All Posts

Many early-stage lenders hesitate to test new product ideas because of the perceived expense. That's a shame, because great testing doesn’t require massive fixed investments. It's often possible to get the feedback and data that you need without investing in infrastructure. If you're testing brand perception, for example, polished execution does matter, and you'll need to build some infrastructure to get that feedback. But when you're probing interest rates, risk response, o

Do you want to accurately and consistently segment risk, therefore enabling your entire credit strategy? Then yes, you need to monitor your model! We see robust monitoring save our clients real money all the time: A shadow scoring test flagged PSI anomalies arising from a difference in a vendor's data at month-end (a quirk that wasn't visible in the development data set). We were able to make adjustments to the model in production. A live model suddenly received drastically d

Same form factor. Very different infrastructure. They look identical in-hand, but credit and debit cards are very different on the back end. Prepaid cards are stored value products. (The customer is spending their money.) Credit cards are revolving loans. (The customer is spending your money.) From the infrastructure to the required day-to-day management, they're entirely different products. Some vendors blur the lines in their marketing, making it seem as if you can build a

We recently worked with a fintech trying to increase the profitability of their loan product. The obvious lever: price. Lower the interest rate, increase uptake. Raise the interest rate, increase the interest income, but volume takes a hit. Econ 101. So, this lender ran tests: increasing or decreasing the interest rate 100 bps for different groups. The results were underwhelming, with customers barely responding to the new interest rates. This puts the lender in a frustrating

"If it ain't broke, don't fix it." Lenders often push back when we suggest exploring a new model build. It's fair—model builds require resources, and it can feel silly to fiddle with an underwriting model that "works," especially if origination volumes are on track and losses seem manageable. But at Ensemblex, we know that "works" often means "leaves money on the table." What Does It Mean for a Model to "Work"? In technical terms, a credit model is effective if it "slopes ris

Launching a credit card (or other lending product) can make you feel like you're spread pretty thin. There's a myriad of vendor, partner, and infrastructure decisions to make, and all of them are consequential. You're juggling a lot: Searching for a sponsor bank. Writing the credit policy that will define how your first loans are made (and how your first returns will look!) Convincing debt providers to take a chance on you. Contracts. So many contracts. Assembling the tech st

What Exactly Is a Leaky Variable? A leaky variable is any feature in your training data that contains information you won't have at decision time. The most extreme example would be using default status to predict default. That creates a perfect model in development with zero real-world utility, as the model will simply learn to predict the outcome with itself. Of course, any serious data scientist would catch an error that massive. But leakage can be subtle: Post-application

Running tests has technically never been easier. With highly configurable back-end tech and sophisticated data analysis tools widely available, even early-stage lenders with small teams can run a sophisticated testing program. The ease is a double-edged sword though, as it's also become easy to drown your insights in noise with sloppy testing. To make sure testing brings meaningful results, follow these four principles. 1. You need a learning agenda. And a budget, too. There

We see so many lenders get stuck in the same paradoxical situation: they're trying to expand their reach to the millions of underbanked or thin-file customers, but their own policies stand in the way. For good reason. Governance structures designed for lending to prime, salaried borrowers shouldn't be recklessly expanded to serve borrowers with no credit history. That would be trying to serve a motorbike mechanic using the same credit policies built for a mid-career engine

The Short Answer: Yes, and Don't Cut Corners When investors, debt providers, and sponsor banks ask to see your credit policy, it may seem like just another box to check. Can’t we just explain the model and include some example flows? Can I whip one up with some help from ChatGPT and a few templates? But while it might seem like a boring document, the credit policy is a terrible place to skimp. Those dozens of pages prove that you're a serious lending business ready to lend re