First published in the Daily Maverick 168 weekly newspaper.
Kevin Frisch, the former head of performance marketing and customer relationship management at Uber, has revealed how the ride-hailing company was in the midst of the #DeleteUber crisis when it uncovered advertising fraud worth more than $100-million. Uber is now suing about 100 mobile exchanges for fraud, either for falsifying placement reports or for fabricating them.
Frisch told the Marketing Today podcast last week that in 2017, Sleeping Giants, a liberal social media activist organisation that describes itself as “a campaign to make bigotry and sexism less profitable” by persuading companies to remove ads from conservative news outlets, repeatedly tagged Uber founder Travis Kalanick on Twitter, questioning why the company was advertising on alt-right website Breitbart, which bolstered Donald Trump’s presidential campaign.
Kalanick, Frisch suggests, went postal – he had left Trump’s controversial economic council and wanted to be viewed as “more neutral”. “I would explain it’s not like we’re buying ads on Breitbart, there are networks and displays … But three days later, our ads would pop up again on the site. And the phone calls and yelling would start again.
“So I’m looking at these ads, to figure out which network is putting them on there, to ask why they weren’t respecting our blacklist,” he explains.
Return on investment?
Desperate to stop the yelling, Frisch turned off 10% of Uber’s $150-million annual ad spend on the rider side and … nothing happened. The number of riders signed up didn’t go down. “I was happy that it didn’t make an impact, so I started digging in – to understand what [was] going on. We started pulling our own log files, asking the networks to tell me exactly which apps the person was in when they saw the advert that they clicked on and we started gaining these reports.”
Uber’s own analytics team started seeing things that didn’t make any sense. Websites with small numbers of monthly active users (MAUs) – a key performance indicator used to count numbers of unique customers who interacted with a company’s product or service in a month – would show hundreds of thousands of installs. Or, users would be shown to have seen and downloaded the app in two seconds, an impossibility.
The tech firm, he said, then discovered that it had an attribution fraud problem: claiming credit and payment for downloads that would have happened organically. “Normally when you think of ad fraud you imagine that it’s impression fraud [bots creating impressions that you’re paying for] but we weren’t playing for clicks – we were playing on the first trip, when an actual human took a ride, swiped a credit card so we knew there were humans involved and the normal sense of fraud wasn’t applicable.
“Attribution fraud is where ad networks were taking credit for installs that would have happened organically; they get inside that path to get credit.”
Frisch said that on the Google Play Store, some ad networks created apps to monitor battery power but the apps had root access to devices, so when users downloaded the Uber app, another app in the background would fire a click on the phone, making it look as if the user had clicked on the Uber ad.
“Then the ad network says, ‘Thanks, Uber, you owe us $20.’ That’s just one of the many amazing methods that they have. It’s not an accident – it’s highly intentional. They spend a lot of time to figure out how to hide what they’re doing and make up where the clicks came from – and often they don’t do it well.
“For instance, we’d have an app with 1,000 MAUs that would apparently drive 20,000 installs. That seems a little unrealistic,” he said.
Here things got even more curious. When Uber cut two-thirds of its advertising budget, it saw no change in its rider app installs. But a lot of installs that the company thought had come through paid channels suddenly came through organically.
Uber is now suing the mobile exchanges Hydrane SAS, BidMotion, Taptica, YouAppi and AdAction Interactive for buying “non-existent, nonviewable or fraudulent advertising”, although it is only able to name five defendants – the rest are cited as fictitious “John Does”. Users who did see the ads merely experienced pop-ups and auto-redirects. The ad tech companies are believed to have been paid more than $70-billion between 2015 and 2017 for performance campaigns to encourage rider installations of the Uber app.
According to the complaint filed, the inventory they bought was junk: some of the ads ran on sites (such as Breitbart) that Uber had explicitly blacklisted. The companies are alleged to have engaged in numerous fraudulent practices, such as click spoofing, ad stacking and spamming, then falsifying their reports to cover it up.
In his end-of-year wrap of the “Top 10 marketing follies of 2020”, Bob Hoffman, aka the Ad Contrarian, noted wryly that “among many stories of clueless ‘performance marketers’ getting their shorts swiped by the crooks who have colonised the programmatic advertising dreckosystem, my favourite came from Uber”. Hoffman notes most “performance marketers” have no idea how deeply they’re being penetrated by online ad fraud. “They don’t even know where to look. They have no clue how untrustworthy or irrelevant the numbers they’re getting are.” But the most disturbing aspect of the story, he says, is the description of how “nobody gave one-tenth of a flying shit how much money was being pissed away”.
Marketers need to re-evaluate their thinking about ad fraud, because the problem is rife, Hoffman says: “Independent researchers … tell us that ad fraud is a massive problem (recently estimated at over $60-billion) that is becoming harder to identify and is growing dangerously.”
With more than $300-billion spent on digital advertising globally, state-sponsored hackers can penetrate some of the most secure systems, undetected. “Gaming the programmatic ecosystem (which transacts about 80% of online ad activity) has been shown to be astoundingly simple.”
Citing last month’s penetration of 250 US government agencies by Russia – which was undetected by notable agencies such as the military’s Cyber Command, the National Security Agency, and the Department of Homeland Security – Hoffman says it’s no “stretch to assume that fraud detection software can also be fooled”.
“It would be amazing if state-sponsored cyber criminals didn’t view the ad tech marketplace as ridiculously easy pickings and even more delicious since there are no consequences for being discovered.”
Where does it all go?
A PwC study commissioned by British advertisers trade body ISBA and the Association of Online Publishers into the UK’s £2-billion “premium” programmatic ad market (including high-profile advertisers, publishers, agencies and ad tech) has shown that 15% of all marketing spend disappears into a supply-chain black hole, with only 50% of investment making it to publishers.
Digital fraud investigator Dr Augustine Fou notes in a Forbes column that marketers should assume that fraudsters are “feasting” on their digital ad budgets, until proven otherwise. “Vast, scalable botnets can easily create billions of ad impressions out of thin air, literally since it’s all bits and bytes flying around. The resulting CPM [cost per thousand impressions] prices are substantially lower now than before.”
With fraud priced in, Fou says marketers justify continuing to buy digital media because it’s cheap. This is not only causing marketers to spend even more on fake ads on fraudulent sites and apps, it is also harming society. “As big advertisers continued to shift money from TV, print, radio and billboard advertising into digital, we witnessed more and more cases of their ad dollars flowing to porn sites or worse, unbeknownst to the advertisers,” he says.
Websites offering free child abuse images and videos have proliferated, Fou argues, because they are able to make money off digital advertising – and technology from Google, Microsoft, and Facebook enabled ad dollars to flow to these sites and apps.
“At the very least, marketers should consider it their moral duty to take a closer look at their own digital ad spending via programmatic channels to see if their dollars are funding disinformation, hate speech, fake news, porn, and child abuse.”
The industry is not blind to ad fraud in the supply chain, with a tremendous amount of awareness and education going into exposing the problem, explains Jarred Mailer-Lyons of The MediaShop.
He says that although there are anti-fraud vendors seeing a decrease in ad fraud, “you need to remember that the law of supply and demand remains and as digital spends increase … so does ad fraud. It’s believed that by 2022, the total cost of digital ad fraud is expected to grow to $44-billion globally.”
Using anti-fraud verification services has become best practice, with a move towards ads.txt – an IAB initiative to improve transparency in programmatic advertising – and the recent App-ads .txt, which lists the ad sources authorised to sell a developer’s inventory, “should hopefully set us on the right path”.
“Market fragmentation plays into fraudsters’ hands so it’s about consolidation of supply and relationships with trusted partners.
“I also believe that while programmatic has its place, it’s also about working with publishers directly, as they have a higher percentage of fraud-free inventory compared to an open exchange buy.”
Mailer-Lyons says supply chain optimisation helps improve bidding efficiency in online auctions and can also bring any shady deals to the forefront.
There’s also no substitute for human intervention to ensure that campaign-level optimisations are applied to monitor unusual occurrences like click clusters and to ensure that blacklists are enforced. DM168
This story first appeared in our weekly Daily Maverick 168 newspaper which is available for free to Pick n Pay Smart Shoppers at these Pick n Pay stores.