Business Case Interview Questions

Sample Answer

You need to define churn operationally, as paid cancellations within a cohort and window, excluding trial expirations unless stated otherwise. Then you confirm the unit and denominator, for example churn rate = cancellations divided by starting paid members, and whether you treat monthly versus annual plans differently. You also lock the comparison baseline, for example month over month versus same month last year, and confirm if any policy, pricing, or billing changes occurred that would mechanically shift churn.

Sample Answer

You could frame this as total earnings per driver per week, or as earnings efficiency like $\text{earnings}/\text{online hour}$. Earnings per week captures livelihood impact, but earnings per online hour wins here because it separates supply changes from pay mechanics and demand, and it matches how drivers experience idle time. You set a primary metric like net $/online hour$ with definitions for net of fees, tips inclusion, and treatment of promotions. You add guardrails such as trip acceptance, cancellation, rider wait time, and surge frequency, and you state assumptions about seasonality, events, and product changes that affect demand or time online.

Sample Answer

You could do a top down approach from metro population and grocery spend, or a bottom up approach from households and shopping frequency. Top down wins here because you can anchor to total grocery dollars, then convert dollars to orders with an average basket size and a delivery eligible share. Then layer adoption as $\text{eligible households} \times \text{orders per household per month} \times \text{penetration}$. For growth drivers, isolate penetration, order frequency, and average basket size, because they map cleanly to product levers like selection, fees, and delivery speed.

Sample Answer

First I would define the core unit as rides per week, then break demand into residents plus visitors. For residents, I would estimate population, the share who can afford ride hailing, average trips per person per week, and the share of those trips that are substitutable for taxis, transit, or driving. For visitors, I would use weekly arrivals, average local trips per visitor, and substitution share. I would pressure test affordability and substitution share first, because they swing the output most and are easiest to triangulate from wages, car ownership, and transit quality.

Sample Answer

This question is checking whether you can separate total demand from shifted demand, avoid double counting, and choose drivers you can actually measure. You start from total annual eligible orders in the region, then estimate the share where same day changes customer choice, not just delivery speed. Model incremental volume as $$\Delta V = V_{eligible} \times (p_{new} - p_{old})$$ where $p$ is the probability an order is placed given the promise. The key drivers are eligibility coverage, incremental conversion uplift by category, and how the offer changes basket mix toward urgent items.

Sample Answer

Start with baseline orders: $1{,}000{,}000\times 0.08\times 0.50=40{,}000$. Baseline gross profit is $40{,}000\times 42\times 0.25=420{,}000$. New purchase rate is 10% higher relative, so orders become $40{,}000\times 1.10=44{,}000$, revenue per order becomes $36$, so gross profit becomes $44{,}000\times 36\times 0.25=396{,}000$. Then subtract incremental shipping: $44{,}000\times 1.50=66{,}000$, net new is $396{,}000-66{,}000-420{,}000=-90{,}000$, meaning it is likely value destructive on week one unless retention or frequency offsets.

Sample Answer

This question is checking whether you can translate retention into expected lifetime orders without mixing a cohort flow with an active user stock. You want expected active months for a new cohort: month 1 is 1, month 2 is $0.55$, then month 3 onward is geometric with factor $0.80$. Expected active months is $$1 + 0.55 + 0.55\cdot\frac{0.80}{1-0.80}=1+0.55+2.2=3.75.$$ Expected orders is $3.75\times 2.2=8.25$, so LTV is $8.25\times 8=66$. You get tripped up if you treat 80% retention as applying from month 1, or if you multiply orders per active month by total months in calendar time rather than cohort survival.

Sample Answer

The standard move is to convert impressions to ad impressions, then to revenue via eCPM and fill rate. But here, the impression drop couples with ad load, so you must apply the 4% reduction before computing ads. Baseline ad impressions are $50\text{M}\times 0.012=600{,}000$, test total impressions are $50\text{M}\times 0.96=48\text{M}$, test ad impressions are $48\text{M}\times 0.015=720{,}000$. Revenue is $$\text{Rev}=\frac{\text{AdImps}}{1000}\times \text{eCPM}\times \text{Fill}$$ so baseline is $600\times 6\times 0.95=3{,}420$, test is $720\times 6\times 0.95=4{,}104$, delta is $684$ per day. Pitfalls: double counting ads if ad load is defined per view not per impression, using eCPM that already bakes in fill, and ignoring auction dynamics where eCPM can change with more supply.

Sample Answer

This question is checking whether you can connect an operational cost line item to measurable business outcomes, not just say "faster is better." You quantify cost as fully loaded $\$/agent$ and incremental staffing, then quantify benefits through fewer refunds, higher driver retention, reduced cancellations, and improved marketplace reliability. You make a simple funnel, tickets affected per day, percent of tickets that cause churn or lost trips, and average contribution margin per trip to get $$\Delta \text{Profit} \approx \Delta \text{Trips} \times \text{CM per trip} - \Delta \text{Support cost}$$. You also call out diminishing returns, the 12 to 2 hour improvement may not move retention if most issues are non urgent, so you validate with a pilot and guardrails.

Sample Answer

The standard move is to compare incremental profit, coupon cost is immediate and scales with redemptions, model cost is partly fixed and accrues over time, so you compute LTV uplift versus total cost of ownership. But here, pay attention to second order effects: coupons can train customers to wait for discounts and can cannibalize full price orders, while models can change long term engagement and have compounding benefits. You pick the horizon based on customer repurchase cycle, if the median cycle is 30 days you can evaluate 90 to 180 days, if it is 1 to 2 weeks you can evaluate faster. You also ask about constraints, if churn is spiking now you may need the coupon as a bridge, while you build the model for durable gains.

Sample Answer

The standard move is to lead with the decision, the estimated impact, and the uncertainty, then propose a test and guardrails. But here, cross-functional misalignment matters because Finance optimizes margin, Product worries about user trust, and Ops worries about volume stability for dashers and merchants. You should translate the model into scenarios, for example base, downside churn, upside mix shift, and recommend a staged rollout with explicit stop conditions like $\Delta$ churn or $\Delta$ order rate thresholds. Close by assigning owners for measurement, instrumentation, and rollback so the group can commit to one plan.

Sample Answer

Get this wrong in production and you can spend millions to buy temporary supply that disappears, while training the marketplace to require incentives. The right call is to recommend a limited launch only if you can show the ETA improvement converts into durable demand or retention, not just nicer metrics, and if unit economics stay within a predefined bound like cost per incremental trip. You should flag risks like cannibalizing existing supply, shifting drivers across regions, and confounding from seasonality or local events. Your 2-week plan should define primary metrics, guardrails, and counterfactuals, plus daily monitoring and a clear rollback trigger.