Attribution companies vs. fraud detection vendors: Who does it better?

Are you doing all you can to deter ad fraud? With ad fraud costing the industry an estimated $6.5 billion to as high as $19 billion), according to eMarketer, more than ever you have to watch for the signs of stolen budgets and damaged data.

But your work is cut out for you: fraudsters and their deceptive tactics present an enormous issue industry-wide. (Learn more about how ad fraud works and its impacts here.) Ad fraud detection or prevention tools are clearly part of your toolkit (and have been for years), but their importance has increased with the number of new vendors that have emerged with solutions aimed at reducing ad fraud. Many advertisers have been biting — attracted by the promise of detecting and reducing fraud’s damage through analysis of fraudulent tactics. But what are the differences between the two emerging solutions? We take a closer look.

What makes ad fraud detection vendors and attribution companies different?

Postbacks in real-time vs. one month later

Ask yourself: if you had a choice, would you prefer to know your results in real-time or one month later? It will always be the former. Attribution companies have the most updated results and are best-suited to detecting (and stopping) ad fraud as it happens. Adjust, for example, provides CrossInstall and other DSPs with real-time knowledge about which conversions have been withheld for fraud.

Imagine a typical scenario where CrossInstall receives secondhand reports from our client as opposed to right away. This makes our models work off those potentially fraudulent conversions between the action and the report. This translates to non-actionable reports, advertiser money wasted, and difficulty stopping ad fraud in its tracks.

Enormous amounts of data vs. minimal data

Adjust works with an impressive roster of clients that gives the company insight into an enormous amount of impression, click, and post-install data. Meanwhile, a third-party ad fraud detection service (because it receives reports and data from attribution companies) has a limited view of the data.

Take these two scenarios for example:

• Imagine that a third-party vendor’s machine learning, with its minimal subset of data, identifies an anomaly and flags it for fraud. In six cases out of six we recently investigated, the traffic flagged as fraud turned out to be a technical issue. Variables such as new device OS versions, GDPR, human tastes, publisher app mediation, SDK versions, User Agent issues and more could all be the source of the anomaly instead of fraud.
• Another example is when an exchange had a problem with one of their SDKs that caused the click to redirect to the app store after a 5-10 second delay. The frustrated user would tap repeatedly thinking his or her device was frozen. This triggered fraudulent-looking clicks according to one third-party fraud vendor. The supposed click fraud was actually a technical problem with an exchange’s SDK. Without reliable data, this would have been flagged down.

Objective vs. Subjective

What if your fraud prevention service wasn’t being transparent? We’ve experienced vendors that are afraid of releasing their “secret sauce” while partners like Adjust provide neutral third-party transparency and collaboration with an explanation of why a conversion might be considered fraudulent.

Third-party ad fraud vendors can capitalize on incentive-based pricing structures too, which makes it difficult to collaborate with DSPs for objective investigations. Remember that example about the 5-10 second click delay? Third-party fraud detection vendors can become a positive ally for the industry with some work.

For example, in one report, a vendor claimed that over 400 installs appeared fraudulent. However, CrossInstall’s independent investigation showed that some versions of an exchange SDK were sending the native user-agent instead of the webview user-agent. This triggered an alarm on the anti-fraud vendor’s end, though this was a purely technical issue that we were able to resolve after talking to the exchange.

Third-party vendors may claim their ad fraud detection machines are impartial, but they look at only a subset of the full picture making their impartiality closer to inaccuracy.

Partner collaboration to fight fraud

In good news, there is already industry collaboration in place. Companies are partnering in independent investigations and DSPs are receptive to transparency and finding out about rejected installs. CrossInstall’s engineering team leverages fraud signals and detection notes from Adjust to preserve the sanctity of our DSP against fraud.

Standards need to be in place for third-party vendors who should agree to share click IDs, device IDs and adhere to standards decided via industry collaboration for false positive and false negative thresholds. With guidelines and a unified approach, it is possible for third-party vendors to join the good team in this fight.

How does this play into the larger ad fraud prevention picture?

Moving forward, industry collaboration to fight mobile fraud will be necessary. We need to set standards, work together to build a plan of action and hold each and every company accountable for combating fraud. Third-party vendors should contribute to the fight with accurate information in the best interest of the mobile marketplace.

As the number and usage of fraud detection vendors continue to grow, industry-wide commitment becomes more imperative to preserve advertiser safety and the stability of our mobile ecosystem. There is much work to be done, and we need everyone to step up and be part of the solution.

Take a peek at this guide that gives you a step-by-step approach to asking the tough questions: Whodunit: A How to Guide for Detecting Bad Ad Networks.