Fraud remains a major problem for the market research industry, with organized networks employing sophisticated technologies to emulate people and devices and evade security checks. Given Cint’s global footprint, we see and block these attacks daily.
With data quality becoming an even more urgent industry challenge, and AI stepping into the spotlight, we must harness the power of new technologies positively; collaborating effectively to fight these threats together.
Leveraging AI to detect fraud
To this end, we now unveil an important component of our multifaceted response to this continued threat: Cint’s Trust Score – a proprietary, scoring model which uses the power of machine learning and advanced AI algorithms to predict when a session may result in a reconciliation.
A fresh aspect of our holistic quality initiatives which include multiple, targeted screening and detection solutions, Cint has launched Trust Score to maximize quality and minimize fraudulent activity across our research ecosystem, and to reduce the number of reconciliations due to invalid completes.
How does the Trust Score work?
Cint Trust Score’s predictive logic rates respondents to gauge authenticity; terminating sessions if a score is indicative of fraud. We have been rolling it out since the summer, starting with the US, Germany, and the UK, before expanding it globally – having noted a 15% reduction in the overall reconciliation rate in the initial markets.
Cint Trust Score is distinct from scoring efforts made by other market research providers, which typically assess respondents based on past behavioral patterns. In contrast, Trust Score operates in a constant state of readiness thanks to adaptive machine learning which identifies irregular patterns in rapidly changing data sets, and learns accordingly. When a respondent returns as part of a new session, they’re scored again, with the predictive model taking into consideration a range of factors including IP and device related information; profile information; and session level information, such as survey attempts.
A dynamic learning model based on cutting-edge data science, Cint’s Trust Score is having a significant impact on catching bad actors.
Our continued investment in data quality
We’ve learned a great deal over the past months and years when it comes to detecting and taking action on fraud. But, as an industry, we’re still fighting a very difficult war. And so our dedicated executive team meets each week to look at trends in the data – encompassing customer feedback and potential next steps. Our goals are threefold.
- Fortification: To make our platform resilient to fraud attacks by securing the entire supply chain workflow;
- Intelligence: To be even smarter by leveraging our vast data and tech infrastructure, via AI prediction models, advanced analytics and enhanced data signals; and
- Scalability: To create value through automation and scalability.
Cint Trust Score is a great example of this three-pronged approach. As one of the largest programmatic marketplaces in the industry, we have access to a vast amount of data which can feed AI prediction models. With relentless focus, we aim to enhance verification, optimize systems and enhance quality checks – and we intend to expand Cint Trust Score yet further, with incremental releases and fresh innovations happening all the time.
With this data model growing and learning continuously, we’re continuing to invest in predictive technologies to understand and act on fraud, with proactive quality monitoring through data-driven product design.
Reach out to our team for more information. As the nucleus of automated digital market research that connects organizations to the world’s largest choice of trusted consumer opinions and data, we ask you to work alongside us, and to report bad data as we continue to optimize this dynamic, AI-driven model. We are in this together.