Katkat
Back

Investigate a 7% Monthly Active Riders Drop and a 20% Wait-Time Increase

LyftData Scientist · Technical Screen
MediumData ScientistTechnical ScreenAnalytics & Experimentation

Last updated

Tests your ability to conduct a structured root-cause analysis on simultaneous marketplace metric regressions, decomposing supply-demand dynamics and separating correlation from causation.

Scenario

During a monthly executive business review at a rideshare company like Lyft, the core marketplace monitoring dashboard flags two simultaneous anomalies across the past 30 days:

  1. Monthly Active Riders (MAR) dropped by 7% month-over-month (MoM).
  2. Median Rider Wait Time (time elapsed from ride request to driver pickup) increased by 20% MoM.

Leadership is concerned that worsening reliability is driving rider churn, but there is also debate about whether external demand headwinds or recent platform dispatch updates are at play. You are tasked with leading the diagnostic investigation.

Investigation Framework

Structure your response to address the following areas:

1. Data Integrity & Verification

  • What checks would you run first to ensure these shifts reflect genuine user behavior rather than instrumentation bugs, logging dropouts, or definition changes?
  • How does the reporting calendar (e.g., 28 days vs. 30/31 days, seasonality, holiday shifts) influence these metrics?

2. Metric Decomposition

  • Deconstruct Monthly Active Riders: Break down riders into meaningful lifecycle cohorts (e.g., new vs. retained vs. resurrected) and journey funnel stages (open app -> set destination -> request ride -> complete ride).
  • Deconstruct Wait Time: Break wait time into sub-intervals (dispatch matching latency, driver accept time, driver arrival/transit time) and explore distribution shifts beyond the median (such as p90 or p95 wait times).

3. Hypothesis Generation & Supply-Demand Dynamics

Formulate at least three mutually exclusive hypotheses that could explain both observations simultaneously, covering:

  • Supply constraints: Reductions in driver supply or changes in driver behavior.
  • Pricing / Matching changes: Algorithmic adjustments to dispatch radius, batching windows, or dynamic pricing multipliers.
  • External / Compositional shifts: Geographic concentration changes, severe weather, regulatory shifts, or competitor promotions.

4. Isolating Causality & Recommendations

  • How would you test whether increased wait times caused the drop in MAR, versus both metrics being downstream symptoms of an unobserved confounding factor?
  • What targeted exploratory analyses (e.g., session drop-off rates, cancellation rates by wait-time bin) would confirm your primary hypothesis?
  • Outline actionable mitigations or experiments to address the issue once identified.

Hints

Hint 1

Start by slicing the aggregate metrics across key dimensions such as geographic market, ride tier, and time-of-day before forming algorithmic hypotheses.

Hint 2

Deconstruct wait time into dispatch delay, driver acceptance rate, and actual transit time to locate exactly where the 20% delay originates.

Hint 3

Evaluate whether rider search-to-request conversion dropped sharply in sessions where initial quoted ETAs exceeded historical baselines.

Follow-up questions

  • How would you determine if riders are abandoning the platform due to quoted ETAs (pre-request drop) versus actual wait times (post-request cancellations)?
  • If the root cause is driver supply contraction driven by rising fuel costs, what short-term marketplace interventions would you test first?
  • How would you establish leading indicator metrics to detect localized marketplace imbalances before they compound into a 7% top-line monthly drop?