Go/no-go criteria and success metrics for robotaxis on Uber's mixed fleet

Uber is getting ready to add self-driving cars to its network and dispatch them alongside cars driven by people. ### Clarifying Questions - Does Uber own and operate the self-driving fleet, or does it dispatch vehicles that a partner owns and runs? - Is this decision about a first launch in one city, or about rolling out across many markets at once? - Can a rider choose or decline a self-driving car, or does dispatch assign whichever vehicle fits best? - Is the scope rides only, or also food and package delivery? ### Part 1 — The launch decision How would you decide whether to go ahead with the launch? ```hint Not every gate can wait for real trips Separate the conditions that must hold before any member of the public rides from the ones you can only measure once real trips happen. ``` #### What This Part Should Cover - Why Uber would put self-driving cars on its network, and what it risks by launching or by waiting. - Go or no-go criteria across safety, regulation, rider experience, operations and economics. - A staged rollout with an explicit decision at each stage and conditions that stop it. ### Part 2 — Success metrics If the launch goes ahead, which metrics would tell you whether it is working? ```hint One marketplace, two kinds of supply Ask what a self-driving trip does to the riders and drivers who never touch one. ``` #### What This Part Should Cover - A primary metric for the whole network in the launch area, not only self-driving trips. - Supporting metrics for riders, fleet operations, the marketplace and economics. - Guardrails for safety, human drivers' earnings and rider trust. ### What a Strong Answer Covers - Treating the decision as a staged, reversible city launch rather than a single yes or no. - Treating safety as a hard stop rather than a metric to trade against growth. - A measurement design that separates the effect of self-driving cars from seasonality and city differences, and accounts for the shared marketplace. - Human drivers considered in both the decision and the metrics. ### Follow-up Questions - In the launch area, self-driving trips get excellent ratings but human drivers' hourly earnings fall. What do you do? - A minor collision involving a self-driving car on Uber goes viral in the second week. How does it change your decision and your metrics? - How would you decide which ride requests a self-driving car should be offered?