What guardrail metrics should Uber track for Driver App at global scale

Problem Statement Description

Product context: Uber is a mobility and delivery platform; its products include rides, Uber Eats, grocery and retail delivery, freight, driver and courier tools, and marketplace pricing.

Uber’s Driver App is the operating interface for drivers across a global mobility, delivery, and logistics marketplace. In this metrics interview, you are being asked to define the guardrail metrics Uber should monitor to ensure changes to the Driver App do not harm marketplace health, driver experience, rider/customer reliability, safety, or business outcomes at global scale.

The question is not asking for a single north-star metric or a growth dashboard. It is asking you to reason about what can go wrong when driver-facing product changes are launched across diverse markets, vehicle types, trip types, regulatory environments, network conditions, and driver segments. The answer should show how guardrails help Uber make safe product decisions while preserving marketplace liquidity and operational reliability.

You should assume the Driver App supports workflows such as going online, receiving and accepting trips, navigating to pickup/dropoff, communicating with riders or customers, completing trips, viewing earnings, and resolving support or safety issues. For business travelers and other time-sensitive riders, downstream reliability matters even though the direct product surface is driver-facing.

The metric framework should consider:

- Clear definitions for each proposed guardrail, including numerator, denominator, time window, and unit of analysis.

- How metrics are instrumented across the driver app event stream, trip lifecycle, marketplace matching systems, payments, support, and safety systems.

- Relevant cohorts such as geography, product line, platform, driver tenure, vehicle type, marketplace density, and connectivity conditions.

- Separation between driver experience signals, marketplace health signals, customer reliability signals, safety signals, and business/unit economics signals.

- Guardrail thresholds, alerting logic, and how to distinguish normal local variance from a meaningful degradation.

- How metrics should be interpreted during experiments, phased rollouts, seasonal demand shifts, incidents, and market-specific operational changes.

- Data quality risks, lagging indicators, instrumentation gaps, and cases where proxy metrics may be misleading.

- How the guardrails would support decision-making, including whether to continue, pause, roll back, or localize a product launch.

The goal is to propose a rigorous, globally scalable guardrail metric framework that helps Uber protect critical driver and marketplace outcomes while still enabling product teams to ship improvements to the Driver App with confidence.