Switching to the underwriting process, I'm curious, especially given your background from Capital One, how would you compare the maturity of the lending process and strategies?

I would say that Nubank actually took Capital One's core idea and made it even more applicable and effective. Being digital-first allowed them to start natively. Additionally, Brazil's relatively liberal regulatory environment regarding data usage for underwriting helped. Nubank was able to iterate models and underwriting strategies more rapidly than any other company I've seen in the lending business worldwide. They are truly world-class, especially in Brazil's underwriting front.

It seems that each part of the playbook is replicable, but the cultural element of speed is very rare. You have an ever-evolving credit process, but the credit model gives you a probability of default or a risk measure. Then you have to make decisions and overlay that with a risk appetite or some sort of rule to decide on extending credit or expanding limits. Could you elaborate on how that interplay occurs?

Yes, the decision criteria are relatively constant. The goal is to make decisions that are NPV positive, subject to some level of risk worsening on the margin. The amount of risk worsening might change slightly depending on the risk appetite. For example, you might want a 10%, 30%, or 50% cushion, which can vary based on macroeconomic conditions. But the core is having a good estimate of the lifetime unit economics of the data point. It's straightforward math

But regarding the model or risk appetite? I understand why he wouldn't have a view.

No, he wouldn't weigh in on the model, and he wouldn't weigh in on the risk appetite. I think he respected the process. That's a core principle at Capital One - don't interfere with credit, at that can lead to negative outcomes. There is respect for the level of the unit economics and allowing the process to determine the outcomes in terms of lending.