A full 25% of device identifiers guaranteeing specific consumer data sets are invalid, according to research by data exchange platform BDEX. Much of this bad data stems from ad fraud activities, such as spoofing, bots and click farms, but also human error, with one of the leading email offenders BDEX found being firstname.lastname@example.org.
BDEX analysts used data from more than 100 different suppliers to cross-reference mobile advertising IDs (MAID), hashed email (MD5), and IP addresses and compared these data points with other identifiers and location information in order to evaluate the validity and quality of the data.
Fraud is prevalent across all forms of so-called personalized identifiers. A full 20% of MAIDs sold in the U.S. data market are invalid and 21% of email MD5s are linked to more than 10 MAIDs, a sign of fraudulent information.
Researchers also discovered 2% of consumer IP addresses sold in the U.S. are from non-U.S. countries, U.S. governmental IP addresses, or are otherwise invalid. In addition, 1.1% of email MD5s sold in the U.S. are invalid.
“Ad-fraud will continue to become more advanced, as fraudsters find creative ways to evade detection,” says David Finkelstein, CEO/co-founder, BDEX. “However, big data has also become a robust tool powerful enough to combat it.”
The report suggests several key actions marketing teams can take to help avoid ad fraud ,including being aware that 95% of all IP addresses have six or fewer email addresses associated with them, and 95% of all email addresses have four or fewer IP addresses associated with them. Hence, companies should be very wary of IP addresses associated with more than these numbers.
Also while not a direct indicator of fraud, the use of proxy servers can be a red flag for fraudulent activity, and companies should consider this factor when evaluating the validity of an identifier.
And fraudulent identifiers may have obvious tells, such as accounts linking thousands of pages, unnaturally quick traffic on web pages or excessive traffic for extended periods of time.
See more from the report here.