What if there are no contextual references to guide decisions?
Equifax identified an opportunity in a new sector of the credit markets... This is a segment where traditional credit risk profiles were weak or non-existent. Potential borrowers were out there, but how to identify them and automate the processes so as to be transactionally effective, and as close to real-time as possible?
In order to be able to offer credit to a new tier of customers, Equifax had to go beyond the typical and established criterias. Factors such as distance from a borrower's workplace to their home, the number of years they had lived in the same location etc. had proven to be effective in determining credit worthiness...but the data was simply not readily available.
AQUIOM identified sources of unusual data. Some sources were scraped from the internet and others from public records. The solutions were innovative and resourceful. The information had to be acquired and validated in real time, without much human interaction.
The scope of the project was well outside of the realm of most typical data science initiatives.
While longterm results will be the ultimately test of the data systems and analytics, the short term metrics are very promising. These are lending channels and customers that constitute new markets. Equifax's competitors don't even have an answer to this. So revenues were almost purely additive to the bottom line.
See Risk analysis below.
STORY OF THE GRAPHIC VISUALIZATION: showing the outcome and where metrics have indicated the result.
Unusual data types, and unique systems to validate and interpret them
MACHINE LEARNING OPTIMIZING UNSTRUCTURED DATA
Multiple sources of unstructured data were gathered, validated and analyzed using unique criteria. Machine learning systems optimize accuracy, and enhance the criteria. Not only are the systems designed to introduce new business opportunities, opening up new markets to credit offerings - but they are a new model of efficiency and compliance. One of the most attractive aspects of the systems are the way they are designed to self-optimize: the algorithms take aggregated data and results from each transaction to make the systems stronger, smarter and better at selecting good candidates. Speeding transactions, minimizing underwriting costs and costly errors.