Steven Snell on the fight against sophisticated fraud in market research

The integrity of market research data is under threat. Fraudulent survey responses have become more sophisticated, making it increasingly difficult for researchers to separate good data from bad. In a recent episode of Now That’s Significant, Steven Snell, PhD, Head of Research at Rep Data, sheds light on the evolving landscape of fraud in market research and what the industry must do to combat it.

The rise of “good-looking fraud”

Fraud in market research is no longer easy to spot. Traditional methods like attention checks and open-ended response analysis are often ineffective because fraudsters have adapted. As Snell explains, “If it was patently obvious what was bad, we could throw it out and move on. But in reality… tech makes fraud possible, and so we need tech to fight fraud.”

Many fraudulent responses appear completely normal, blending seamlessly into datasets. In fact, Snell’s research suggests that “as much as 10% or 20% of fraud gets through data cleaning because it looks good on the surface.”

Why does fraud matter?

Some may argue that if fraudulent responses still yield positive business outcomes, there is no real harm. Snell pushes back on this idea: “Not all bad data is fraud, but all fraud is bad data, even if it looks good.” Fraudulent responses distort key business metrics, leading to misguided decisions. He elaborates: “If it comes from a fraudster, if it comes from a click farm, a bot, a bad actor, we want it out of the dataset, even if it passes the initial sniff test of data cleaning or open-ended analysis.”

One of the most concerning trends is how fraud skews key performance indicators (KPIs). Rather than introducing extreme or random responses, fraudsters often pick moderate, non-controversial options. This creates a situation where brand awareness and perception data become artificially compressed toward the middle. Fraudulent responses dilute meaningful differentiation between well-known and lesser-known brands, making it harder to extract actionable insights.

Taking control of data quality

The responsibility for ensuring data integrity is increasingly falling on individual researchers. With the democratization of research tools, more people are designing and running their own surveys, but not everyone is equipped to detect fraud. According to Snell, “Everyone has an interest in increasing data quality… but the researcher themselves at the end of the line is ultimately responsible.”

Rather than relying solely on post-survey cleaning methods, researchers need to be proactive, using technology to detect and block fraud before it ever reaches the dataset. One solution is leveraging metadata and paradata—tracking elements such as IP addresses, device information, typing speed, and copy-pasting behavior. “We see it, we block it before they touch your survey,” Snell explains, referring to the tools available to flag fraudsters in real time.

The future of fraud prevention

Market research fraud prevention is a constantly evolving challenge, requiring innovative solutions. Snell warns against relying too heavily on retrospective data cleaning: “Traditionally, researchers have really focused on after the survey. We’re going to clean our data, we’re going to look at open-ends, we’re maybe going to ask some attention checks and trap questions. And those things have limited effect because the fraudsters are getting better all the time.”

Instead, Snell advocates for a three-stage approach: pre-survey, during-survey, and post-survey fraud detection. This multi-layered approach aims to identify fraudulent respondents before they even enter a survey, monitor suspicious behavior while they are taking it, and refine the dataset after collection.

Staying ahead of fraudsters

As AI and generative models continue to evolve, fraudsters will find new ways to infiltrate market research studies. However, the industry is fighting back with increasingly sophisticated fraud prevention technology. The goal, as Snell puts it, is to make market research an unappealing target for fraud: “Hopefully, we get sophisticated enough… that the fraudsters figure out it's not worth their time. Let’s go defraud some other industry."