Online Lending: the Good, the Bad and the Ugly
Plus: Mortgage Value Chain, Cash App, Turkish Banks
|Marc Rubinstein||Aug 7|| 17||11|
Issue #12 of Net Interest, my newsletter on financial sector themes. Welcome to 571 new subscribers who joined since last week, particularly those from the equity research community who gave me great feedback on last week’s note. Every Friday I go deep into a topic of interest in the sector and highlight a few other trending themes underneath. If you like what you are reading, please share and invite friends and colleagues to sign up. I appreciate it!
Online Lending: the Good, the Bad and the Ugly
They were founded in an era of low rates and benign credit conditions. So for as long as they’ve been around, people have been waiting to see if online lenders can withstand the impact of an economic downturn. Now we have one, and the results are mixed.
The Good. Newspapers report that Affirm, a US-based point-of-sale lender, is looking to IPO at a valuation of US$10 billion.
The Bad. Monzo, a UK neobank, filed its annual report. It noted, “material uncertainties that cast significant doubt upon the Group’s ability to continue as a going concern.”
The question is: what determines the success of some and the failure of others?
Because loans are inherently a digital product, lending is quite suited to an online world. However, outside of China, the penetration of specialist online lenders is quite small. The Cambridge Centre for Alternative Finance collates the best data. Excluding China, online lenders did volumes of only US$79 billion in 2018. Within the US they have a share of around 4% of total lending.
In the early days of online lending, user experience was enough of a draw for customers. Compliance can make the loan application process cumbersome, so any company that is able to streamline it will have an advantage. Typically though, online lenders exploit one of three additional features to drive a competitive advantage against incumbents:
The first is price. Like in any industry there are pockets of inefficiency in lending which startups are able to target. SoFi was founded on the basis of one such inefficiency in student loan pricing in the US. Prodigy Finance is another example—it offers loans to students who are studying abroad and find it difficult to get credit in spite of their superior earnings prospects. Elsewhere, interesting price arbitrage models have emerged to shift credit risk from employees to employers in order to reduce the credit risk premium.
Second, online lenders navigate gaps in the regulatory landscape. Ever since the financial crisis traditional banks have had to operate under tighter regulation. Specialist lenders are often not subject to the same conditions, either because they operate a platform rather than a balance sheet, or simply because they are small enough to fly under the radar. One paper estimates that the increasing regulatory burden faced by traditional banks can account for about 55% of recent shadow bank growth in the US.
Finally, online lenders are able to innovate new lending structures with entirely different characteristics to traditional loans.
Much has been made of the opportunity to price loans using alternative data sources, like Ant Financial does with social data in China. But in markets where credit data is mature, the opportunity diminishes. One study comparing new-fangled credit scoring models with more traditional ones found that although the new model was better, its advantage diminished for borrowers with longer credit history. Plus it’s not like online lenders have access to any more information than banks—HSBC has 169 petabytes of the stuff.
Rather, the advantage comes in using data to fashion new products. There are lots of examples. They include user acquisition loans where lenders advance funds for companies to acquire new users, utilising real time data on user economics. Or payroll loans which allow employees to draw down their salary on a daily basis, rather than the biweekly or monthly basis that employers are set up for. A lot of my time as a fintech angel investor is spent looking at opportunities like these.
However, even with these features, new lending companies have a tough time.
One of the reasons is that lending businesses are actually two businesses. There’s lending on the one side and there’s funding on the other. Many other businesses are two businesses as well. Payments is an adjacent business to lending, and that consists of a merchant side and a consumer side. However, the two sides there are a bit more complimentary—more merchants drives more consumers and more consumers drives more merchants. That’s not necessarily the case in lending. In lending the dynamic can work in reverse – bad lending can cause funding to dry up – but it isn’t a flywheel for growth because the model isn’t just about demand.
Another reason is that unit economics can be difficult to pin down. Again, contrast lending with payments. In payments all three dimensions of unit economics – customer acquisition cost, frequency, retention – are structurally more favourable. Customer acquisition costs are helped by the proximity of payments to commerce; the flywheel then helps to push them down. They are also more stable. In lending, you don’t know your true customer acquisition cost until the customer has fully paid back their loan, which may be months down the line.
It is no surprise that most fintech value has been created in payments. Two weeks ago we talked about Ant Financial, valued at US$200 billion, and which started out as a payments company. The largest fintech unicorn outside Asia is Stripe, valued at US$36 billion. In public markets, PayPal and Square have a combined market cap of US$300 billion. It’s true even at the smaller end of the scale. According to a recent survey of UK fintech startups, the median IRR in payments is 98%, which compares with 64% in lending.
Compared with payments, lending is hard!
So let’s look again at our three case studies.
The Good, the Bad and the Ugly
The marketplace model of lending was born in the UK in 2005. In its simplest form it provides a low-cost standardised loan application process and facilitates matching between borrowers and investors (lenders). Once a loan is originated, the marketplace company acts as an agent for investors by servicing the loan in return for ongoing fees. It maintains records, collects borrower repayments, distributes cash flows, and manages the recovery of unmet obligations. To improve the information investors have when selecting individual loans, most marketplace platforms provide additional services such as borrower screening and loan pricing; they also offer partial protection against loan default, and secondary trading facilities.
Ratesetter was founded as a marketplace for personal loans in the UK in 2009. Since then it has originated nearly £4 billion of loans. The problem most marketplace lenders face is that they are not seen to have ‘skin in the game’. Their incentives are geared towards volume growth rather than credit quality. Ratesetter innovated a mechanism to reconcile these competing interests via a ‘provision fund’ which absorbs missed repayments. However, the current economic situation is depleting the fund. The fund sits on cash reserves of £26 million, yet the company anticipates £38 million of losses. Credit losses are expected to hit 4.0% versus the 3.5% pencilled in when the loans were originated.
Faced with the prospect of higher losses, investors are skittish. Back when demand exceeded supply, marketplace companies like Ratesetter attracted institutional capital to fund loans. That’s reflected in the concentration of the platform’s investor base—there are 86,920 investors financing 833,343 loans. Lending Club, which operates a similar model in the US, reported origination volumes down 90% in the second quarter compared with last year as demand on the investor side evaporated. The company has said that although some demand has returned, five out of its top ten investors are still out of the market.
Covid flipped Ratesetter’s problems. While investors are the problem now because of concerns around credit quality, beforehand the business had plenty of investor demand, but struggled to find borrower demand. It really is like managing two businesses!
As a result Ratesetter never really grew that quickly. In each of 2016 and 2017 it originated around £600 million of loans. That grew to just over £700 million in 2018 and just below £800 million in 2019. Not exactly tech-like growth. And the cost of finding these loans made the company loss-making. Ratesetter hasn’t disclosed its marketing expenses, but other marketplace lenders spend around 40% of revenues acquiring borrowers. (The beneficiaries of that in the US are demand aggregators like Credit Karma, which has 100 million customers and which announced its sale to Intuit for US$7 billion in February this year. Although the deal is yet to complete – it’s being investigated on antitrust grounds – it’s clear where most of the value capture happens.)
Last week Ratesetter announced it is selling itself to Metro Bank at a valuation of between £2.5 and £12.5 million, depending on performance over the next three years. That compares to its last valuation of £261 million in 2018. Metro Bank revealed the logic of the acquisition: “RateSetter was struggling to find funding. The reality is that we have plenty of funding.” Metro Bank will look after one side of the business and perhaps if it can use its branch network to drive loans through the channel, it can solve for the other side, too.
While Ratesetter was negotiating with Metro Bank, OnDeck announced that it’s selling itself to Enova International (International because it launched a business in Brazil but most of its business is done across 40 states of America). It, too, went at a distressed price, selling for less than 10% of what its market cap was in 2015.
OnDeck’s funding profile is more diversified than Ratesetter’s—it deploys its own balance sheet as well as its marketplace. So the issue of ‘skin in the game’ is less relevant. However, it suffered the same problems of higher credit costs and higher acquisition costs. Its 15+ day delinquency rate increased from 10.3% in March to 39.5% in June. That’s on top of higher customer acquisition costs. They were already up last year across all three of the company’s distribution channels. By May 2020 the company’s ratio of customer lifetime value to customer acquisition cost was only 1.4x to 1.6x across the channels, which compares with 3-5x as an industry benchmark.
OnDeck’s journey also sheds light on the relationship between startups and incumbents. In 2015 OnDeck had signed a distribution agreement with JPMorgan in a move to reduce its customer acquisition costs. The partnership was terminated last year. JPMorgan explained: “It’s been a great collaboration with OnDeck… They helped us create and launch an online loan application process that gave business owners faster decisions and easier access to credit, something we will continue to do on our own platform.” While JPMorgan got the know-how, OnDeck was left with a US$900,000 impairment charge against technology it had used to support the bank. A lesson learned.
As we discussed in More Net Interest last week, the company rang a cautionary tone in its annual report: “the Directors recognise there are material uncertainties that cast significant doubt upon the Group’s ability to continue as a going concern.” It was never clear which of the three additional features highlighted above Monzo stood for over and above user experience. However, user experience was enough to win it over 4.4 million customers—until Covid hit. The company said, “we’ve seen organic customer growth slow as word-of-mouth drops.”
Like OnDeck, Monzo has suffered serious credit losses in the past few months. Even before Covid, at the end of February, a fifth of its loan book was either in default, in arrears or in high or very high risk categories (defined as >5% chance of default). The company took a £20 million provision against its loan book, exceeding the £18 million in interest the loan book yielded in the year. The company would have been better off not lending a penny.
Monzo’s ability to wow customers was clearly better than its ability to manage credit. But its customer acquisition costs were superficially low once credit costs are baked in. In past financial statements Monzo had tracked its per user contribution margin. This rose steadily from a loss of £65 per customer in September 2017 to a profit of £4 per customer in May 2019, including expected credit losses. Those unit economics have now been turned upside down.
Monzo is also getting hit by regulation. The annual report reveals that regulatory reviews are being undertaken around financial crime and may result in lower forecasted customer numbers and revenues, and increased costs associated with correcting areas of concern. In addition the Bank of England allegedly hiked Monzo’s capital requirement, prompting it to raise capital at a 40% discount to its last valuation. I argued last week that the invisible asymptote that stands in the way of fintechs is regulation. As they get larger, regulation matters more, and it’s a big fixed cost.
For all these online lenders the past few months has been challenging. Calibrating between lending and funding, dealing with greater regulatory scrutiny, adjusting to higher customer acquisition costs, managing a turn in the credit cycle.
And then there’s Affirm.
Founded in 2012 by Max Levchin, the company offers installment loans to consumers at the point of sale. A couple of weeks ago it announced an exclusive deal with Shopify to provide ‘buy now, pay later’ installment purchases to customers of Shopify merchants.
According to newspapers, it’s looking to IPO. On its last funding round the company was valued at US$2.9 billion so an IPO at US$10 billion would represent a significant markup. Until an S-1 is filed, financial information is not available. However Affirm’s main competitor Afterpay is public and its financial information is available. Afterpay is trading at an all time high, up eightfold from its low in March.
The market opportunity is an interesting one. Customers of merchants who have signed up are presented with an option at checkout to ‘buy now, pay later’. They fill in a short form to get an instant credit decision and if they are approved, rather than pay upfront, they can pay in four equal installments over a two month period. The loan is interest free as long as the customer pays on time; if they don’t then late fees are incurred; the merchant pays a base fee regardless.
The model has all three of the additional features that give fintech lenders a competitive advantage:
On price, it is cheaper than a credit card, but offers the delayed settlement benefit that a debit card doesn’t. The company has essentially unbundled some of the features of a credit card and by using these features alongside their debit card (as 85% of customers do in Afterpay’s home market of Australia) customers can recreate their benefits at lower cost. The company also prices customers individually by waiting until they are late before levying a fee. There is therefore no cross-subsidisation between customers as there is in the credit card model.
On regulation, the company is not regulated as a credit provider (since it doesn’t impose a charge for the ability to pay) nor as a payment system (since its relationships are structured bilaterally between consumers on the one hand and merchants on the other). However, it's not in the clear yet since its regulatory status is under scrutiny. Plainly, this is the least sustainable feature of the three, and provides a window for growth rather than the core of a business model.
As a reflection of how much faster regulatory response is these days than in the past, the company recently complained that “the dominant international card payment systems...were launched in Australia in 1984 and were not subject to RBA [Reserve Bank of Australia] regulation until 2004.”
Finally, the model is clearly innovative. Merchants report increased customer engagement which is why they are happy to pay the fee, which across the industry can be 4-6% of the transaction value compared with 1-3% for credit cards. For example, LVMH’s Fresh brand has seen average order value increase by 18% since adding Afterpay, with 82% of users new to the brand or ‘win-back’ shoppers.
Afterpay has leveraged all the benefits of payments economics into lending. Customer acquisition costs are low because they are subsidised by the merchant; frequency is high (customers in Australia and New Zealand who signed up with Afterpay in 2015-2017 are now purchasing, on average, ~22x per year); and retention is enhanced by availability—the two sides of the business are complimentary.
Its challenge should be credit. Here it is in conflict with its customers. Merchants will want approval rates to be high so that they can close their sale and consumers similarly will be influenced by their early application outcome. The danger is that as competitive pressures mount between providers, approval rates creep up. Right now around 30% of Afterpay’s order requests are not approved (a rate that can be as high as 50% for first-time customers). This has kept defaults low at 1.1% in the last financial year.
Since the onset of Covid, Afterpay has seen demand skyrocket. It has added more than 1.6 million US customers since March as online shopping has grown and as consumers become more careful about budgeting. Analysts and investors face a constant struggle discerning between the cyclical and the secular. Covid has been good to Afterpay in a way that it hasn’t to OnDeck or Ratesetter. But Covid is simply the catalyst rather than the cause.
Based on Afterpay’s performance, Affirm will likely be a sought after IPO. It’s success is predicated on low acquisition costs and a model that fosters better unit economics than traditional lending. Ultimately one of the problems faced by online lenders is the long lag between business decision (i.e. loan origination) and outcome. All companies contend with feedback lags but in lending they’re longer. Affirm sits at the shorter end of the scale. That’s perhaps all it takes.
More Net Interest
Mortgage Value Chain
Ever since pick and shovel makers made it big in the Gold Rush we’ve understood that value is not captured equitably along a value chain. Sometimes an outsized share of the value accrues to where the customer demand is. Sometimes it’s at the back end with the infrastructure suppliers as in the Gold Rush.
The US mortgage market has a very fragmented value chain. Brokers sit at the customer end. Then there are originators, insurers, appraisers, servicers, warehouse lenders, investors. Interspersed throughout these segments there are technology providers—at the front end providing white labelling origination services, at the back end providing processing services, in the servicing segment doing the servicing technology, and in the investor segment doing data and analytics.
This week Rocket Companies, the parent company of Quicken Loans, completed its IPO. It was a feature of More Net Interest a few weeks ago. The IPO didn’t go as well as expected. Its price and size were both cut and it went at a valuation of US$36 billion rather than US$44 billion. The problem is that even with a large market share of 8% it operates in a competitive and cyclical market.
Meanwhile, Ellie Mae, a provider of mortgage origination technology has just been sold to ICE for US$11 billion, eighteen months after it last traded at US$3.7 billion. ICE already has interests in the mortgage market and the deal is premised on “how we could hook the front end of the mortgage industry to the back end of the mortgage industry… the combination of the two of us together moving into the cloud with the entire industry attached to us would be very powerful.” Ellie Mae has a 50% share of a smaller market (US$10 billion total addressable market) but it’s stable and the company is making itself as critical to the industry as pick and shovel makers.
Two weeks ago we looked at Ant Financial. In the West, Square is probably closest to replicating Ant’s model. This week it filed its 10-Q and for the first time reported financial information separately for its two ecosystems, Seller and Cash App. Cash App was launched in 2013 and now has 30 million users, double the number it had at the end of 2018. Square makes money by charging businesses to use the app and by charging consumers fees to access additional services. It’s the consumer piece which is beginning to drive growth. The number of products customers adopted within their first month of activation increased during the second quarter and customers using two or more products generated 2-3x more revenue than customers using only P2P payments.
Companies play around with their segmental reporting structures all the time. New segments are created, others are combined; it’s a never-ending process that reflects company strategy and company politics. Square isn’t Ant Financial yet because it operates two distinct ecosystems. Cash App will have truly made it when it no longer exists as a separate segment.
A running Net Interest theme is how banks have been turned into a tool of policy across most markets. In emerging markets this is nothing new. But the cost is being laid bare in Turkey right now.
Against the backdrop of a Covid fuelled slowdown, Turkish authorities urged banks to lend more. They provided credit guarantees, monetary policy incentives and regulatory relaxation. Loan growth duly ramped, to ~30% year-on-year, and banks reported robust second quarter numbers. But higher loan growth led the current account deficit to ramp as well.
Now banks find themselves at the epicentre of another crisis. The Turkish lira is hitting record lows while the central bank depletes its FX reserves trying to defend it. Banks may find that their FX liquidity, on deposit with the central bank, is no longer there. And if jacking rates back up is the ultimate policy response, they will suffer higher funding costs and likely higher credit losses, too.