Does fraud have a pattern? Inside the research Rep Data brought to ESOMAR 2026

Rep Data Team · September 15, 2026 · 2 min read

Fraud detection often starts with the idea that bad respondents will reveal themselves in recognizable ways. They might fail a behavioral check, show suspicious device activity, provide a poor open-ended response or trigger another signal that indicates something isn't right.

Rep Data wanted to know whether those signals add up to a recognizable fraud profile. We analyzed 11,778 respondents across eight quality signals, looking at whether different checks consistently identified the same people and what that could tell us about how fraud should be detected. Our team presented the findings at Esomar Congress 2026 earlier this month in Valencia, Spain.

Want to dig into the data? Read the full research: Does Fraud Have a Pattern?

There isn't one fraud profile

The research did not identify a distinct type of respondent who consistently triggered multiple quality signals. Poor-quality respondents don't all behave the same way, and the characteristics that identify one suspicious respondent may be very different from those that identify another.

A respondent can also look acceptable through one quality check and raise concerns through another, making it difficult to define fraud through a predictable set of behaviors or characteristics.

Different signals catch different people

Across the eight quality signals we studied, there was near-zero correlation in the respondents they flagged. Someone who passed one check could still be caught by another, whether through device signals, behavior or the quality of their responses. Different checks picked up different problems, making it risky to rely on any single signal to judge respondent quality.

Quality needs more than one layer

The findings make a strong case for evaluating quality in multiple ways and at multiple points in the research process. Signals available before someone enters a survey can identify certain risks, while in-survey behavior and the responses they ultimately provide can reveal others.

No individual check captured everything we found across the eight signals. Researchers need to understand which quality checks are being used, where they are applied and what each is designed to identify.

We brought this research to Esomar Congress because these are questions the industry is actively working through as fraud becomes more sophisticated and approaches to data quality continue to develop. Let’s continue the conversation: https://repdata.com/contact

There may not be one recognizable fraud profile, but the lack of a pattern tells us something important in itself. Protecting research quality requires looking at respondents from multiple angles rather than relying on any one signal to determine whether their data can be trusted.