Big Data Scoring Case Study
Big Data Scoring is an essential tool for use by banks and financial institutions to determine the creditworthiness of individuals based on data secured online. It does this by tapping into the broadest source of information from across the internet, using all publicly available information. This intelligent research brings lending into the digital information age allowing lenders to make informed credit decisions.
Big Data Scoring wants share with you another case study about how big data can improve the accuracy of an in-house scoring model.
2 main results:
(1) 34.7% reduction in credit loss rates
(2) an expensive credit bureau score can be replaced with Big Data Score™
Over the course of more than a year, we applied many layers of big data models to the best-in-class in-house scoring model of one of our clients. For each of their potential borrower, we were able to collect tens of thousands of additional data points from various public data sources (web search, address investigation, online behaviour, etc). The collected data was structured and processed to relevant variables.
The client is a non-bank consumer lender in Central Europe and had so far been relying on a standard approach to underwriting – all borrowers with bad credit history were denied and then an in-house scoring model was applied. The in-house model used data from loan application and a local credit bureau (positive credit data) as inputs. That type of approach is what we’ve seen in use at most lenders in the region.
The in-house scoring model saw a 22% improvement in GINI coefficient after addition of big data. For the lender, it meant an astonishing 2.6 percentage point drop (that is an astonishing 34.7% reduction) in the overall credit loss rate, which saves them hundreds of thousands of euros a year (they accept clients in scoreboards 7-10). What they also could have done was to keep credit loss rate at the same level and instead accept significantly more loan applications.
Comment: percentages in the table indicate credit loss level in each scoreband. Credit loss is defined as the share of issued loan that couldn’t be recovered.
As part of the in-house underwriting model, the lender was also purchasing expensive positive credit info from a local credit bureau. We tested whether that data can be replaced with our Big Data Score™. As a result of the exercise, we proved that the predictive power of Big Data Score is even slightly better than that of the positive credit information. For the lender, it meant they could gain significant savings from dropping an expensive data source. For us, we proved that behavioural big data is at least as good (actually even a bit better) as hard data from a credit bureau.
Want to see how much Big Data can improve your underwriting? Contact our head of sales to find out – Joaquin Gual, +34 629 466 225, email@example.com