Data talks big in Brazil – for now

It’s impossible for me to include the impact of big data in my valuation model. If I ask the banks, they all say they are using it – that it’s a very important tool – but it’s impossible to know using publicly available data how they are using it and how big the impact is. I have absolutely no idea.”

This blunt assessment of the impact of big data from an equity analyst of Latin American banks is near universal. The banks, they say, are all talking about big data initiatives. They say it is going to drive performance, but any metrics about how the issue is affecting the business today, or even meaningful projections about how big data could improve results in the near future, are not forthcoming.

This is no surprise. The banks themselves don’t know. Unlike the previous wave of digitization, when banks implemented additional, lower-cost transaction channels (ATMs, online and now mobile), there are no clear metrics, even for internal dissemination.

The promise of big data is more powerful and more ethereal. But isolating its direct impact is almost impossible. Big data promises to touch all areas of banking – from retail to corporate and investment banking. Wealth management, trading and even the support areas of legal and HR are beginning to see its positive effects. 

When talking with Brazilian banks about big data, the first imperative is to ground the terminology. Concepts like machine learning and artificial intelligence mix easily in the conversation. After all, banks have always been data-driven organizations – Itaú says it has 60 million clients. Millions of transactions generated big data warehouses.

But capitalizing the term ‘Big Data’ indicates an evolution from these warehouses, which are fantastically expensive to maintain and limited in the scope of application. Data in this format needed to be structured. Often only individual departments could make use of this structured data for insights; broader analysis was virtually impossible.

Then came the new technology. Google created Hadoop and a new data revolution was born. Now data can be stored in ‘lakes’, not warehouses. It can be poured into these lakes in whatever format it comes – audio files, movies, simple texts, social media excerpts can all be thrown into a large pool of data along with transactional data and other more traditional sources. Data scientists can analyze the data in these Hadoop file systems using R code or Python. 

The static variables used in previous data analysis – those fields defined in the structured approach such as income, credit scores, etc – are now just one of many. Now new forms of data can be played with. Itaú recently bought part of ConectCar, a payment system for toll roads and car parks. It gives the bank the ability to know when clients enter shopping centres; algorithms can be created to automatically send credit offers to those people before they have left the car park. 

Three stages

Estevão Lazanha is director of data engineering at Itaú Unibanco. He says there are three stages of development of big data for banks. The first is to be able to use the bank’s internal data and combine it with external sources to improve the bank’s understanding of the client to enhance the effectiveness of its customer relations management, fraud and credit risk systems. This improves cross-selling insights and speeds up signals of deteriorating credit metrics.

“This is the kind of initiative that everyone is doing,” he says. “This won’t differentiate a bank, although maybe all banks haven’t got to the same depth at this level yet.”

New cloud computing means that this technology is widely and inexpensively available – data is no longer a barrier to entry for financial institutions (although the large banks still retain an advantage because of the huge amount of internal data generation).

“The second wave is to use the data to create a better experience for our clients – not just sell new things,” Lazanha says. He gives an example. Currently clients need to inform the bank when travelling abroad so the bank can unlock their debit and credit cards for international use. Some people forget and some were never aware of the need to do this in the first place. However, combining geodata from clients’ cell phones could enable the bank to automatically unlock these cards as clients travel. “We will become invisible most of the time,” says Lazanha. “Things will just work.”

Banks in Brazil are not doing this yet. The technology is not the issue, rather it is a cultural one. “We are looking for the best way to interact with the client,” he says. “The broader rule is that we always have to ask if he wants that experience. This is even more important for us when we think about what kind of data we are going to capture. When we talk about moving beyond transactional data, we understand that privacy is very important. All these solutions only make sense if the client sees a benefit – if they are feeling their privacy is being invaded it’s not going to work.” 

Itaú is working on this second stage. But Lazanha is already looking to the third stage of big data, which he argues will bring new business models.

“The third wave is, for me, the most powerful,” he says. “When you combine a lot of solutions for the client that are not just financial solutions. You make the client’s life easier.”

This third stage will lead to banks extending beyond their financial worlds and seamlessly merging into a client’s broader life. The bank will partner with other companies and have such rich data on an individual that marketing will become purely customized to their interests (rather than the propositions for static clusters as was the case in the past). To take a crude example, an individual who has transactional data from wine merchants, and who has a wine-based app and who has geodata from a visit to the Napa valley would be offered special wine packages. 

Sci-fi future

Bradesco is also working towards this sci-fi future. The potential for added revenues and lower costs is enormous. Luiz Kamogawa, model research manager at Bradesco, says the big data revolution will have an impact on the rationalization of resources and efficiency.

“At first, the complexity of implementing this new data structure may be higher than the traditional analytics structure,” says Kamogawa. “However, the benefits on efficiency and coverage gains outweigh these costs. Currently, our managers and front-line employees can only focus on a relatively small number of customers, from a large pool of over 26 million people. Algorithms and big data will allow these employees to focus on those customers with a higher need for attention.”

How much will big data increase efficiency or reduce costs? “I can’t say the numbers, but the potential for efficiency is obvious,” replies Kamogawa. 

Bradesco seems to be somewhere between stages one and two by Lazanha’s definition. It is implementing data lakes and developing machine learning. It too is working with geodata. It is also looking at data it pulls from its digital interactions with clients: the navigation history on the website alters the promotions they see there, for example. 

Henrique Arutin de Albuquerque, data innovation manager for Bradesco, says the bank is focusing more on the customer experience and relationship applications of big data for a number of reasons. First, this is an easier place to test algorithms, with tests on containable client groups (plus controls). It is also safer than working with core systems, where failures or mistakes could have dramatic implications. Practically, it is also easier. When clients update app versions, they can agree to new terms and conditions that enable the experimentation. 

Arutin says new customer-experience and relationship models can be developed and tested in three to six months with a data lake in place, whereas in the past, the time line was about 15 months – an example of the velocity and agility of big data.

Some of these tests give indications of the financial benefit big data might bring. The banks are reluctant to discuss specifics – and there was a sense that the precise financial impact might not have been measured. However, Itaú gave Euromoney an example of a credit-card project where the bank trialled an origination exercise using new external data sources to both increase origination rates and improve the credit performance of new customers. 

“In this exercise, we tripled the volume of origination and reduced the risk by half – so we improved our typical result by six times,” says Lazanha. This specific example included machine learning. The algorithm used found new analytical relationships within the data to drive improved volume and credit performance. 

Big data also goes beyond retail. Bradesco and Itaú are both using it for corporate and investment banking. BTG Pactual is also developing its own big data programme with trading and CRM applications. 

Gustavo Roxo, BTG Pactual

“Big data gives us a new perspective on asset prices,” says Gustavo Roxo, CIO at BTG Pactual. “Rather than, for example, looking at the pricing data for one oil company like we used to do, we can now look across the industry. We can see how the whole supply chain evolves and we can generate different economic perspectives; and this deep analysis helps us to better see if we are reaching a moment when we should be changing our strategy.”

BTG is also using new sources of unstructured data. Emails, both internal and external to the company, are being used to inform perceptions about companies’ prices and financial health. “We have started to collect non-structured data to use with the structured data,” says Roxo. “It’s getting more complex. We aren’t just looking at one-dimensional data analysis now – now there are a bunch of them. The next stage is how we capture the value of all that data, and that will come through machine learning – that’s our bet. That’s how we will be able to leverage it.”

Internal challenges

The biggest challenges appear to be internal: leveraging the data, which is down to data scientists and statisticians, and providing the organizational structure to allow big data insights to flow throughout the organization.

There is also the internal organizational challenge that is related to a larger human resources issue. “You need people who understand what the data means,” says Roxo. “You need people who understand the technology of very large amounts of data. And then, third, you need people with the analytical skills.”

That challenge is common to all the banks. And all the banks agree that trying to find all these skills in one person is hard, if not impossible. But even breaking the skill sets down still creates recruitment challenges. Data scientists are in short supply and huge demand. And the banks are not just competing among themselves, they are also competing with fintechs and data companies like Google, Facebook and Uber – and they are competing internationally. All the banks have responded by partnering with universities and launching programmes to develop internal talent.

Itaú, Bradesco, BTG Pactual and Santander have all got their own fintech outreach programmes. The strategies vary. Itaú has its Cubo project, which physically houses start-up technology companies. Bradesco has a competition-style outreach to try to find the best companies. Both say the aim is to find the best possible technology partners out there, but, practically, it is also a great way to recruit. 

At least in theory: BTG’s Roxo is not so sure: “We are collaborating with fintechs, just like all the banks, but the key thing is that most of these start-ups want control of their own futures. The key talent wants to leave college and start their own companies and develop their own software. Convincing them that they are better off joining a bank is very challenging – we have to create something similar in terms of environment, challenge and motivation.” 

This also applies to the banks’ engagement with the fintechs. “These start-ups don’t necessarily want to engage with large companies where they could be taken over,” Roxo adds. “It’s a tricky situation, and although the way the banks will make money in the future is by engaging with those fintechs, I don’t think it is something that has been addressed well so far. This relationship will evolve, but it’s not that simple.”

Some analysts seem to think the banks will be able to stay ahead of the start-ups. “None of the new fintechs has gained any market share and the banks are so worried that they are investing so much in this area that – given their large market share – they are in a good position to protect themselves,” says one.

After UBS’s annual fintech conference in São Paulo in April, analyst Frederic De Mariz wrote that the Brazilian banks are well placed due to the “cordial” style of disruption: “Brazilian fintechs are less aggressive in their desire to disrupt the status quo than US/Euro peers in our view. Incumbents are most likely to acquire rising fintechs or absorb their technology. Fintechs want to collaborate with banks or request a banking licence to offer a wider range of products.”

Driving transformation

Credit rating agency Moody’s seems to be one of the leaders of external analysis of the Brazilian banks’ digital strategies. It thinks market leaders, not start-ups, will drive the transformation of the banking landscape. At the end of April, it published a report that outlined – as far as it could – the digital strategies of Itaú, Bradesco and Banco do Brasil. It too concludes that fintechs would likely be pulled into the big banks’ gravitational field rather than create sufficient mass to stay within their own financial orbits. 

“Fintechs do not have the capacity to fully replace large traditional bank franchises because they are small and tend to specialize in offering limited services,” says the report, which was led by Ceres Lisboa, senior vice-president at Moody’s in São Paulo. “Though these relatively new entrants may increase competition in certain segments of the banking industry, including payment systems and asset management, in general, start-ups will likely opt to work in partnership with leading banking institutions.”

Lisboa agrees with the general consensus about the transformational aspect that will come from adoption of technology in the banking system. “Brazilian banks have historically been early adopters of new technologies, but the digital solutions being implemented today will likely prove to be even more transformative, producing a lasting impact on customer relationships and competitive dynamics.”

Lisboa points to the wider evolution that will support this trend. Today, 59% of Brazilians have access to the internet, either through mobile devices or personal computers, up sharply from 41% as recently as 2010. A regulatory change in 2016 also allowed accounts to be opened remotely and the central bank is fostering a benign climate for digital banking as it seeks to encourage competition and lower the system’s credit spreads. 

Quantifying the impact of market and institutional change is extraordinarily hard. Projection has always been part of equity analysis of course, but the speed of change and the potential size of that change and the impact on companies and markets is bigger than ever before. And now try to understand how big data – that has yet to be implemented and is full of unknowns but even greater promise – will change bank performance. 

Off the record, some of the bankers expressed deep concern that companies whose day-to-day business is managing big data could enter the market either through a start-up or by acquiring an existing bank to obtain its licence. The banks believe they have a window of opportunity – while old-fashioned retail branch networks and brands are still central to retail banking – to scale up their big data capabilities to see off this competition. 

How long that window stays open is anyone’s guess.