Summits Yellow

How brands like Uber can be saved from ad fraud – Adjust

Mobile Marketing

In light of Uber’s click fraud lawsuit against Fetch, Adjust CTO and co-founder Paul Müller asks why fraud continues to be such a major issue – and how much of the fault sits with advertisers.
Last week, a large ridesharing app filed a $40m click fraud lawsuit against its mobile agency, alleging misrepresentation on the effectiveness of their mobile ads in acquiring users for its mobile app. The case is particularly notable because it speaks to the sizable impact of ad fraud and how brands, agencies, ad networks can be at odds with each other when fraudsters run away with millions. If this case goes in the client’s favour, it could lead to more brands fighting click fraud after the fact with hefty lawsuits.
So, how do we start preventing fraud before it strikes? Let’s look at each part of the industry in turn.

When we released our first fraud prevention filter almost two years ago, we couldn’t have predicted the uphill battle it would be to save our clients from spending millions on fake inventory.
The very first beta test revealed two issues we didn’t anticipate. First, we were asked to double and triple check our numbers, as clients claimed they must surely be wrong and couldn’t possibly be that high. We didn't understand. Why wasn't the client happy to save almost $100,000 every day?
On speaking with the client further, it dawned on me. He didn't want us to be right.
The UA (User Acquisition) manager overseeing the test was measured on ability to buy new users at a certain price. We made it impossible to buy the volume at the price he needed, as simulated installs are readily available at any volume for CPIs as low as $1, complete with a $0.99 margin. Even worse, we exposed that he had likely spent a few million dollars on fraud. Overall, he was not incentivised to remove fraud from his inventory.
Only after learning that UA managers have spent large sums on fraudulent app installs do clients often take action. On one particularly bad occasion, the head of UA was fired after it became clear that more than $10M had been wasted.
Herein lies the rub – as long as UA managers are incentivised based on how many users they can buy at the lowest price, effectively penalising them for removing fraud from their inventory, our ecosystem will suffer from massive levels of fraud, with everybody turning a blind eye.
Ad networks
Which leads me to the second culprit of this mess: networks and their sub-publishers. Often, even basic due diligence for publishers is thrown out of the window in favour of being able to offer more inventory.
Selling organic users
Surprisingly, I can sell you your own organic users at quite a margin. The newer and much more common type of click spamming and click injection fraud, which game attribution and claim credit for organic users, has only made the problem worse. Guaranteed, any big app with a sufficiently large userbase is currently buying their own organic users to some degree.
The root of the problem here again is the incentive. Sure, the majority of networks are genuinely interested in keeping their first-party inventory clean, but as long as a sufficient number of players profit from turning a blind eye, it's a disadvantage to be the only good guy.
Over-attributing for partial impressions
Another problem is the blatant push to over-credit and over-attribute their traffic. Many video networks send only video loading impressions to vastly improve their conversion rates and give them much more credit. Without an effective real-time filter against click spamming, advertisers are fooled into believing ridiculous conversion numbers.
But, as we already mentioned, the advertisers actually want to be fooled.
Incomplete attribution to too many networks
The ultimate pinnacle of overstating the influence of a campaign is multi-touch attribution. While I believe that the user’s journey should not be reduced to the last click, this is a huge challenge on mobile. In-app attribution can only work by combing through billions of clicks and views, and trying to match the advertiser ID with any candidate.
In-app inventory has a huge chunk of self-attributing API partners that disclose only the last engagement of a user journey, essentially cutting off 50 per cent of all potential touchpoints. This leaves us with a system that is 50 per cent incomplete and further over-values click spam, leading to even more useless data.
Attribution companies
Without naming names and pointing fingers, I’m going to summarise the key actions we should and should not be doing as an industry, because we’re only stronger in fighting fraud if we’re all doing it together. 
As the saying goes: “The only thing necessary for the triumph of evil is for good men to do nothing.” Blow the whistle, and confront the ad networks when you find fraud. Don’t shy away just because your network sends client referrals. It’s critical to push back and fight for credit on fraudulent traffic. I can tell you from personal experience that enabling our fraud prevention typically leads to someone yelling, preferably at me, preferably over the phone.
Fight the right fights
Fraud prevention that doesn't combat click spam is nothing but a fig leaf to hide behind. Simulated installs and click spam are the types of fraud to fight, and fight proactively before it gets paid out.
The ecosystem
Fraud is worse than ever for those left undefended. And I mean much, much worse. Anywhere from 20 per cent to a whopping 50 per cent of fraud outside the big three networks seems to be the norm. Yet, fraud is a zero-sum game. Every time a big company enables fraud prevention, the fraudsters stop being rewarded.
Additionally, we need every facet of the advertising ecosystem to work together in battling fraud. The recently-launched Coalition Against Ad Fraud consists of 15 networks and app marketing platforms that have agreed to global standards and guidelines in order to participate in the fight against fraud together. These organisations have agreed to set up rejection callbacks for all campaigns, ensured network partner’s average server-to-server postback latency lies under three seconds, and complied with kicking out most fraudulent sources.
It’s a first step in the direction of real change in prevention, and we need more joining continuously, so that we can start saving millions from going to fraudsters – and lawsuits – in the battle of ad fraud.