‘I keep things running smoothly between modelers and data analysts’
Ron Keur, Senior Project Manager, Credit Risk Modeling
Ron Keur (36) is the project manager for the team that is developing a new credit risk model for Rabobank. The team collects, interprets, and organizes a huge amount of data for large business clients around the world. The team then uses the data to develop smart algorithms to predict risks.
Sharp, ambitious people
“I’m currently working on a mega-project: building a new credit risk model. We do that together with a team of around 50 employees that includes modelers and data analysts, as well as colleagues who are directly involved in providing services to our clients. They’re all sharp, ambitious people with a wide range of backgrounds and talents. As the project leader, I’m the lubricant between all of the team members. It’s important that they can all communicate effectively with each other, that they understand one another and can clearly explain complex subjects. My responsibility is to make sure that happens.”
Estimating risks to stay strong
“Credit risk models are vital for a bank, because they let you accurately estimate whether a client is credit-worthy or presents too great a risk to the bank. For example, how likely is it that a large commercial client won’t be able to pay back a loan? And if that happens, how much of the loan will we lose? Accurate predictions help keep us reliable and strong as a bank, so it’s understandable that so many people at Rabobank are looking over our shoulders to see how the project is progressing. We think that’s cool, because it underscores that we’re doing extremely relevant work.”
Keeping the banking world safe
“Another group that are looking over our shoulders are the supervisory bodies for the financial world. So the team is laser-focused on only using the data that we’re allowed to use for that purpose. Strict regulations have been drawn up to that effect, which is only natural because those regulations protect our clients and keep the banking world safe. The challenge for us, then, is to build a credit risk model that’s accurate and statistically supported within the framework of all of those conditions.”
Past is prologue
“In order to develop a model like that, we need vast quantities of customer information from the past. That’s the power of data: we can make reliable predictions about the future based on data from the past. But first we have to collect, correctly interpret, and organize the data. That’s where our data analysts come in. They can present useful data sets that our modelers can use to build the statistical model. To do that, our data analysts utilize SQL Server, Python, and dashboards to quickly show which data is available, and to provide insight into the quality of the data and how it will be interpreted during the model construction phase.”
A model you can trust blindly
Our customer advisors, account managers and financing specialists want to present our clients with an attractive offer for financing. If the model doesn’t work properly, and clients are estimated to be riskier than they actually are, then we’ll have to ask a relatively high price for our products. And then there’s a good chance that the clients will go to another bank to get a better deal. But if the model estimates the risk as lower than it really is, then we won’t comply with the supervisory body’s regulations. So it’s very important that we build a model we can trust blindly. If we can do that, then we’ll be able to help our clients faster and more efficiently.”