Katkat
Back

How to target commute coupon users?

LyftData Scientist · Technical Screen
HardData ScientistTechnical ScreenAnalytics & Experimentation

Last updated

This question tests your ability to design an experiment to measure the incremental lift of a coupon program and your strategy for segmenting users based on commuting behaviors to maximize ROI.

Scenario

Lyft is considering launching a "Commuter Pass" coupon program, offering a 15% discount on rides taken during weekday morning (7-9 AM) and evening (5-7 PM) commute hours. You are tasked with determining how to target users to maximize incremental revenue rather than just subsidizing rides that would have been taken anyway.

The Task

  1. Segmentation: How would you identify the target audience for this coupon? Propose at least three behavioral features that distinguish a "commuter" from a casual rider.
  2. Experiment Design: Design an A/B test to measure the incremental impact of these coupons. How would you handle potential cannibalization (e.g., users shifting non-commute rides to commute times) and spillover effects?
  3. Success Metrics: Beyond simple conversion rate, what metrics would you track to determine the health and success of this campaign?

Constraints

  • Assume you have access to historical ride data, including timestamps, origin/destination coordinates, ride type, and user spend history.
  • The company is highly concerned about "deadweight loss" (coupons given to users who would have booked a Lyft regardless).
  • Your solution should address how to validate that the observed growth is causal and not due to seasonal trends or natural commuting patterns.

Hints

Hint 1

Consider using features like ride frequency, consistency of route start/end points, and time-of-day regularity to build a commuter propensity score.

Hint 2

Think about using a randomized rollout at the user level, but consider if "cluster-based" randomization might be more robust against interference in urban environments.

Hint 3

Focus on the 'incremental ride count' and 'contribution margin per user' as primary KPIs to distinguish between profitable growth and coupon-subsidized loss.

Follow-up questions

  • How would you account for a user who shifts their commute from a private ride to a carpool/shared ride due to the coupon?
  • If the experiment shows high lift in a pilot city, how would you decide which other cities are the best candidates for scaling the program?
  • How would you quantify the risk of a 'burnout' effect where users become dependent on the discount and stop booking rides once the promotion expires?