banker01

BBVA Data & Analytics Awarded by The Banker Technology Projects of the Year Award

Fco. Javier López Peñalver and Jordi Nin

d&a blog

Interview with Francisco Javier López Peñalver and Jordi Nin
 
jordi-nin

Jordi Nin — Data Scientist at BBVA Data & Analytics

Jordi holds a Ph.D. in Computer Science, with over 75 research publications in journals, conferences and book chapters. He was granted with the Outstanding Ph.D. Award of the UAB computer science department. Jordi’s thesis was devoted to use several machine learning methods to correctly calculate the disclosure risk of the official statistical surveys published by national statistical agencies.

 

Francisco Javier López — Data Project Manager at BBVA Data & Analytics

Javier holds a master’s degree in Computer Science and Innovation Management. He’s now leading Risk & Fraud analytics projects at BBVA D&A to boost BBVA’s digital transformation. Background as R&D engineer, innovation consultant and international public funding manager. Enthusiast exploring technologies to create new opportunities.
 
 
 

Financial Times awarded BBVA Data & Analytics with The Banker’s Technology Projects of the Year in the category of Risk Management. Francisco Javier López Peñalver (Data Project Manager) and Jordi Nin (Data Scientist) at BBVA Data & Analytics explain the inception of the RedeX project, developed in partnership with BBVA and BEEVA between June 2015 and May 2017.

What is RedeX and why was it awarded?

This project aims to improve the risk assessment method for new and current customers during the loan admission process, and represents a meaningful solution for customers that have a thin credit file . BBVA receives thousands of loan applications each month and many are denied because there is no information about the customer’s credit history, a frequent problem especially among young companies, start-ups and foreigners that have recently accessed the financial system.

Building on different types of data such as bank transfers, official gazettes and interactions with other entities, and without having to request any additional information, RedeX established a very rich source of information about customers’ ecosystem complementing the information provided by customers when they request a new credit product.

What were the objectives of this project?

The objective of this project was threefold: to obtain all necessary data to build a network of relationships; perform distributed relational analytics to extract the most relevant customers’ attributes and the most suitable metrics to analyze the customer context; and build a self service tool to allow risk analysts explore the data, visualize it and obtain new customers metrics.

How was Data Science relevant for this project?

Due to the complex and adaptive nature of the financial system, Complex System Theory provided a mathematical toolbox to analyze Complex Systems at micro, meso and macro scale. A network is a mathematical abstraction that represents systems of interacting entities. Naturally, entities are symbolized by nodes and the interactions by edges or connections between them.

Network Theory and agent-based simulation were useful to understand systemic risk in the banking system, which suggests that default  risk propagates through a network following very complex patterns at macro scale but also mention that such patterns can be decomposed to simpler ones when the network is analyzed at microscopic level.

How is RedeX useful to risk analysts?

RedeX provides a self-service tool with an interactive interface to assist risk analysts  explore the context of customers through graphs and network analysis, assess the level of risk of a customer and make decisions about loan applications. It can be also be applied to other areas such as fraud management, referral systems or user experience in digital channels analysis. The screenshot below demonstrates how easy it is for any risk analyst to access customers’ ecosystem data and use this information to assess the financial health of a customer had thin credit file  (data and names have been obfuscated):

ewsrwg