Fraud Puts a Premium on Trust for DSPs
In the eyes of mobile advertisers, programmatic offers tremendous upside and considerable risk. A 2016 report by ANA estimated $7.2B in ad fraud the previous year during a period where total spend outside of Facebook and Google was $12B. Despite growing awareness about the risks fraud presents for advertisers, programmatic continues to grow - particularly in mobile.
Given the pervasiveness and growing awareness of fraud, trust could be the most valuable commodity for DSPs. Advertisers with the resources to invest in robust fraud detection are more likely to be the biggest spenders. This dynamic makes establishing and retaining trust with a DSPs most important customers even more tenuous and important.
Further complicating the issue, DSPs are hard pressed to establish trust through transparency due to the necessarily opaque nature of their business. Offering too much information to advertisers to put their minds at ease can mean revealing the “secret sauce” that a DSP relies on to differentiate in a fiercely competitive space.
How to Build Trust & Combat Mobile Ad Fraud
Each time an advertiser detects fraud, perceived risk compounds. If that becomes a pattern overtime, the likelihood of an advertisers divesting from your DSP or DSPs altogether increases.
Current solutions aim to remedy fraud after an impression is already delivered and trust is already tarnished. These reactive tactics are not sufficient in mitigating against the ever-evolving tactics of bad actors in the space. The best complement to a robust blacklisting capability is a more informed approach to targeting.
DSPs are particularly vulnerable to fraud because they tend to rely on post-click optimization. Without access to signals beyond their database that can verify the identity of a (valid) human behind a given device ID, it’s nearly impossible to take a proactive role in mitigating against fraud.
A Two-Prong Approach to Mitigating Against Fraud Pre-Click
To effectively combat fraud, DSPs should consider approaches that mitigate risk pre-click. For years, Appnique has applied it’s NLP, AI, and Machine Learning technology to the app store to establish a robust affinity graph and improve ad performance on Facebook. Now we're applying the affinity graph to DSP performance and believe our model combats fraud pre-click in two major ways:
Analyze Historical Behavior
In addition to tracking post-click behavior to identify anomalies indicative of bad actors, DSPs must consider behavior that takes place before a click that might signal fraud and weed them out preemptively. App store and DSP log data in combination can provide insight into cross-installs, reviews, timelines and affinity. The richer and more established the data profile for a "user", the less likely that it’s a bot.
Advanced modeling from Appnique considers a rich and diverse set of signals beyond a DSP's database. It then marries those signals to unique device IDs. While the barriers for bots to emulate basic qualities to appear human are relatively low, it is significantly more challenging to replicate a signature combination of interests and behaviors that would qualify them for an affinity-based audience segment. Furthermore, a given audience is often comprised of several unique segments. For bot fraud to be lucrative, it would need to build a high volume of profiles to scale.
Fraud is not going away anytime soon. DSPs have an opportunity to distinguish themselves in a crowded marketplace in the midst of exploding revenue potential by doing more to build trust with advertisers. That means being proactive in addressing fraud before it affects their advertisers. Tools that catch fraud once it has occurred is one half of the equation. Now it’s time address the issue pre-pre-bid.