FitBank powers next-generation growth with MongoDB

THEIR CHALLENGE

Managing data complexity at FitBank

Brazil is at the vanguard of digital banking. The country’s vast size and young population means many banking customers are happy to skip traditional physical branches in favor of a digital experience. Today, 90% of Brazil’s population has at least one bank account and 77% of financial transactions are carried out digitally.

This embrace of digital, and Brazil’s buoyant economy—with a GDP growth of 200+% since 2000—has attracted a raft of fintech start-ups, including FitBank. Founded in 2014, FitBank is now an established player in Brazil’s financial sector. It provides digital wallets, payment processing, credit and debit cards, and a white label banking-as-a-service to corporate customers. In fact, FitBank remains the only 100% native cloud platform integrated in the Brazilian Payment System.

Recognizing the appetite for digital banking, FitBank identified this as an opportunity to demonstrate innovation while appealing to current customer demands. FitBank's business involves the daily processing of millions of banking transactions of various types through numerous APIs. Initially, all these transactions were stored in multiple SQL databases. As expected, this high volume of data created performance challenges, raising concerns that continued growth would lead to increased processing times, directly impacting the company's efficiency, availability, and reputation.

Additionally, FitBank faced the operational and financial burdens associated with running and maintaining relational databases. “We have a relational database that reaches several terabytes, and with our security, availability, and backup policies, this volume can even quadruple,” explained Gustavo Castelo Branco Crisóstomo Ramos, Head of Infrastructure and Cybersecurity at FitBank.

Ramos also highlighted the difficulty of retrieving information from a relational database. “For our engineers, extracting data was a time-consuming process, and performing complex queries on the relational database became challenging, as large volumes of data compromise application performance. Although query and application tuning is a constant focus of our SRE teams, the pace of database growth is rapid and challenging.”

FitBank recognized it urgently needed to restructure its data flow.

OUR SOLUTION

Enhancing performance with MongoDB Atlas

At the heart of FitBank's next-gen data architecture is MongoDB Atlas. The organization has adopted a new approach that retains operational data in the SQL server database while redirecting metadata to MongoDB Atlas. Utilizing APIs simplified this workflow, allowing Ramos’s team to rewrite only select code to manage metadata routing. “Now our APIs know that some data goes to the relational database, and other data—the metadata—goes to MongoDB,” said Ramos. “The migration was really very easy.”

For the data that stays in the SQL server, Ramos and team began the data tiering process with Parquet, a column-oriented data storage format. MongoDB leverages Parquet to make analyzing data in external systems easier. “We call it a data migration facility (DMF),” he says. “So DMF queries the relational database, gets the data, puts it in a Parquet file. Whenever we need the data, we can seamlessly query it via MongoDB.”

MongoDB Atlas on AWS offers a cost-effective, scalable and robust integrations with AWS Services so that its reduce the complexity of the architecture so that developers and architects can focus on driving the business. Thanks to Atlas Data Federation, FitBank is able to query the same way the data stored in SQL Server using MongoDB query language using Amazon S3.

“The simplicity of the solution was a pleasant surprise,” added Ramos. “We expected it to take months, maybe years—in fact we achieved it within weeks, and it took little effort.”

Now, FitBank relies on MongoDB Atlas for essential internal operations. The document-based model of MongoDB Atlas allows for flexible and dynamic schemas, making it easier to store and manage varied and complex data structures. This was a crucial requirement, given FitBank’s diverse metadata.

The MongoDB architecture efficiently handles large volumes of data and high-throughput operations. This makes it ideal for fast-paced banking applications. MongoDB’s capability to scale across multiple servers effectively resolves the performance issues FitBank faced with its legacy solution. Additionally, MongoDB's compatibility with over a dozen programming languages and 65+ third party platforms simplifies integration into FitBank's existing technology stack. This ease of integration enables FitBank to accelerate the implementation of new features and improvements by shortening development time and reducing cost.

OUTCOME

Smart approach to metadata underpins future growth

Having successfully completed redirection of its metadata, FitBank now has 70% of its internal systems connected to MongoDB Atlas. FitBank aims to have its ‘customer phase’—creating native apps in MongoDB—completed by the end of 2024.

Ramos said there are many ways to judge progress, and for FitBank one of the criteria is adoption rate. Currently, 20% of its development teams are using MongoDB Atlas, with more onboarding each month. “Sometimes, we introduce technology, but the adoption is not as we’d hoped,” said Ramos. “I'm proud when I introduce something that’s as widely accepted and used as MongoDB.” FitBank's adoption of MongoDB Atlas has transformed its system performance and operational efficiency. By adopting a flexible data model, the organization has achieved notable improvements, including:

• Reduced Database Load: 20% decrease in contention by migrating queries from SQL
• Faster Queries: Enhanced speed for improved user experience
• Lower Maintenance: Streamlined operations reduce maintenance efforts
• Scalability: Ensures optimal performance across database systems
• Cost reduction: Down by 20%.

By leveraging MongoDB's powerful data processing capabilities, FitBank can offer faster, more reliable services to its corporate customers, and an improved experience that aligns with its customer-centric business goals. For example, Ramos and his team are developing a system where customers can seamlessly access their latest balance and transactions from a mobile app.

It is a vastly improved scenario. “Now we have the best of both worlds,” says Ramos. “We lower our need for storage, for licensing, and we speed up our queries. We’ve likely reduced costs by around 20%, thanks to more efficient data management and flexible scaling options that eliminate the need for expensive storage redundancy.”

By far the biggest win, Ramos continues, has been breaking the architectural paradigm. “Before, the paradigm was such that we were only using relational databases,” he says. “Now, we can visualize new advancements and opportunities.”

Going forward, FitBank plans to deepen its integration with MongoDB, utilizing advanced features and capabilities to support evolving business needs. This includes exploring MongoDB's latest tools for analytics, machine learning, and real-time data processing to stay ahead in the competitive fintech landscape, and deliver superior customer experiences.

Discover why leading finance companies trust MongoDB Atlas to power their business by visiting MongoDB for Financial Services.