Banco Bradesco

In-database R and Python models accelerate time-to-value.

Banco Bradesco Financial Services

Achieving digital transformation requires dramatic change. In small areas, you remain equivalent but in so many other areas, the company is represented in entirely new ways.

Founded in 1943, Banco Bradesco is one of the largest financial services and insurance institutions in Brazil and for the last 20 years, it has enjoyed rising ranks in the Fortune 500 through steadily evolving and growing. Their mission is “to contribute to the fulfillment of people and to sustainable development, by offering solutions, products and financial and insurance services that are widely diversified and accessible.” Banco Bradesco has built their business on a traditional product-centric approach, selling relevant products and services to customers.

Undergoing radical digital transformation

Banco Bradesco first recognized a shift in customer expectations and behaviors, where data management and analytics became the cornerstone in creating a customer-centric and data-driven culture. Transforming Banco Bradesco required: 

Digital Innovation: through omnichannel, digitalization of products and services, 100% digital on-boarding, and cross-channel transactions.

Digital-Native Platforms: like Bradesco Next, a fully digital bank designed for younger audiences; Bitz, a digital payments platform; and Agora an investments brokerage platform.

Open Banking: where the bank becomes a platform.

The bank views data as a key raw material in understanding, relating, and offering products and services to clients. To promote a data-driven culture in the bank, Banco Bradesco has formed multidisciplinary squads on strategic fronts that are transforming Bradesco from a product-centric to customer-centric organization. All of their initiatives have the customer at the center of their attention.

Banco Bradesco by the numbers

69.5M 
clients
31.9M
account holders
21M
digital account holders
98%
transactions carried out on digital channels

The beginning of the digital transformation

In 2019, Bradesco approached the CRM and Analytics teams under Andre Duarte’s responsibility, tasking them to apply modern analytic approaches to transform customer relationships to be more personalized and relevant, solving customer needs first, rather than leading with products.

Chief Analytics Officer, Rafael Cavalcanti, explained that Duarte “decided to build a couple of squads and focused them on developing models in two perspectives.” The first was to improve upon what analytics Banco Bradesco already had. “There was a lot of analytics going on in the bank, but not in the most efficient way. Our approach was, ‘we need to make sure that we can pick out some very specific use cases and show the organization that we can work together to produce immediate value’,” recalls Cavalcanti.

Second, Duarte tasked the teams to think and do things differently, challenging the way Banco Bradesco approaches modeling.

“For example, our product-oriented approach had a model for customers who want loan type A, another model for loan type B, another model for overdraft, and another model for credit cards. We needed to challenge that because, at the end of the day, customers in their financial journey and at a moment of life—they need loans, they need money. If that money is going to be expressed in a one-time payment loan, a monthly payment loan, or in overdraft, that's more related to the suitability of a product,” says Cavalcanti.

Project CRM 2.0 leads to customer-centricity models

Head of CRM, Gian Cantarella, is tasked with customer relationship management for all of Branco Bradesco. According to Cantarella, this means, “applying techniques to provide a better chance for our customers to have better offers, and for the bank to be more accurate in those offers across all channels. This includes all the branches and the entire sales team.”

Under Cantarella’s leadership, Banco Bradesco has implemented relationship hubs based on data and analytics in Teradata and integrated with Salesforce as the transactional CRM system.

“The vast majority of our data is in Teradata. So, our data foundation is there. Teradata is where all the data is located and where we do most of the analytics. We score and segment all of our customers in Teradata, which informs decision-making and customer journeys with Salesforce as the distribution engine,” explains Cantarella.

Over 8,000 client clusters have been identified for relationships and sales through digital channels and account managers. According to Caio Quini, senior data scientist in CRM, detailed and highly specific client clusters allows the bank to “find specific profiles of clients and know what they like based on the rest of our population of clients, or maybe even more. We can then correlate those and try to offer the same product.”

By emphasizing customer behaviors, Banco Bradesco customer-centricity meets their clients where they are in their individual financial journey. “We have been able to lift sales 2-2.5x, in some cases 3x, when compared to not having this kind of segmentation behind our analytics,” says Cantarella.

Enterprise analytics ecosystem makes data their (second) greatest asset

For organizations as large as Banco Bradesco, it’s common for teams to work in silos. This natural evolution of data analytics leads organizations to treat their data as business unit-focused, able to only serve a single purpose or use. The data is constrained, only able to answer specific questions because it’s isolated and not combined with other data to drive more answers and deeper insights.

To transform and become a data-driven culture, Banco Bradesco understood the value of an integrated core.

The Teradata Vantage platform empowers users to have access to the right data at the right time, free to use tools like PowerBI, Tableau, and MicroStrategy, to access that reusable, managed, secure, correct, integrated, enterprise scalable data.

Banco Bradesco data engineers’ sole mission is guaranteeing the data will be available to its thousands of users. “We have a specific team focused on guaranteeing the analytics teams don’t use the raw data. Rather, they have specific curated data labs with data that has already been prepared for the analytics teams to use. The data engineers’ efforts speed up the analytics development process,” expanded Fernanda Souza, data science specialist.

Through these Teradata Data Labs, analytics teams have access to the same physical instance without compromising the speed, freedom, autonomy, and power to do their work through best-in-class software that protects other mission critical workloads happening at the same time.

When more people and processes in your business have frictionless access to all the data, the more value is created. And with so much data coming into the modern enterprise, users need a flexible, agile, untethered way to access the right data, analytics, models, and advanced functions based on their requirements. Creating an effective and widely adopted data culture, requires a platform that can accommodate functionally unlimited users and concurrency while managing tactical and strategic queries with sophisticated workload management at scale.

Teradata Vantage is the data analytics platform enabling Banco Bradesco’s data culture at enterprise scale. 

Vantage’s differentiated workload management makes the most efficient use of fixed resources, whether in the cloud or on-prem. While other data analytics platforms require scaling through adding additional compute or hardware—a costly and inefficient way to scale resources to support growing workloads.

Cantarella stopped short of calling data their greatest asset. Instead, offering, “I would say that data is our second greatest asset. Our first greatest asset is our customer. But to satisfy our first asset, the customer, we have to have the second asset, which is data. Putting the customer at the center of everything we do—using data to improve our customers’ satisfaction—and as a consequence, generating more revenue to the company, that’s how we are transforming to be customer-centric through data.”

Data science in Teradata accelerates time-to-value and puts the customer at the center of decision-making

It’s common for data scientists to perform their data science outside of their data analytics platform, choosing to extract data from the data warehouse to run R and Python models in analytics applications separate from the data. This approach leads to data sampling. While sampling data is acceptable in many instances, it comes with risk of excluding important outliers or other unique data features that could be important to the models. More importantly, creating models in an environment where there is no easy access to the entire database makes it difficult for quick model operationalization to occur. Model creation is important, however, operationalizing in near real-time is when models become most useful.

The Teradata Vantage Advanced SQL Engine includes data science analytic functions, allowing users to develop and run Python and R models in-database; a feature called Script Table Operator (STO) supports a library of roughly two hundred R and Python models to run directly in-database in Teradata Vantage. When such modeling is done in Teradata, the Banco Bradesco data scientists have the advantage of the entire data set—in addition to the benefits of Teradata’s workload management, query optimization, and multidimensional scalability. All of which support more users, performing more complex queries, against more data, at speed and scale that other data analytics platforms can’t match. Additionally, this approach of in-database analytics rapidly accelerates the operationalization of models because the models no longer have to be manually brought into Teradata. Faster operationalization of models leads to more rapid answers and insights—ultimately, realizing potential business value sooner.

“If the data is in Teradata, it has become a clear choice of the team to perform analytics.”

“We used to use open source for R and Python but it wasn’t in-database. Now, with Teradata Vantage, we’re able to do more sophisticated analytics inside of Teradata. Effectively bringing the analytics to the data. It’s enabling us to create new models, challenge our existing models, and use different techniques we weren’t able to do in Teradata before,” said Cantarella.

Cavalcanti added, “if the data is in Teradata, it has become a clear choice of the team to perform analytics.”

Cavalcanti’s experience speaks to the value of in-database analytics, carrying the notion that data processing is conducted within the database by building analytic logic into the database itself.

Doing so eliminates the time and effort required to transform data and move the data back and forth between a database and a separate analytics application. When you separate where data resides and the analytic applications, time delays are introduced. This impacts the ability to gain critical business insights. Moreover, it creates needless overhead on data scientists and business analysts to create the bridge that would be used to operationalize analytics. This can make all the difference between being a market leader and a laggard.

On the subject of accelerating analytics operationalization, Marques shared, “now that we have incorporated Python, R, and nPath into our Teradata data analytics platform, we can develop a lot more models in shorter periods of time. Our data scientists have developed more than 20 models over the last year. Just by putting the right people in the right model operation, we could scale. Before, it would have taken 10 years to do 10 models because we were limited by techniques and processes. For example, implementing a linear regression model required converting variables in SQL, equating to a thousand lines of code. Now, with Python in-database in Teradata, you just promote the code to production, and it's implemented.”

No longer requiring data sampling, Bradesco is now implementing their data science models on all of the data within Teradata. This opens up the possibility to address a much larger opportunity of use-cases than ever before.

For example, identifying potential customers for a mortgage. Rather than running a simple propensity model, the team first applies nPath analytics with features such as demographic variables. The nPath analytics output is added as an additional variable to a mortgage propensity model, like a linear regression model, random forest, or XGBoost model.

“The beauty of this case is that we are talking about two different things but, together, they come out with a better result. One part is that we’re now able to use more sophisticated tools in Teradata to do analytics. And two, using Python and R with different techniques by including a new variable, coming from the nPath, inside our new Python models. The combination of these two things is where we really see the positive impact that we can have,” spoke Cantarella.

Data scientist, Quini, is credited with the innovative approach. His efforts allow the CRM team to see a clear path of a single client’s, and multiple clients’, behavior. The common behaviors are variables added in the Python model as a way to improve the model.

As a leader of the analytics team, Marques sees the acceleration of productivity as a key reason to continue to perform their advanced analytics in Teradata.

“If I were to ask any of my data scientists, ‘what is your experience using Teradata versus another available analytics application?’ They're going to say, ‘wow, we cannot compare that, Teradata is much faster.’  For example, our credit card transactional data is in Teradata and we can access that data rapidly to perform queries from one year. If you were to try to do the same thing elsewhere, it’d take a month,” declared Marques.

Quini sees multiple benefits from minimizing movement while increasing agility, to model curation, and document libraries. “having everything in the same platform is the best part. We can work in the model, we can go back to Teradata, we can have SQL, we can have Python, and then we can also add everything that we want to the documentation. That's not just the best thing for us when we are developing or creating a model, but that's also good for other people to see our work and run an algorithm curation.” The result, he said, is that “everything works better and faster that way.”

However innovative and creative the approach may be to multi-genre analytics in their Teradata data analytics platform, the real benefit is in adding value to the customer and the revenue generation Banco Bradesco sees. Improving model effectiveness becomes more than a ‘badge of honor’ for an individual’s data science model success.

“We can see that this is creating new opportunities and better results. Within our mortgage propensity models, we were able to go from a 30% KS to a 50% KS score. Just by using these new modeling practices,” claimed Cantarella.

Banco Bradesco’s improvement in KS scores increases the ability to correctly understand every customer’s unique need or wants of certain products, informing future products, and making the life of retail bank merchants and managers easier. “at the end of the day, we are trying to offer a product for someone, and you want this offer to be accurate. The analytics behind all of this becomes very important. With every percent increase in our KS score means we’re a little bit more accurate in the potential offer, generating a little more in sales, and a little more in customer satisfaction,” concluded Cantarella.

A look at recent Banco Bradesco earnings releases and investor presentations underlines the importance of data management strategies, and the impact of this multidisciplinary squad of analytics and CRM teams.

Their financial guidance for the upcoming year included a forecast for the bank’s expanded loan portfolio to grow between 9% and 13%. In addition, the bank’s yearly growth in mobile (23%) and digital users (15%) is in part because of the modern real-time decisioning platform. Contextualized offers for personalized omnichannel journeys have become a vital requisite.

“Every single challenge and opportunity I see for the bank, I believe my analytics team can help. We can help if we want to identify anti-money laundering actions, improve the quality of our sales (not only in the amount we sell but the quality that we do), and we can help optimize our branch networks. I see our analytics team in every single page of our CEO’s investor presentation—whether he's talking about wholesale, efficiency, or digitization challenges,” proclaims Cavalcanti.

In the end, transforming Banco Bradesco into a customer-centric organization, where data drives decisions, is about the people. The people are in multidisciplinary squads, engineering the data and features, building the models, and taking action on insights to drive noticeable gains., Souza summarized best, “I believe we are part of a revolution inside Bradesco. We're part of a squad that is pushing innovation to the entire company and we're making a real difference inside of the organization.”

Other customer stories