Katkat
Back

How Should Stripe Capital Be Evaluated?

StripeData Scientist · Onsite
MediumData ScientistOnsiteAnalytics & Experimentation

Last updated

This problem tests your ability to design a performance measurement framework for a lending product, focusing on balancing merchant growth with credit risk.

Scenario

Stripe Capital provides merchant cash advances, where eligible businesses receive a lump sum of capital in exchange for a percentage of their future credit card sales. As a Data Scientist, you are tasked with evaluating the health and profitability of this product portfolio.

The Task

Design an evaluation framework to assess the performance of Stripe Capital. Your response should cover the following areas:

  1. Key Performance Indicators (KPIs): Define at least three core metrics that track the balance between product adoption and credit risk. Justify why each metric matters.
  2. Cohort Analysis: Describe how you would set up a cohort analysis to track loan performance over time. What specific segments or cohorts would be most meaningful to compare?
  3. Risk-Adjusted Returns: How would you measure the "success" of a specific credit underwriting model? If the default rate is 0%, is that necessarily optimal for the business? Explain.
  4. Experimentation: Propose an A/B test to optimize the offer presentation (e.g., changing the fee structure or the repayment rate). What is your primary metric, and how would you handle potential sample ratio mismatches or selection bias?

Hints

Hint 1

Consider the difference between 'gross originations' and 'net losses' when defining your KPIs.

Hint 2

Think about the trade-off between strict underwriting (lower defaults, fewer customers) and loose underwriting (higher revenue, higher risk).

Hint 3

Account for the 'repayment velocity' of the merchants; faster repayment could indicate both healthy business growth and potential early payoff risk.

Follow-up questions

  • How would you account for macroeconomic shifts (e.g., a recession) when evaluating the performance of loans originated in different years?
  • If we introduce a new feature that allows merchants to 'refinance' their advance, how would this impact your existing metrics?
  • What data points beyond credit card transaction volume would you integrate into your model to improve predictive power for merchant default?