Risk Management in the Age of Big Data

By Renato Fiorini, Risk Management Specialist at SAS

 
New solutions involving Big Data technologies are already impacting the way financial institutions handle their business. Obtaining faster results and making data available almost in real time make it possible, for example, to calculate the risk position several times a day, not just once, as traditionally happens. In this way, the bank can redefine its strategy in real time in the face of adverse events.
 
High-performance technologies allow the calculation of the risk position of the entire portfolio with thousands of risk factors in a few minutes or even seconds. And it's easy to see how useful this skill is in a crisis situation, like in 2008, when markets were crashing. In these cases, financial institutions have little time to make decisions, such as: “What to sell first?” “What assets to keep?” “To what value?”
 
This agility made possible by Big Data brings the possibility of orchestrating and executing a coherent strategy, instead of unplanned actions motivated, basically, by the emotions of the moment. Of course, such a skill is not only relevant in moments of crisis. With greater speed and precision in the calculations, banks can operate with a smaller liquidity cushion and thus seek greater profitability even in times of calm.
 
In the area of credit risk, high-performance technologies also make a difference, as just developing a good predictive model is not enough. It is important to be able to quickly implement this model in the daily operation, comparing credit proposals with updated data. This time between preparing the data, developing the predictive model, validating the model and deploying it to the operation is known as the modeling cycle.
 
Currently, the largest Brazilian banks have modeling cycles between 9 and 18 months. This means that in many cases, after identifying that a model needs to be replaced, it takes more than a year to do so. That is, more than 12 months running a lower quality model and suffering from higher default rates than would be possible through a more efficient process.
 
With changes in interest rates, exchange rates, household indebtedness, among other factors, old predictive models may no longer be adequate. It is important to be able to react quickly and develop and implement models that are suitable for this new reality. With high performance technologies it is possible to reduce this cycle to three months. And this can represent a reduction of up to 5% in credit risk provisioning. Technologies such as these allow preventive actions by banks, avoiding default and contributing to the proper pricing of loans.