The tool, called Second Shield, provides a layer of machine learning defence that sits between Rep Data’s Research Defender fraud prevention software and live fieldwork.

Second Shield aims to identify multivariate patterns that signal fraud and then block fraudsters, and uses a continuously refreshed 180‑day reconciliation warehouse, decision‑tree‑based models and anomaly‑detection techniques.

The tool also learns from every survey response that is later overturned in reconciliation, with each reversal then becoming a labelled training point feeding a classifier that refreshes nightly across DIY sampling tool Research Desk and Research Defender traffic. 

Second Shield draws on signals such as open‑end length, device/OS, geo data, timestamps, supplier and historical behaviour across nearly 200 sample sources, and can alert Rep Data’s services team in the instance of a predicted‑fraud threshold, so that supply can be paused, switched or price‑adjusted before issues escalate.

Future updates will allow for autonomous threshold tuning, meaning API connections will allow Research Defender to adjust, auto‑tighten settings or swap suppliers in the background.

Pat Stokes, founder and chief executive at Rep Data, said: “Second Shield transforms closed‑loop data from our DIY sampling platform, Research Desk, and Research Defender into predictive power, blocking the sophisticated, hard‑to‑spot fraud patterns that slip past the current limits of fraud prevention technology.”