Nubank: Mexico Credit Strategy

This is a snippet of the transcript, sign up to read more.

How do the models differ from your prior experience at Citibank?

At Citibank, we used traditional methodologies like logarithmic regression, which are predictable and robust. Nubank, however, uses new techniques like machine learning, which adapts to customers or applications. Traditional banks rely on credit bureau data or internal information, while Nubank captures non-traditional data to explain credit behaviors. Model development at Nubank is fast. Traditional banks like Itaú or Citibank or BBVA take nine to 12 months to read a model, but at Nubank, we can develop and deploy a model in three months due to a robust data platform. Nubank's ability to create scenarios and simulate "what if" scenarios is powerful. Their technology platform allows for experimentation, always testing and discarding what doesn't work.

This is a snippet of the transcript, sign up to read more.

That's a completely different approach from Citibank, right? My understanding is that no one else has this global platform that is being localized. Everyone else has local ERP and local systems.

Yes, which is a nightmare from a global perspective. For example, we have a global FICO license and a decision engine strategyware. Citibank in different countries requires different changes. The strategyware isn't even available in FICO anymore; it's now called Blaze. Citibank has to change many things, and even FICO said, "Okay, it's not my child anymore," because it's so different from its inception. Any adaptation project takes a long time, and Citibank still has centralized operations in places like Singapore or the US. Understanding the nuances or idiosyncrasies of each country was really hard. In my personal opinion, this is why Citibank was not successful on the consumer side. The corporate side is more stable, like a global company, with headquarters in the US, UK, and Asia, such as Hong Kong, where every branch operates the same way. Consumer banking differs from country to country, making it very hard to centralize platforms. I think one of Citibank's failures was trying to compete on a global scale when local nuances in operations are necessary.

This is a snippet of the transcript, sign up to read more.

In terms of the maturity of the credit process and underwriting, by the time you left, how confident were you and the team in your ability to underwrite and scale the credit book?

It was very robust and strong. There were some developments needed, like cash flow underwriting. Some providers gave us customer cash flow information, which is crucial for credit risk to understand customer capacity and ability to pay. The collection process was more robust; we didn't have collections before, but now we do, like Nubank has collections in Mexico. The combination of great origination, acquisition models, a low and grow strategy on credit line management, and a strong collection process is important for managing credit risk, especially in Latin American countries. Customers often need education on managing credit cards and revolving lines. Collection is crucial because customers frequently change their numbers. One day you're talking to someone with a number, and months later, it's a different person with the same number. You have to manage the pace of changing customer demographics and contact numbers, so a strong collection process is essential.

This is a snippet of the transcript, sign up to read more.