This is a free version of Net Interest, my newsletter on financial sector themes. For additional content and supplementary features, please consider signing up as a paid subscriber.
A Nu Dawn, A Nu Day
One of the best – and easiest – trades I ever did took place around ten years ago. I’d been in Brazil, scouting for investment opportunities, and had visited management at the big banks headquartered in São Paulo. The city is a tough one to navigate – one of the most traffic-congested in the world. I moved slowly in an armoured vehicle between meetings, only later finding out that high rollers commute by helicopter.
But it was worth it. Management described to me a banking sector that was one of the most profitable in the world. A history of hyperinflation kept interest rates high. Simply buying and holding government bonds was a profitable enterprise. Most private lending was reserved for large corporations, and these relationships fuelled banks’ consumer business – employees typically maintained a bank account with whichever bank their employer used. Interest rates on consumer loans were justifiably high because charging overdraft fees was prohibited and a lopsided credit scoring regime made it difficult to underwrite credit.1
On top of all this, Brazil’s banking sector was highly concentrated. A couple of years before my trip, two of the largest banks – Itaú and Unibanco – had merged to form Itaú Unibanco. Over the next few years, concentration would only increase, with Citigroup selling its consumer business to Itaú Unibanco and HSBC selling to Bradesco, another of the large domestic banks. After that bout of consolidation, the five largest banks would account for over 80% of banking assets, up from 50% in 2000.
Shortly after my return to London, Bank of America announced that it was selling its stake in Itaú Unibanco in order to shore up its capital position. It had acquired the stake before the financial crisis when it merged its Brazilian operations with those of Itaú. As a forced seller, Bank of America offered a very attractive price. Optimistic about the prospects for the sector, we bought some; it was a nice simple trade.
It turns out I wasn’t the only one looking at Brazilian banks at the time. David Vélez was a partner at venture capital firm, Sequoia Capital. He was looking at them too. Except he came to a very different conclusion. A few years later, he founded Nubank.
In mid-2012, I entered the branch of one of Brazil’s largest banks to open my first Brazilian bank account. As I approached the first bulletproof door that was flanked by armed security guards, I sensed this was not going to be easy. During the following four months I spent long hours in queues, calling the call center, and returning to the bank branch with an increasing number of documents, until finally a bank account that would charge hundreds of reais per year in fees was approved in my name. The entire experience was incredibly frustrating. As I tried to reconcile this experience with the immense profitability of Brazilian banks and the low penetration of banking in the country, I realized that this was possibly the entrepreneurship challenge I had been looking for…
Nubank’s founding story shares many similarities with Tinkoff in Russia, the subject of a Net Interest post earlier this year. Both Vélez and Oleg Tinkov, founder of Tinkoff, found inspiration in Capital One, which had achieved success as a new entrant in the US financial services market in the 1990s. In each of their respective countries, credit cards were an underpenetrated market, with significant potential for growth. Neither founder had a background in either banking or tech. And they each confronted a wall of skepticism when they laid out their plans. Oleg Tinkov recounts in his memoir how bankers scoffed at his plan to start a credit card bank. Vélez, the same:
“I spent a lot of time talking to the experts, to the CEOs, to the consultants. The overall sense – their conclusion – was that it was impossible.”
This week, Nu proved it wasn’t impossible, filing its prospectus ahead of a listing on the New York Stock Exchange that may see it valued in excess of US$50 billion. Today, the bank offers credit card loans, personal loans, deposit accounts, payments services, insurance and investments to 48.1 million customers, mostly in Brazil.
Why not become a paid subscriber of Net Interest? As well as the free post, subscribers get supplementary features. This week, they include comment on Metro Bank, Stablecoins and Sell-Side Analysts, plus an Excel model of Nubank. To join, sign up here:
There are two ways to look at a bank. One – let’s call it the customer view – is through the eyes of a customer. Like any business, a bank would be nothing without its customers; they represent the indivisible unit of account. The customer view dominates in the tech school. It tracks the number of customers, the cost to acquire customers, contribution per customer and lifetime value of customers as the route to evaluating the business.
The other way – let’s call it the balance sheet view – is more top-down. It recognises that banks have a unique role in the economy to transform risk and create money. As they get larger, this role becomes more central and entropy leads to banks getting more complex. Bank of America has 40.9 million digital active users of its consumer banking services, not that far behind Nu. But it also has a US$3.1 trillion balance sheet, which is the thing to analyse if you want to understand what’s going on.
With that in mind, let’s take a look at Nu. We’ll start with the customer view.
The Customer View
Vélez is a big proponent of the customer view. As well as being influenced by early Capital One, he cites the influence of the big technology companies. This leads to a difference in the way he thinks about the business compared with incumbent banks.
“There are a number of different things that we’re doing differently. But I would say the number one is having a culture that is obsessed about customers and doing the right thing for the customers, from doing the right decisions, to giving the right customer service, to building products that are really actually good for them. I would say that's number one.”
Nu’s first product was a purple credit card, launched in September 2014. To apply for a card, prospective clients had to download an app and enter some identification details; they would receive a preliminary decision within two minutes. The company had raised US$15 million of series A funding from Sequoia and Kaszek Ventures, but was loath to spend it on marketing. Rather, Nu harnessed customer referrals. A referral programme was built into the app, and it turned into a very effective customer acquisition tool. Within four months, the company had 1,200 credit card clients; one year later it was at 350,000.
The referral programme has helped to keep customer acquisition costs very low. The company claims to have acquired 80-90% of its customers organically on average per year since its inception. In the first nine months of 2021, customer acquisition costs were US$5.0 per customer.
Unlike other local credit card issuers, Nu doesn’t levy an annual fee, so it absorbs that customer acquisition cost up front. Payback comes from its share of interchange fees that merchants pay on transaction volumes. In the first nine months, that fee was 1.04% net of associated costs.
Average spend so far this year is around US$115 per active customer per month, but it takes a while for customers to warm up. When they first use the card, customers typically make around 4 transactions per month. Over time, that goes up. Average usage per customer tends to peak out at around 18-20 transactions per month. Alongside that, payment volume goes up. After two years, customers put about twice as much spend on their Nu cards per month as they did in their first few months as customers, and after four years they are spending around three times as much.
All this means that customers become more valuable the longer they stick around. The one wrinkle is that interchange fees are slipping. Back in 2018, the net interchange rate was 1.11%. Part of the reason is a shift in mix towards debit transactions, where the fees are lower. Indeed, a regulatory change (Public Consultation 89/21) may drive fees lower still as regulators look to cap debit rates. Given its current mix, that would knock around 0.10% off Nu’s fee rate.
So Nu works hard to make sure customers do stick around. In the first nine months of this year, it lost an average 0.06% of its customers per month on a voluntary basis (and kicked out 0.07% per month due to risk or fraud concerns). It does this partly by enticing them with new products. Since its early days, Nu has continuously launched additional products:
In 2017, a bank account
In 2018, a debit card attached to the bank account
In 2019, a personal loan product and a business account
In 2020, insurance brokerage services and investments, following the acquisition of Easynvest, a retail investment platform
In the future, an e-commerce marketplace to offer products beyond the bank’s core capabilities in cooperation with partners
While the earliest cohorts of customers started out with one product, the latest cohorts are starting out with around three, and customers across all cohorts are ratcheting up the number of Nu products they have to between three and four. As of September 2021, 53% of customers have a deposit with Nu and 4% have a personal loan. Charts the company lays out in its filing show the positive impact that new products have on customer activity levels. Personal loans are looking particularly profitable. Among those customers that have a credit card, a personal account and a personal loan, “ARPAC [Average Revenue Per Active Customer] ... has increased substantially in the last several months as we started to offer these customers access to our personal loans product in June 2020.”
In addition, with more products, the company hopes that customers increasingly regard Nu as their primary bank. It estimates that around half of those active customers who have been with the bank for more than a year use Nu as their primary bank. That means they pass a greater volume of funds through their Nu accounts, which means higher profitability. The same may happen as customers get older, and potentially more affluent. Right now, the average age of a Nu customer is 34, but as growth slows, that may start to creep up.
Bringing it all together, the company estimates that a customer breaks even on their acquisition cost after around 12 months, before their contribution ramps up. After 24 months, their cumulative contribution is around US$13; and after 36 months, it’s around US$37.
Assuming the customer sticks around for ten years, and discounting their contributions back at a rate of 12%, Nu estimates that the value of a customer is equivalent to over 30x their acquisition cost.2 On the basis of the current customer acquisition cost of US$5.0, that makes each customer worth around US$150 to Nu. With 48.1 million customers, that’s US$7.2 billion of value.
A Nubank financial model is available to paid subscribers of Net Interest. To get access, sign up here.
Yet Nu is hoping to fetch a valuation considerably higher than US$7.2 billion in the market, so where’s the rest of its value coming from? There are three possible sources.
First, Nu is still still growing its customer base. It already has approximately 28% of the population of Brazil over the age of 15 as customers, but it is expanding in Mexico and Colombia, where it currently only has 762,000 and 37,500 customers respectively.
Second, the company may be underestimating the cumulative contribution of a customer as it rolls out new products. It puts its current serviceable addressable market – consisting of retail financial services in Brazil – at US$99 billion of potential revenue, but its total addressable market – retail financial services and marketplace services across all of Latin America – is 2.7x the size, at US$269 billion.
Finally, customers may stick around for more than ten years, particularly if they come to utilise Nu as their primary bank. (You know the old line: you’re more likely to get divorced than change the bank account you have on your wedding day.)
The Balance Sheet View
The customer view dominates Nu’s filing document. There’s lots of discussion of Average Revenue Per Active Customer even though that doesn’t include customer deposit costs. There’s only one mention of net interest margin – and its acronym “NIM” is put in inverted commas. (It was 6.2% in the first nine months of the year, in case you’re wondering; full calculation alongside other performance indicators available to paid subscribers).
It helps that Nu’s balance sheet is pretty simple. It has collected US$8.1 billion of customer deposits and it has made US$5.3 billion of loans. It’s total balance sheet size is only US$15 billion (a far cry from Bank of America’s!) and it is backed by US$2 billion of equity.
Nevertheless, there is one risk in there and that’s credit.
Accounting regulations require Nu to take provisions against credit risk up front, so while it is growing its loan book, it suffers the drag from putting up those provisions. As at the end of September, the bank had US$425 million of provisions squirreled away for bad loans. In the first nine months of 2021, it added US$281 million (net of reversals) to its provisions, yet the loan book only generated US$406 million of interest income, before financing costs. That wouldn’t be good business if it was like-for-like, particularly as financing costs would absorb most of the profit that remained, but the US$281 million reflects incremental credit losses the loans will likely suffer over their life, while the US$406 million is just what they threw off in interest in those nine months.3
The provisions Nu takes reflect its current outlook for credit. In its credit card portfolio, for example, it classifies 12.6% of loans as being at “higher risk” of default (over 20% probability) and in its personal lending portfolio, that proportion is 15.7%. Interestingly, the company doesn’t view its credit quality as being especially sensitive to macroeconomic conditions. Its provision is predicated on a base case estimate of 5.4% GDP growth in Brazil; a more pessimistic estimate of 4.7% doesn’t have that much impact on the provisioning requirement.
Of greater risk is the lending model itself. Like other data-driven lending businesses (e.g. Upstart, discussed in More Net Interest two weeks ago), Nu reckons its models provide superior credit underwriting capabilities. It describes an internally developed credit engine called NuX which collects more than 11,000 data points per monthly active customer. One such datapoint is how an applicant came to Nubank – the model shows that someone recommended by a well-performing cardholder is likely to be a better credit risk.
Its credit engine allows Nu to optimise between approval rate and risk. The company claims that for a given approval rate, its model reduces risk by 60%. Alongside a higher risk appetite, it has moved up its approval rate from 15% in September 2020 to 42% currently. This does mean going into lower income brackets, though. Around 56% of the company’s credit balances are to borrowers earning less than 3x the minimum wage; traditional banks allocate only 35% of their credit in that bracket.
Nu shows that its loss rates by income bracket are lower than industry averages and that its overall delinquency rate in credit cards is lower than the industry average. But with faster growth than the industry, it’s difficult to make a comparison without seeing credit cohort curves. There are a lot of cohort curves in the filing; sadly none of them reflect credit! That said, unlike data-driven lending businesses in the US, Nu did operate through a recession in Brazil, during 2015 and 2016, so its model has been stress-tested.
In the past I’ve talked about the trade-off at the core of banking regulation – between stability and competition. Brazilian policymakers historically erred on the side of stability. They allowed a profitable oligopoly to form – the one I was introduced to on my trip around São Paulo – and in many respects it paid off. The Brazilian financial system sailed through the financial crisis and numerous recessionary periods without collateral damage. But it had a cost, which was financial inclusion. Too many Brazilians were excluded from the financial system. David Vélez made the case for change:
“We made a very honest case saying, ‘Regulators, you have five banks that own 90% of these markets and they're charging the highest interest rates in the world. You have 40 million Brazilians outside this banking system. How is this type of concentration good for Brazil?’ And they agreed. They said, ‘It’s not good. We want competition.’”
Nu represents that competition, and it’s going to be exciting seeing it go public. The company is giving away free shares to its clients, so if you are one, well done! Be warned, though: there are a couple of errors in the company’s filing document… All is revealed in More Net Interest, available exclusively to paid subscribers.
Prior to 2011, Brazil operated only a ‘negative’ credit bureau. A ‘positive’ credit bureau was introduced in 2011, but it required consumers to opt-in. In July 2019, the law was changed in favour of opt-out. Since then, banks, retailers and utilities have contributed data to bureaus, allowing a continuous credit score to be generated rather than the binary decision that thwarted even David Vélez, founder of Nubank: “In my case, I moved apartments and the cable company still sent me a bill for $10 and I never got that bill. So I became a delinquent for them. They sent me to one of these credit bureaus. If I needed a loan from one of the big banks, I would have been rejected.”
Nu defines “Lifetime value” or “LTV” as the present value of estimated contribution margin generated by a customer during the lifetime of a customer’s relationship with its business. It calculates LTV based on the following key assumptions: (1) 12% per annum as the discount rate applied to the projected stream of contribution margin generated by a customer; (2) estimated lifetime capped at 10 years; and (3) growth and churn estimates based on historical analysis across our cohorts and estimated inflation rates.
There’s something else going on here, too. The credit card product works differently in Brazil compared with elsewhere. Just as in other countries, customers have a choice to pay off their balance in full every month or to allow credit to ‘revolve’ on which they pay interest. But in Brazil, there is a third mechanism, which is to pay by installments. This mechanism allows consumers to divide the cost of an item into a number of equal installments which they pay monthly through their credit card. So a R$2,000 sofa, for example, can be paid over ten installments of R$200 per month. Rather than incurring interest, the merchant reflects the financing charge in the original cost of the product. Each month, the consumer pays their credit card issuer and the credit card issuer passes the funds on to the merchant.
In Nu’s case, installments make up 47% of its credit card receivables, which is roughly in line with the industry. But because it doesn’t generate interest on most of these, nor on the 47% which are paid off in full every month, the company generates relatively little interest income on its credit card portfolio. It does carry the credit risk, though, as it is obliged to make the payments on behalf of the cardholder even if the cardholder does not pay.
Consequently, Nu may suffer losses on credit credit receivables from which it hasn’t earned any interest income.