Mobile advertising has a dark skeleton in the closet under the term Fraud.
Appsflyer, the largest 3rd party attribution solution estimated that attribution fraud accounted for $75B of ad spend wasted in 2018.
Attribution fraud was extremely large as mobile app advertisers were often engaged with performance advertising - paying publishers based on the number of installs they “generated”. The credit was given to whichever publisher brought the last click before the user installed.
This was a giant loophole. The entire system was incentivized to give credit as being “the last click”.
Fraudsters saw this as a major opportunity. The opportunity was huge and the reward was much larger than robbing an average bank.
High profile lawsuits, market education and awareness didn’t fix the problem itself. The market’s response was to consolidate ad spend to the Self Attributing Networks (SANS, SRNs, Walled Gardens).
These were large companies, with much scrutiny, and no real reason to mess around in generating fake clicks, or injecting clicks to win the credit.
In 2017, Google, Facebook and the other giants took over 90% of the growth in mobile ad spend.
Fast forward to 2020, attribution fraud seems like it was a thing of the past, as the market starved most of the fraudulent publishers, but this was the year that Apple announces “App Tracking Transparency” and SKAdNetwork
The role of attribution solutions became more about reporting and cost consolidation rather than actually doing the attribution itself. Apple’s SKAdNetwork was a privacy centric attribution solution that was free to use by app advertisers.
While SKAdNetwork placed the future of 3rd party attribution solutions at risk, it also gave systematic fraud a big opportunity.
(we already predicted this in September 2020)
Check the video below:
What this video is showing is a popup rewarded video ad, which the user (me!) triggered.
Once the video ends, the user is prompted to click the ad, or close it.
If the user (me again!) would have clicked the ad and installed the advertised app; the app that served me the ad (the publisher) would receive credit from SKAdNetwork as the one being last before I downloaded the Advertised app.
Since I did not click the ad – the publisher did not receive any credit, other than the value for the impression of the ad itself.
Check out the video below
Similar to the previous video, what you’ll see here is a popup rewarded video ad which the user (me!) triggered.
Once the video ends, without doing anything, an app store window is loaded to the screen, prompting me to download the Advertised app.
In the SKAdNetwork framework – loading the app store counts as a click.
My device was just “marked” by the SKAdNetwork as having clicked this ad, while I did no such thing.
Now, had I gone to the app store, the following day, after having seen other ads for the advertised app elsewhere (but not clicking), the publisher app (in the video), and as a result, the ad network, would have received credit for me installing the app.
Given that Apple isn’t sharing click or individual install data with anyone – the system must work within the framework for it to function as prescribed.
While the legacy attribution fraud was mostly dumb fraud (click spamming, click injection), it was easily traced back to the source. Most of the fraudulent installs were coming from a few bad seeds who eventually got sued, or disowned by the ad networks. The advertisers had almost unlimited access to user-level data, allowing conversions and clicks to be deduplicated and re-traced back to the origin showing which publishers were the fraudulent ones and which weren’t.
However, with SKAD, to conduct the type of fraud as seen above, the ad networks are not only involved – they are some of the ones that are doing it. In order to do what the video above shows – the SDK of the Ad Network is the one controlling how the ad is displayed, and how the app store page gets triggered.
Without the option of even knowing how many “clicks” (app store popups) happened, advertisers are effectively blind to know about the portion of spend they are running that is only showing up as “good performance” by being falsely attributed.
The bad news is: not much. There are no “SKAD anti-fraud” solutions out there, as the whole point of SKAdNetwork was to limit access to user-level data, making traditional anti-fraud solutions obsolete.
You can of course demand contractual clauses in the agreements with ad networks that ban the use of automatic app store popup , but I have a feeling that the networks who are already engaged in such actions are likely not respecting contractual terms.
You could do an easy incrementality test. Stop a network for a week in a major country. See if there is an effect over the total number of conversions you have during the week. It’s not a perfect test, but it can serve the purpose.
You can also do your research – watch ads, see how they behave, and which networks are systematically popping the app store page without users’ action, vs. the ones who only do this upon user click.
Fraud is a hassle to deal with. No one likes admitting that any part of their ad spend went to fraudulent inventories, however, this is an unfortunate part of the industry that we face regularly.
Someone once said that doing fraud in mobile ad tech is easier, more lucrative, and more appealing than robbing a bank. There never seems to be any jail time, even for those that billed millions, if not hundreds of millions, for traffic without any value.
Fraud was one of the inspirations to why I started INCRMNTAL. Fraud has no incremental value.