CAAF Spotlight: ironSource shines a light on how they tackle mobile ad fraud
Senior Content Manager
Feb 21, 2018
Continuing our CAAF spotlight, we got in contact with Yevgeny Peres, VP of Growth at ironSource, to find out more about their perspective on mobile ad fraud, and their ongoing membership in Adjust’s Coalition Against Ad Fraud.
The initiative has been a great success since its launch, as we continue to work together with our partners to take on fraud in the mobile ad ecosystem.
Didn’t read our first in the spotlight series with Liftoff? Click here to catch up. And for more on ad fraud more generally, head on over to our mobile fraud guide - where we run through the most common types of fraud and how they affect your campaigns.
What was the biggest breakthrough ironSource has made to fight fraud?
In the past year we decided to double-down on our fraud research, which fell into two main categories:
The first - how do we keep our SDK network clean from publishers trying to gain access to campaigns and manipulate ad serving and attribution?
This part of our research and analysis drove us to a completely new publisher onboarding process, focusing on premium publishers and implementing even more stringent validation processes, which evaluate their SDK integrations in full. This not only adds an improved security layer, but also a viewability verification layer which ensures that the full ad-serving funnel is completely secure and under our control as the publisher goes live, and every time our SDK calls for ads. These improved security features on the publisher side, combined with our proprietary anomaly detection, and CPA/ROAS optimization, ensure that things stay fraud-free from the advertiser perspective.
The second - how can we fight ad fraud outside of our network?
Over the last nine months we’ve developed technological capabilities which enable us to identify whether or not an advertiser, while running a campaign with ironSource, has been impacted by ad fraud from outside of the ironSource network - i.e. from buying on higher risk channels like affiliate networks, DSPs or less secure SDK networks. By intercepting these fraudulent instances out in the wild, combined with a full analysis of the overall paid and organic installs driven to an advertiser, we’re now able to measure the impact of ad fraud per individual advertiser, and provide tools and a proven framework to eliminate that fraud.
What were the first signs for ironSource that fraud was affecting campaigns?
What triggered this new angle of research were instances of video campaigns running on our network seeing significant drops in volumes and eCPM without any changes to bids, creatives, targeting, or anything in the competitive landscape.
Witnessing this phenomenon made it clear that even though we’ve been able to maintain a clean and fraud-free network, if an advertiser is buying on fraudulent channels, they are giving bad actors destructive access to their campaigns. This can lead to widespread attribution manipulation and, ultimately, organic installs being flagged as paid, and paid installs not attributed to the last touch point with the user.
The results we saw were so shocking we realized we had to act on this information. We have since been running a series of workshops and educational sessions for all our partners with an emphasis on what processes and framework can easily be implemented to tackle this issue.
Beyond lost campaign spend, how have you seen mobile ad fraud impact the industry?
The damages we’ve seen go far beyond just lost spend. When it comes to the biggest type of fraud, which is focused on manipulating the attribution of real users and which involves a click event reported from a user who hasn’t actually engaged with an ad at all, or hasn’t engaged with the ad associated with the last-click URL - there are four main issues.
- A significant amount of spend is wasted on paid users who have never seen an ad, i.e. organic users. But beyond the wasted spend, this also causes a fundamentally skewed understanding of an advertiser’s paid vs. organic ratio and the creative reporting associated with the misreported clicks.
- Attribution manipulation not only cannibalizes organics and colors them as paid users, it also takes credit away from legitimate sources driving users through ad engagement. Since a different entity wins last click attribution, the channel that really drove the user will not convert as well in terms of eCPM. With time, the volume of users seeing your ads on that channel - a channel that could be bringing you great users - will start to decrease.
- Click spamming and click injection-driven installs will have amazing downstream performance, since they are most often the closest to (real) organic users and therefore approximate their behavior. Since ads aren’t actually shown in those cases, advertisers can significantly drop their bids, which will ultimately drive much higher ROI.
- Our research has also shown how some ad networks and DSPs take an active part in manipulating attribution by misreporting clicks in order to increase the attribution window and attribution footprint. The most common practice we’ve seen is ad impressions or video views reported as clicks. This means that most of the clicks in a campaign aren’t really clicks. The intent and time distribution of last-click-to -first-app-open is similar to click spamming channels. Other more complex implementations have included forced auto-redirecting ad units and manipulated clicks forcing fingerprinting attribution.
Unfortunately, this is a catch-22 situation, which most marketers struggle with even once they realize that most of the suspicious traffic is indeed fraudulent.
When marketers look at the attribution picture, their best-performing channels are actually fraudulent ones which don’t engage real users through ads, but rather manipulate attribution to win the last click. This is dangerous since with time, as the share of spend of fraudulent channels continues to grow (because of a misunderstanding of channel effectiveness), the overall install growth will actually slow, as fewer ads are shown to drive actual user engagement.
The situation is further complicated by the fact that marketers are often also using self-attributing channels, adding more obscurity to the overall attribution picture. Savvy marketers must base their evaluations of campaign performance on data gathered from independent attribution platforms - where the playing field is level - in order to access a true understanding of campaign performance.
The difficulty is that no matter how savvy marketers may be, if they are buying on fraudulent channels, the accuracy of the entire attribution picture is compromised. The cost of fraud is not only the wasted spend in the here and now, but the erroneous understanding marketers have of their channel performance, which will inform (wrong) decision-making in the long-term.
How do your research findings impact the ad monetization side of the equation?
All too often, the discussion around fraud centers on lost spend and not on lost revenue, but this is an issue that meaningfully and significantly impacts both sides of the mobile app equation - monetization and marketing.
As a leading video ad monetization platform, it's our mission that premium developers who integrate our SDK with the purpose of maximizing their ARPU and retention metrics get the attribution credit they deserve if a user engages with an ad in their app.
Yet, our data and full-funnel analysis has shown that our top publishers are losing at least 22% of their ad revenues, mainly due to attribution manipulation executed by other fraudulent channels like gaming attribution by stealing the last click, even though a user engaged with the ad shown by the publisher last. By stealing the attribution credit for that install and that user, the fraudulent channel effectively takes money from a publisher who rightfully deserves credit and payment.
What does ironSource do to proactively fight fraud in its systems?
The first step is controlling the ad serving funnel completely and absolutely. From the level of our ad mediation platform and down onto the network level, we’ve built a new publisher onboarding process which includes a thorough business review and technical due diligence to ensure that our SDK has been integrated correctly by a premium publisher, before we allow them access to ad serving.
Once a publisher is in, we then monitor the standard industry metrics. We focus on the user intent and its origin, along with a few proprietary anomaly detection solutions (which focus on revealing in real time whether there’s any suspicious ad request/user/publisher behavior). If anything is flagged as suspicious, ads won’t be served and an alert is immediately triggered to our anti-fraud team who investigate immediately.
To make sure all our bases (and our advertisers’ bases) are covered, we also use our proprietary solution to monitor per advertiser, per campaign for suspicious activity outside of our network and alert the advertiser of anything flagged as suspicious.
How does ironSource go beyond to help advertisers when it comes to education?
Fraud is a constantly-evolving problem, and we believe that a core component to fighting it is to invest heavily in ongoing research that goes beyond just our network and looks at the industry as a whole. We consistently share these findings with our advertiser partners, whether in formal one-on-one workshops, industry meetups in different local markets, or in on-the-fly reports on specific campaigns. Our goal is to increase awareness by being as transparent and inclusive as possible with our findings - from attribution partners, to platforms and of course advertisers themselves - whether they’re currently engaging with fraudulent channels or may be about to.
What motivated ironSource to join the Coalition Against Ad Fraud?
The main motivation was the common goal that CAAF has to fight fraud, and since ironSource has the resources and the passion to try and solve the issue, it was a no-brainer. It should be clear to everyone in the industry that we have to come together to fight a problem of this magnitude and prevalence, and that’s what makes initiatives like the CAAF so important.
What's been the best thing about joining CAAF?
The biggest benefit we’ve seen from joining the CAAF has been the open channel it has created between industry players. The level of knowledge sharing and communication - whether that’s during global meetups or day-to-day on the Slack channel - has been a real gamechanger.
Do you have any predictions on how ad fraud will develop in 2018?
I think we have to be cautious about sharing predictions around the future of fraud, since we’ve learned how sharing details around fraud detection and prevention can easily translate to new fraud schemes on the part of bad actors in our ecosystem.
What I can say is that we’re not going to see fraud fade or decrease at all over the coming year. Instead, I think we’re going to see current fraudulent practices evolve into increasingly sophisticated new methods to game attribution, or steal credit. This means that every marketer should seriously invest in setting up the right processes internally to make sure that their tracking URLs can’t be accessed by fraudsters. I think the key to doing that is to master attribution and ad serving, and then ultimately think like your enemy. Identifying the possible loopholes or dark spots in the ad serving funnel and thinking like a fraudster will ensure you’re always one step ahead.
With CAAF, our aim is to educate the industry about the problems of ad fraud - because education is the first step to getting rid of it for good. Interested in learning more about the issue? Start by reading our mobile fraud guide here.