Evaluate UberEATS priority delivery and membership

You are a Data Scientist at UberEATS evaluating two monetization features. Assume the marketplace has customers, merchants, and couriers sharing limited delivery capacity within each city zone. The business goal is to maximize long-term incremental contribution margin, not just gross revenue. Use the following definition throughout: incremental contribution margin = subscription fees + delivery and service fees + incremental order margin - courier incentives - fee waivers - refunds - promotions - support costs. ### Case A: Priority Delivery Users can pay an extra fee to receive higher fulfillment priority and a faster delivery ETA. 1. Why should UberEATS consider launching this feature? What benefits could it create for customers, merchants, couriers, and platform economics? 2. How would you determine a fair and profitable price for the feature? What data would you need, and how would you balance willingness to pay, operational cost, and negative externalities on standard deliveries? 3. Design an experiment to evaluate the feature. Specify the unit of randomization, treatment and control, duration, primary success metric, guardrail metrics, and key sources of interference or bias. 4. If the experiment looks strong, would you launch it broadly? What disadvantages or rollout risks might still prevent full deployment? ### Case B: Membership Subscription Users can pay a monthly fee to receive 0 delivery fees and a reduced service fee from 20% to 10%. 1. How would you evaluate whether this membership product is strategically attractive? 2. Design an experiment to measure whether the feature creates incremental revenue and profit. What historical and experimental data would you use? How would you randomize users into treatment and control? 3. Suppose the treatment group receives a free trial to encourage adoption. What metrics should be measured during the trial and after the trial ends? How would you estimate causal impact when take-up and post-trial conversion are voluntary? 4. Based on the experiment, what evidence would you require before recommending launch, and what long-term risks would you still monitor?