According to the Office of the Inspector General, Fraud is defined as the “wrongful or criminal deception intended to result in financial or personal gain.” There are different types of fraud. There is fraud committed with the intention of making money, such as insurance fraud, account takeover fraud, investment-related fraud, social security fraud, credit card fraud, and return fraud. Fraud for personal gain is another type of fraud like impersonating a famous person or falsifying academic and/or professional accomplishments.
Organizations globally lose trillions of dollars due to fraud. A 2014 report referenced in fraud.com’s article, “The Actual Cost of Fraud”, estimates that fraud is responsible for a 5% loss of revenues for businesses annually, along with global losses of $3.7 trillion. This figure is greater than the annual GDP of most countries! A 2022 LexisNexis study reports that fraud costs for U.S. retail and e-commerce merchants have risen over 19.8% since 2019. U.S. retail isn’t the only sector being impacted by fraud; financial services, leading firms, and real estate businesses are also dealing with a substantial increase in fraud attacks. These statistics are evidence that fraud will continue to be a major challenge in the coming years, especially as businesses navigate a post-COVID world.
Thankfully, when society and organizations become aware of fraud, they fight back. Laws have been passed that define various forms of fraud and the consequences that come along with them, particularly financial fraud. Costly civil litigation is frequently used to address fraud committed for personal gain. These laws help protect organizations, as well as individuals, from fraudulent activities.
Unacast continually investigates location signals to determine if they are good or fraudulent. Bad data can be costly to organizations and businesses. That is why we have developed processes for detecting fraud within location data.
Location Fraud and the Need for Location Data Forensics
Early on, our engineers at Unacast identified fraud in the raw data we were processing. More ads or more data containing location information meant more revenue, regardless of its accuracy, as app publishers shifted to monetizing location data from apps through advertising and selling analytics solutions.
Our business began processing location data from a variety of sources in 2017, and we quickly picked up on suspicious data and potentially fraudulent data in the supply. In order to increase data quality and transparency for our customers, we started to streamline our data validation procedures and introduced our first Location Data Forensic algorithms in the same year. Today, we use over 15 algorithms and over a dozen related forensic flags to improve our data. Some of these algorithms find fraudulent devices and spoofed locations, and these algorithms remove that information from our supply. Other algorithms simply flag suspicious data so that our customers can choose which information is still relevant to their particular analysis.
Is This Suspicious Data Really Bad?
We often get asked the question: “Why keep potentially fraudulent data?” For Unacast there are several reasons:
- The full data set has been requested by the customers so that it can support their algorithms.
- We want to maintain the highest level of data quality for our clients and continuously enhance Unacast’s own algorithms.
- Unacast’s algorithms conservatively favor false negatives over false positives, but some false positives will still happen. Users of the data can form their own opinions when the data is preserved and appropriately flagged.
The majority of data types, including location data, inevitably contain fraud. Detecting fraud can require processing large amounts of data to detect patterns that suggest it is occurring. These significant fraud detection efforts present an opportunity to use pattern detection techniques to also extract useful information from data sets, which can be turned into monetizable products or used as a data-driven mechanism to guide business strategies. At Unacast, we have found these techniques to be of such high value that the information-extraction algorithms we are currently running outnumber our substantial collection of quality and fraud-detection algorithms.
In order to find these patterns for our clients and spare them the time and labor required to validate and enhance the data they already have, Unacast uses machine learning and heuristics algorithms. Additionally, a customer’s own first-party data can also be cleaned and improved using Unacast's validation and forensic algorithms. This can be applied to different kinds of data sets as well.
In the current environment, organizations must take action to combat fraud or risk basing crucial business decisions on inaccurate data. The key is collaborating with a sophisticated analytics provider like Unacast who makes investments in information extraction techniques and data quality. Organizations can save time and money by relying on a dependable analytics partner’s processing skills to identify suspicious data while also gaining access to trustworthy insights.
