To conduct the analysis, we reached out to their borrowers in various states of repayment and asked them to opt-in to the Neener analytics app. From all of the respondents that registered their various social media accounts with Neener Analytics two different groups were created for analysis. The first group consisted of 75% of the collected records with an 80/20 split between good borrowers and those behind in their payments. We then tested the trained our AI models for predicting default using the remaining 25% of records. On this held-out group we found that our models could correctly identify 35% of the individuals that were Late Pays or in Default. We also had a "false positive" rate of 10% (those that are Paying-As-Agreed, but we predicted as Default). But false positive doesn't necessarily mean wrong. Because we predict default (not success) a false positive suggests individuals that may default, miss-a-payment, late-payment or generally exhibit default outcomes that may not have yet manifested during the POC period.