We are excited to bring Transform 2022 back in-person July 19 and virtually July 20 – 28. Join AI and data leaders for insightful talks and exciting networking opportunities. Register today!
If there were any doubts left in the hearts and minds of retailers and lenders about the viability of buy now, pay later (BNPL) platforms, they were laid to rest this past holiday season. By the end of 2021, shoppers had spent over $20 billion using these point-of-sale lending offerings to make purchases immediately and pay for them at a future date through short-term financing.
Since then, BNPL has been dubbed one of the hottest consumer trends on the planet, projected to generate up to $680 billion in transaction volume worldwide by 2025 and spurring all manner of banks, fintechs, retailers, and ecommerce platforms to get in on the action. For many, however, the path to developing successful BNPL programs has been littered with obstacles that quickly expose the central challenge of the BNPL proposition: It’s not like any other form of lending that’s come before.
From executing real-time credit approvals based on scant customer data to scaling loan offerings to delivering a seamless customer experience, real-world BNPL implementation presents a complex set of operational challenges with which few lenders and merchants have had much experience. As a result, many fledgling efforts have struggled to get off the ground.
Fortunately, there have also been some successful early forays into the space that have established some best practices for implementing strong BNPL programs. Based on my team’s work developing large-scale BNPL initiatives, I’ve learned that the single most important lesson is to start small, taking a crawl, walk, run approach to BNPL program rollout, which lets the program learn as it grows.
Step 1: Widen your credit spectrum, narrow your loan offering
The biggest challenge in any BNPL scenario is quickly determining risk appetite based on minimal customer data. This is not the realm of traditional credit decisioning, with its detailed credit applications and credit bureau-based risk scoring standards. In a typical BNPL scenario, a largely unknown customer is browsing items online, adding them to a shopping cart and expects to complete the transaction in as few clicks as possible. The retailer must be able to offer a BNPL payment option, make a split-second credit decision, and execute the transaction in a matter of seconds.
That’s an inherently high-risk proposition that is focused more on building customer lifetime value than on immediate profitability. In the early stages of the program, a retailer will want to cast a wide net that will likely include approving customers in comparatively higher-risk tiers. This may sound counterintuitive, but taking more up-front risk initially is critical to maintaining the attractiveness of the BNPL offering, and the customer data collected in the process will help inform and guide the future of the program.
That risk is offset by diligently controlling the dollar amount for BNPL offers shown to each customer and keeping guardrails in place to limit the scope of the program based on total risk appetite.
Step 2: Incorporate alternative data sets
As the program gets up and running, it is critical to start ingesting and capturing merchant-specific data, such as customer purchase history, offer acceptance behavior, loyalty membership tier, etc., which can feed into the optimization of underwriting and identity verification processes. This information needs to be integrated directly into lender risk algorithms, along with other alternative data sources, such as bank statements, utility reporting, and income reporting to “train” the system based on real-world data.
Ultimately, BNPL programs need to get comfortable moving beyond the traditional credit score by recreating their own real-time screening and risk rating tools based on data generated from each new transaction. This allows the system to get smarter as it grows.
Step 3: Optimize to manage risk
Once the system has been operational for several months and retailers and lenders have been vigilant about collecting and analyzing consumer behavior, it will be possible to develop an optimization model that aligns personalized BNPL offers to customers based on their individual risk scores. This is where the real power of the program begins to reveal itself.
With this real-time, model-driven approach to underwriting, merchants and lenders offering BNPL platforms will not only be able to fine-tune special offers at the individual customer level; they will also have developed a proprietary risk framework for understanding customer behavior that is far more detailed and nuanced than anything that has come before.
Realigning our relationship with risk
Getting the BNPL formula right requires a fundamental overhaul to our conventional understanding of credit risk. Most traditional credit products involve one-time risk assessment for a single product, whereas BNPL programs need to manage multiple transactions at a customer level that occur at different points in time. Where traditional consumer lending models are focused on assessing up-front risk, BNPL programs require a calculated leap of faith on the front end in exchange for a treasure trove of highly personalized data on the back end. Done right, that flip to the conventional wisdom has the power to revolutionize consumer engagement. Done wrong, it creates risks that will make even the most ambitious lending players uncomfortable. The difference between the two is the ability to harness the data necessary to control the risk.
Vikas Sharma is Senior Vice President and Banking Analytics Practice Lead at EXL.
Welcome to the VentureBeat community!
DataDecisionMakers is where experts, including the technical people doing data work, can share data-related insights and innovation.
If you want to read about cutting-edge ideas and up-to-date information, best practices, and the future of data and data tech, join us at DataDecisionMakers.
You might even consider contributing an article of your own!
Read More From DataDecisionMakers
Source: Read Full Article