Credit-Card Cashback Launch: Segment Prioritization and Measurement Plan

Context

You are evaluating which customer segment to launch a new cashback feature to first. For each segment you have monthly averages and rates needed to compute unit economics, stress test resilience, and design an experiment to measure causal lifts.

Assume monthly units unless noted. Interchange is a percentage of spend; APR, cost of funds, net interest margin, and annualized loss rates are annualized.

Inputs (per segment)

• S: average monthly spend per active card ($/mo)
• B: average outstanding balance (OS) per active card ($)
• i: interchange rate (% of spend, decimal)
• a: APR (annual, decimal)
• f: cost of funds (annual, decimal)
• m: net interest margin = a − f (annual, decimal)
• c: operations expense per account per month ($)
• L: annualized loss rate (% of OS, annual, decimal)
• r: proposed cashback rate (% of spend, decimal)

Define R = B/S ("revolve ratio" in months of balance per month of spend).

Tasks

1. Compute unit economics per active account and per dollar of spend:
   • contribution = interchange + interest income − cashback − expected credit losses − opex.
   • Rank segments by contribution. Explain any dominance relationships you can establish from inputs.
2. Derive the break-even cashback rate for each segment. Show sensitivity of break-even to:
   • ±20% change in loss rate L.
   • ±20% change in monthly spend S (state your assumption about whether B is held fixed or scales with S).
   • Identify which segments remain viable across these stress scenarios.
3. Select the single segment to launch first under a hard constraint that contribution margin must be ≥ a threshold you choose (state and justify the threshold). Explain what metric shifts could change your decision.
4. After launch, cashback is expected to increase both monthly spend and OS balance. Propose a causal measurement plan to estimate these lifts separately from selection effects. Include:
   • Randomization unit and design (e.g., account-level vs. customer-level; opt-in/encouragement design if needed).
   • Sample-size inputs and approach.
   • Primary outcome and guardrail metrics.
   • Stratification and variance reduction (e.g., CUPED/baseline adjustment).
   • Handling cannibalization from existing promo cards.
   • Explicit stopping/rollout rule based on risk-adjusted CLV impact.