Poor data quality is a major problem for mobile advertisers, and marketers are forced to bear the costs of this fraud.
While the issue of mobile data cleanliness and efficacy has been the subject of much anecdotal commentary, there has been scant scientific evidence of the true scale of the problem, or of its root causes.
As a consumer intelligence platform able to store, access and analyze more than two years’ worth of mobile data, Mobilewalla has generated exclusive insights into the nature and extent of mobile ad traffic. The goal: Understand the source of mobile ad fraud so marketers and others in the industry can more strategically combat the issue. Here’s a look at those findings:
The major sources of ad fraud
There are four major sources of ad fraud:
Mobile Advertiser ID (MAID): Marketers use MAIDs to determine if they have already served an ad to a specific user, or to retarget or frequency cap the ads shown to that user. MAIDs, however, are often fraudulent. The identifier available in the signal or bid request may not actually refer to a real user. Sometimes this occurs because the device itself is fraudulently generated. Alternatively, the app from which the bid request originates may also be counterfeit or infected by malware, in which case the device is considered anomalous.
Location: As much as 70 percent of location values passed through the bid stream could be inaccurate. When the requesting app is unable to get a lat/long value from the GPS of the device (this commonly occurs because the functionality is turned off by the user), the attribute is typically estimated by reverse-mapping the IP address passed in the originating request. Reverse-mapped locations are highly inaccurate when it comes to local points of interest such as sports stadium, stores, restaurants or specific addresses that advertisers need to know for this kind of targeting.
Media source: The source app of a bid request is a powerful indicator of the demographics and behavior of the consumer using a device; it is one of the most important data elements used to assess the value of a bid request for a campaign. Yet, between 25 and 67 percent of media sources in bid requests are blinded, meaning the values passed are not helpful in tracing the request back to a known app.
Network connectivity: Inaccurate IP addresses are a well-known culprit in fraudulent programmatic supply. Nefarious parties that purposely generate fraudulent users/MAIDs must indicate source IPs for them, leading to the inclusion of forged IPs in the bid stream.
The scale of device and location fraud
Device and location fraud is global, and it can vary widely by location or time. There’s an ebb and flow of fraudulent data in all geographies. For our analysis, we observed bid traffic in two randomly chosen large global SSPs. The patterns reported represent behavior across the entire programmatic traffic spectrum.
Let’s start with fraudulent devices. The average number of fraudulent devices observed globally in the last 90 days is 25 percent, but the distribution varies widely across geographies: the Netherlands only had 4 percent, while Brazil saw 41 percent of its active devices fall into the fraudulent category. In the U.S., the figure was 5 percent.
One thing to keep in mind is that the scale of fraudulent devices varies widely and with time. While it is tempting to classify certain geographies as more fraud-prone, our analysis suggests otherwise. In the U.S., for instance, we have seen sudden spikes in bid reports coming from two-to-three times the number of devices that actually exist. In other words, fraud occurs irrespective of geography.
Now, let’s consider anomalous locations passed in mobile programmatic supply. Globally, about 13 percent of location signals, on average, were fraudulent, with a wide distribution of anomaly percentages across different geographies. Interestingly, the U.S. turns out to have by far the highest fraud percentage of any geography, and this pattern is something we’ve observed persistently.
Mobilewalla has been tracking fraudulent device and location signals, as well as other fraudulent patterns, in mobile ad traffic over the past two years and has become one of the most trusted identifiers of anomalous ad traffic data in the industry. Only by analyzing data over time can indicators of fraud be identified and removed from the supply, even as other sources of fraud rush to fill the void. With this persistent approach to fraud detection, marketers can stay a step ahead of these nefarious elements, saving themselves time, money, and valuable impressions.
