavior.”
Dr. Augustine Fou, an ad fraud researcher and consultant, stressed that the data in White Ops’ report is “factual and accurate,” but that “where the problem lies is in the trade association trying to extrapolate that to the industry and saying that fraud is going down … that’s going to put people at ease and they should be more vigilant than ever.”
“The challenge is they don’t know what they don’t know,” he said, adding that ad fraud encompasses far more than the bots measured in the study. This is particularly true in regards to connected TV and over-the-top streaming devices, where the current metrics of measurability don’t apply—and are also exactly where fraudsters are shifting their focus, according to some analysts.
Dr. Fou pointed out that White Ops shared that conclusion in the report itself, writing out that “less than half of all impressions are fully, transparently validatable.” Direct and programmatic video buys on mobile lag particularly far behind, with 28.8% and 33.6% “validatable at the highest level.”
“White Ops is admitting, correctly, that they can’t measure half of this stuff,” he said. “If you can’t measure it, why would you want to buy it?”
He suggested that agencies and even marketers themselves were incentivized to look the other way regarding ad fraud.
“The agencies are trying to maximize their own profits. They’re not going to look for fraud, because they don’t want to find it … and now I’ve found that even some marketers are doing everything to suppress knowledge [of ad fraud] because they want to keep buying large quantities,” he said. “Everyone believes in the whole reach and frequency thing.”
“If we actually solve the fraud, there’s going to be far less [inventory] for people to buy,” he added. “The agencies don’t like that and the big brands don’t like that, especially those marketers whose badge of honor is how much they spend.”
Tiffany conceded that since the nature of the sample included some limitations, while explaining that White Ops attempted to account for such shortcomings.
“Our sample draws from ANA members, which tend to be larger advertisers, and half of them were repeat participants, which almost guarantees that we face an inherent sample bias skewing towards more sophisticated advertisers,” he explained. “So we had to normalize the data from the study participants with other data sources, such as the programmatic bidstream that we get to see.”
He agreed with the assessment that, despite the areas of improvement highlighted in the report, it’s imperative for agencies to stay vigilant.
At the same time, he stressed that structural improvements such as advertisers directing more traffic through channels with independent fraud prevention measures, initiatives (such ads.txt) that have helped to reduce desktop spoofing and high-profile arrests are “big differences in 2019 that weren’t present when we did our first study in 2014.”
The big question is how the industry continues to respond the threat of fraud, particularly as it takes new forms.
“The frontiers of fraud, highlighted in our study, are growing. But those growth areas haven’t scaled far enough to overwhelm the very real gains from those five big improvements,” Tiffany said. But whether fraud continues to decline or “the frontiers of fraud” overwhelm progress “depends on how, as an industry, we persevere in the fight this year.”