The Center for Data Innovation spoke with Pablo Viguera, co-CEO, and co-founder of Belvo, a company in Spain offering an open banking API in Latin America. Viguera discussed how the API works and how it delivers greater insights and innovation to the fintech ecosystem.
This interview has been edited.
Christophe Carugati: What is the purpose of Belvo’s platform, and how does it work?
Pablo Viguera: Our goal at Belvo is to empower the next generation of financial services in Latin America. And to do so, we have built a developer-friendly API platform that allows companies to quickly and safely connect their apps to financial data from banks, tax institutions, and gig economy platforms.
We currently offer two main solutions.
First, best-in-class data aggregation in Latin America. This allows companies to easily access financial data from their end-users to deliver more engaging experiences using banking and other alternative data sources, such as fiscal data and data from gig economy platforms such as Uber and Rappi.
Second, data enrichment. On top of our aggregation services, we have built a data enrichment solution that provides clients with a layer of intelligence so that they extract more value from their customers’ data, access specific data points to understand their customers’ financial habits better and improve decision-making.
Carugati: How does better access to data and analytics improve access to credit for consumers, and why is this important?
Viguera: Using data analytics, together with open banking solutions that enable secure access to users’ banking data, lenders can automate crucial parts of their underwriting processes, such as income verification. This allows them to improve and speed up their decision-making for risk scoring and, thanks to this, reach a broader customer base.
Income verification is one of the most important procedures that lenders conduct before approving a loan. Yet, the most common way to access this information is still the manual collection of payslips or account statements. This solution entails a high consumption of resources and time. And the data they provide is unverifiable and prone to errors.
To solve these challenges, Belvo has developed a data-science-based solution that allows lending companies in Latin America to identify and predict their income based on their banking data. The new solution uses Belvo’s API to access customer data, then aggregates and enriches this information using data science techniques to identify account movements that correspond to income.
This product is the first one designed to meet the particular needs of the credit lending sector in Latin America and a great example of how data analytics can improve credit access. Thanks to it, lenders can improve their knowledge of their users’ real financial stability and ability to pay to develop more tailored credit services faster and at a lower cost for them.
Carugati: How accurate are your predictive models, and how do you validate their accuracy?
Viguera: We currently offer two products that use predictive models: transaction categorization and income verification.
To keep track of the transaction categorization engine’s performance, we select a sample of transactions per week. The sample size ranges from several hundred to a few thousand. The transactions are then manually categorized by an analyst, new keywords are extracted to improve the model, and finally, the dataset is categorized automatically by the engine. We also generate datasets of a few hundred transactions that are manually and automatically classified, but no patterns are extracted, which gives us an unbiased estimate of the actual accuracy of the engine.
Our current categorization engine’s implementation can automatically categorize in real-time around 80-85 percent of transactions.
In the case of the income verification product, the output is a set of transactions that have been classified as income. Our model analyzes user account movements to find patterns in the frequency and quantity of transactions, as well as the combination of specific keywords that indicate if they correspond to income or not.
To estimate the accuracy, we take a sample of our customers’ requests on a weekly basis. Then, an analyst manually classifies the transactions as income or not. The transactions classified manually are then matched with the transactions classified by the product automatically.
Finally, we calculate our results’ accuracy with two different confidence levels, with a reliability of up to 90 percent.
Carugati: Belvo says it has access to over 90 percent of the bank accounts in Latin America. What were the main challenges in obtaining such widespread access, and what would it take to get to complete coverage?
Viguera: Our platform’s reach was the first thing we knew we needed to address at Belvo. And achieving this great coverage came with interesting challenges while building.
We wanted to build an inclusive solution that could enable the vast majority of the population in the region to link their accounts to financial apps. This meant that our priority was including the largest possible number of institutions into our platform so that, no matter where users’ financial data is stored, companies can access it safely through Belvo.
This was one of the main challenges our product team had to face to grow and integrate quickly. To solve it, we built a very strong local operations team focusing on understanding the requirements for those institutions’ accesses ahead of the development stage. This team is constantly improving this process for future integrations and building comprehensive guides on how to build the connections, ensuring consistency.
Another challenge was to handle all the potential scenarios that users can face when connecting their accounts to such a wide variety of institutions in Latin America. Our mission at Belvo is to make it as easy and secure as possible for users to connect their bank accounts to fintech applications in every possible scenario. Providing a smooth connection experience, given all the possible login requirements for each bank, was challenging.
This is something we completely baked into our Connect Widget, a plug-and-play connection tool for companies that provides an optimized login experience for users, which is fully customizable for each client and covers all the possible use cases they can require.
Lastly, we can mention the challenge of bringing simple, unified, and consistent data to our developers, given that we retrieve data from very heterogeneous sources. This is something critical for us as we want to make it as easy as possible for developers to innovate and extract value out of Belvo.
Carugati: How does Belvo’s API enable innovation in Latin America’s fintech ecosystem?
Viguera: The fintech ecosystem is exploding in Latin America, but there’s still a substantial lack of the necessary infrastructures for financial data to securely flow between users, traditional financial institutions, and new technology-based companies, such as neobanks or digital wallets.
We want to change this by becoming the standard infrastructure that connects all these players together so that innovative financial companies can access and interpret financial data from their users to create more modern, accessible, and inclusive products.
Leading innovators in Latin America are already building next-generation financial services thanks to Belvo by connecting their apps to their users’ bank accounts in a fast, efficient, stable, and developer-friendly way.
Our products help these companies use financial data to build new products and improve decision-making. Some of the most common use cases are: personal finance, where companies use our APIs to connect to access all their user’s bank accounts in one place so that they can provide better spending analytics and proactive recommendations; challenger banks and wallets that build an in-house personal finance manager for their users and instantly authenticate and validate the owner of any account to streamline their KYC processes; credit companies that are building innovative and secure experiences for borrowers while reducing their fraud risk and improving their scoring; and accounting and ERP companies that use our API to directly access their clients’ banking or fiscal data to reduce manual errors and costs via automated accounting.