Quality Corner | December 2023 | Key Terms to Know in the Battle Against Fraud
Blog | November 27, 2023
by Terence McCarron
To keep online survey data trustworthy, the core task on our collective plate is to separate genuine human answers from fake or fraudulent ones as efficiently and accurately as possible. CleanID™, our proprietary fraud and duplication mitigation system, helps us do that by using clear rules and details, then showing our customers precisely what’s happening.
The decision to share the detail was a game-changer for researchers. Previous market options used a Blackbox approach, attributing this to complexity or “intellectual property”. We believed that wasn’t sufficient and decided to more this data quality effort a collaboration between us and Market Research companies.
Our Quality Corner will continue our philosophy with a deeper dive into key topics. Here we break down a key component related to fraud.
Invalid Traffic Type
In the world of fraud identification, there is an umbrella concept used to classify the severity of threats. ‘IVT’ is an acronym for “Invalid Traffic Type”. IVT is not an OpinionRoute language invention. It’s widely used by experts at the Media Ratings Council (MRC) and apply to any online traffic in any arena: surveys, advertising, ecommerce, etc. The elements that form this construct are important within the survey ecosystem. This is particularly true in the programmatic sample era, which is deeply connected to broader digital marketing techniques.
IVT further breaks down into two subcategories: GIVT (General Invalid Traffic) and SIVT (Sophisticated Invalid Traffic). These definitions are used as elements of the CleanID™ functionality and can be found in the data we share with our MRX clients. In the end, the IVT sign is a big clue that, along with other checks, helps us figure out if a malicious actor is trying to get into your survey.
GIVT & SIVT
GIVT is the simple stuff, like automatic programs (some good bots), basic cheating, and privacy tools that make it hard for us to see who’s who. SIVT is the trickier stuff- with clever tricks used to look like real people. Sometimes these things elements are present because of the internet service provider’s technical setup, or because a real person really wants to keep their privacy. When it comes to GIVT traffic, it’s important to analyze as many device and network attributes as possible to give a clue to the risk we carry. Our algorithm executes this on every click.
How common is this?
Not all ‘fraud’ or bad participants are labeled with an IVT designation, but much of fraud does carry an IVT flag. Digging into the IVT construct offers clues for understanding the fraud types.
When we review what % of total volume have these flags, we see…
IVT as a percentage of Traffic:
- More than 6% of all traffic gets a GIVT designation with (**currently) ~25% of those being evaluated as fraud.
- More than 7% of all traffic gets a SIVT designation with (**currently) ~75% of those being evaluated as fraud.
Now, let’s flip the coin.
Here’s the percentage of Fraud that currently has an IVT designation:
- 72% of all fraud has an IVT designation.
- 28% of all fraud lacks any IVT designation at all.
Clearly, IVT is a major concept in the quest to understand and identify fraud real-time. CleanID™ integrates this designation within a total of 800+ data points from the device and network, then we add in behavioral patterns.
By leveraging the MRC-established GIVT and SIVT designations, CleanID™ ensures the reliability of survey data while also assuring transparency for our clients. We believe this approach fosters trust and confidence in the survey ecosystem in a way that translates well to the clients of Researchers.
**Note: We will always qualify numbers within the fraud space with a ‘currently’ given the evolution of fraud and how we evolve with it. As an example, in 2022, we were designating less than 15% of all GIVT traffic as fraud. Conveniently, the fraudster community picked up on this. Thanks to our full feedback loop, we were able to identify poor in-survey data, then adapt our scoring system for these bad actors to keep them from entering new surveys.