Capital Allocation Across a Lending Portfolio: Analysis, Deck and Leadership Q&A

A lender gives you 48 hours for a take-home on its lending data: a portfolio, capital-allocation style case. Decide how the lender should allocate its capital across the parts of its loan portfolio, and support the recommendation with your own analysis. The case is deliberately open-ended: you choose the framing, the metrics and the recommendation. The deliverable is a deck of five to seven slides plus a code appendix. In a later round you present it to risk and analytics leadership for 15 minutes, followed by 10 minutes of questions. The report does not describe the dataset beyond calling it lending data. Assume it is loan-level: one row per loan with the origination month, amount, term, interest rate and any fees earned, the borrower's risk tier or credit score at approval, a product or channel segment, and the loan's monthly performance to date (payments, delinquency status, charge-off and recoveries). If you would want other fields, say which ones and how they would change the analysis. ### Clarifying Questions - What does "capital" mean here: a fixed amount of lending capacity for the next period, funding whose cost rises with volume, or equity held against loans? - What is the objective: maximize risk-adjusted profit, grow volume subject to a loss limit, or something else? Are there limits on the portfolio's loss rate or on any one segment's share? - Does the data include declined applications, or only approved loans? - What funding cost and operating cost should be assumed, and do they differ by segment? - Over what horizon is the allocation set, and how quickly can it be changed? ### Part 1 — Frame the decision and choose the metrics State the decision your analysis supports, the segments you allocate across, the objective and constraints, and the metrics you will use to compare segments. Explain how you handle loans that have not yet reached the end of their term. ```hint Loans of different ages A loan booked last month has had almost no time to default. Ask how you would compare its losses with those of a loan booked two years ago. ``` ```hint Losses are only one line The segment with the highest loss rate is not automatically the worst place for capital. Think about what else differs between segments, including how long each loan ties up the money. ``` #### What This Part Should Cover - A one-sentence decision statement with the allocation unit, objective and constraints made explicit - A profitability metric per unit of capital that nets credit losses, funding cost and operating cost against revenue - A treatment of loss timing that makes young and mature loans comparable ### Part 2 — Analyze and recommend Compare the segments on your metrics and turn the comparison into a specific recommendation: where to add capital, where to cut, by roughly how much, and what the portfolio gains. Describe what goes in the code appendix. ```hint The next dollar The loans added when a segment grows may perform differently from that segment's average loan. Ask what the data can tell you about the loans at the edge of each segment. ``` ```hint Wrong by how much? Your loss projections will be wrong. Ask how far losses could run above them before each part of your recommendation stops making sense. ``` #### What This Part Should Cover - Segment comparison on projected lifetime, risk-adjusted economics, with uncertainty shown - A concrete allocation that respects the constraints, with its expected effect on the portfolio - Stress or sensitivity analysis and the data's limitations - A reproducible code appendix ### Part 3 — Present and defend the recommendation Lay out the five to seven slides for the 15-minute presentation, and prepare for the 10 minutes of questions from risk and analytics leadership: which questions do you expect, and how would you answer them? ```hint One slide Ask which slide the audience would need if they could see only one, and where it belongs in the deck. ``` #### What This Part Should Cover - A storyline that opens with the recommendation and its impact - One message per slide, each backed by a single chart or table - Prepared, short answers to the strongest challenges to the loss projections, the data and the downside ### What a Strong Answer Covers - A framing a risk leader would recognize as the real decision, with objective and constraints stated rather than implied - Vintage-based loss analysis with projections for immature loans, not raw cumulative loss rates - Unit economics that combine revenue, funding cost, credit losses and operating cost per unit of capital, accounting for loan term - A quantified recommendation that survives a stated downside scenario, with its limitations named - Concise, answer-first communication in both the deck and the Q&A ### Follow-up Questions - Your data contains only approved loans. What does that stop you from concluding about loosening approval in a segment, and how could the lender learn it safely? - A leader asks what happens to your recommendation if a recession pushes losses well above your projections. How do you answer in under a minute? - After the new allocation goes live, what would you monitor, and what early signal would make you reverse it? - Two segments have similar projected returns, but one has much less history. How do you weigh that uncertainty in the allocation?