Buy Now Pay Later (BNPL) has shaken up the payments industry as no other technology has in a while. Whilst many have been able to benefit from the new technology, the lack of background checks and flexibility surrounding BNPL makes it the perfect breeding ground for fraudsters, and with Black Friday coming up, BNPL risk managers are left with an important decision: how tightly do they monitor and accept shoppers wanting to use the service?
Joe Lemonnier is Product Marketing Manager at Resistant AI, where he works on product strategy. He’s been engrossed in online privacy and security products for over a decade, working on multiple VPN, anti-fingerprinting, and cybersecurity tools. He loves diving into data to inspire product teams to solve customer problems.
Speaking to The Fintech Times, Lemonnier explains how thanks to advances in machine learning, BNPL risk teams can choose a broader set of options and be more deliberate in their approach, and they do not need to worry as much about protection vs growth, and risk vs revenue:
By any reasonable accounts of her shopping habits, Cheryl Anderson led a pretty unremarkable life. For most of last year, she’d been filling out her wardrobe with a set of sport apparel outfits and occasional sneakers. Sometime during lockdown, she developed a taste for slightly fancier bracelets and earrings. She’d been using a popular Buy Now Pay Later service to make her purchases, and dutifully paying back the four monthly payments on each item.
By the time Black Friday rolled around, her limits on the platform had been increased to the point that she could afford more expensive items. So she splurged on a flatscreen TV and surround sound speaker set bundle. And that’s when Cheryl suddenly became very remarkable. Because she never paid for it. But she didn’t default on payments she couldn’t afford. You see, Cheryl never existed in the first place.
To say that point-of-sales payment markets are undergoing a transformation undersells the impact of BNPL: nothing else has brought the users-over-profits growth model of tech companies to the financial services industry in quite the same way.
Despite none of the major players having really turned a profit in this pioneering new form of fast-paced/low-friction lending, they’re making everyone from MasterCard to Apple and Goldman Sachs chase after them in a land grab for users — and the potential to create the next big marketplace. And the customers are loving it. According to a March 2021 survey run by The Ascent, 56% of Americans had tried BNPL, up from 38% the year before. 62% said it could replace their credit card, and 77% trusted BNPL companies more or as much as credit card providers. Meanwhile, GoCardless found in their survey that 42% of Americans fully intended to use the services this coming holiday season. In other words, BNPLs generate a ton of goodwill from their users, and they fully plan to leverage that into other lines of revenue to turn a profit.
But the mechanics of the space don’t just democratise access to credit, they also significantly increase risks. To avoid impacting conversion rates at the point of sales, credit and KYC checks often get neglected in sign up flows — and customers know this. That same Ascent survey showed that 31% of Americans used BNPL because their current credit cards were maxed out or because they wouldn’t otherwise be approved for loans. Another 24% were specifically seeking to borrow without credit checks — either because they wanted to minimise credit score check impacts, or because they likely wouldn’t be approved if the checks occurred.
BNPLs shoulder those credit risks on their balance sheets. UK equity research house Redburn showed the typical cost structure breakdown in the BNPL industry usually includes 30% of revenue going to cover credit losses — four times the estimated profit margin.
And this is where Cheryl comes back into our story. Because an environment with very few checks, very high customer and transaction volumes, and a high expectation of natural losses is the perfect hunting ground for fraudsters using stolen and synthetic identities.
A report put together for Reuters by IDCare, a not-for-profit consumer support organisation based in Australia, the birth-place of the industry, showed that BNPL-related identity theft doubled year over year, and had an outsized share compared with credit card fraud despite being a quarter of the size. Indeed, the ease of perpetrating these kinds of fraud is generating a seemingly endless string of outrage stories from victims, who end up with bills to deny and credit scores in shambles.
These frauds damage the reputations of the BNPL services just as they attempt to leverage customer loyalty into new revenue plays. But since they hide in the services’ natural credit losses, they inflate them, and invite additional scrutiny from regulators. And so every risk manager at a BNPL faces what seems like an unsolvable dilemma during the holiday shopping seasons that start on Black Friday: Tighten the screws too much, and lock potential new users out and leave money on the table. Loosen up too much and get overwhelmed by a string of new account fraud. But the problem is actually deeper, and they should also be asking themselves how many of their already approved users are actually sleepers — Cheryl’s — lying in wait to pull a coordinated, high-value bust out fraud and blow a hole in their plans and profits.
Choosing between protection and growth, risk and revenue, is the oldest story in finance. It might even be more extreme in the immediate, low-friction world of BNPL. But thanks to advances in machine learning, risk teams can choose a broader set of options and be more deliberate in their approach.
Reliable document forgery detection using machine learning can give trust verdicts on sign up within seconds and at scale, helping to weed out fraudsters before they enter the system. Once in the system, context-aware AI that takes in data not just from individuals, cohorts, and counterparties, but also real-world population densities, crime rates, and more can weave together a new, adaptive understanding of customer risk. They can spot and root out sleepers like Cheryl before they have a chance to deal their damage, and reallocate legitimate customers blocked by unsophisticated security rules into happy, engaged customers.
These approaches can smooth the path of BNPLs towards the new services they are seeking to deploy and achieve profitability sooner.