PureSpectrum implemented PureScore™, a machine learning driven scoring system that rates individual respondent quality on a scale of 0-10, at the beginning of 2020. It represents the next evolution in data quality technology, working to keep up with the constantly shifting landscape. 

Since then, the system has blocked respondents with a failing PureScore™  from taking buyer surveys. The aim is to improve data quality before poorly-ranked respondents can even make it past our screeners.

PureScore™ ensures data reliability, the most inherent risk to online survey research. Buyers will accept completes as long as they meet just a few criteria within the survey, unaware of potential fraud during screening. PureScore tracks answer consistency and captures deviations as they interact with the PureSpectrum platform, keeping serial offenders out of client surveys. 

The model works by finding patterns among behavior including screening consistency, LOI, completion rate, etc. The further that a respondent’s fraud markers and behaviors deviate from the ideal, the lower their PureScore. Only respondents with a passing  PureScore are directed to surveys. 

PureScore has effectively blocked ~3% of all transactions since its implementation in January 2020, leading to a decrease in reconciliation rates. Compared to rates from January 2019, the PureSpectrum platform saw a 15% decrease in rejected completes from researchers. Data reconciliation is an important but isolated data quality measure driven by various methods and policies of the researcher. Nonetheless, PureScore improves upon this costly cleaning process. 

The data also showed that less than 1% of respondents were responsible for 3% of all blocked traffic. Ensuring that these respondents do not get into surveys will lead to continued data quality improvement and further reduction in reconciliation rates. 

“This is our first step, applying innovative technologies and scientific approaches for fraud prevention and quality assurance,” Sushma Vasudevan, VP of Analytics and Data Science, said. “The results we are seeing are quite promising, and only get better with machine learning and deep learning.”

Just as the data landscape changes over time, our commitment to data quality means that PureScore is constantly evolving. We are consistently enhancing and improving the model for better business insights on behalf of our buyers.