← DoorDash Interview Insights

DoorDash·Data Scientist·Technical Phone Screen·Senior

Senior
May 2026

Summary

DoorDash DS interview that went deep into launch metrics and experimentation design. The question was essentially one long case but it covered a lot of ground, so be ready to stay organized under pressure.

Questions Asked (2)

Q1

DoorDash is launching in a new city. Define success metrics and guardrails for three phases: pre-launch readiness, soft launch, and full ramp. For each phase, specify core metrics with precise definitions, distinguish leading indicators from outcome metrics, and set guardrail thresholds that would trigger a rollback or pause.

Product Analytics & MetricsProduct Strategy
Author's notes

The three-phase structure sounds clean until you're actually in it and realize pre-launch has almost no live data to work with.

Create a free account to read the full note

AI HintsAI Generated

Suggested Approach

Structure your answer around the three phases, defining for each a small set of core metrics with precise formulas, explicitly labeling leading vs. outcome metrics, and setting guardrail thresholds that trigger pause/rollback. Tie every metric to a clear business question (e.g., can we acquire supply, match demand, and deliver reliably?) and show how thresholds tighten or loosen as confidence grows. Use a consistent framework like acquisition → activation → retention → marketplace health to avoid missing key dimensions.

Pro tip: Anchor guardrails to the unit economics and customer experience—e.g., delivery time, cancellation rate, and contribution margin—and specify that thresholds should be set relative to mature-market benchmarks with a margin for new-market volatility. This shows you understand that a launch is a learning exercise, not just a growth sprint.

1. Pre-Launch Readiness: Supply & Ops Foundations

Focus on whether the market can support launch: Dasher acquisition and onboarding, restaurant partnerships, and operational setup. Core metrics: Dasher sign-up completion rate (leading), restaurant activation rate (leading), and projected delivery time from simulation (outcome proxy). Guardrails: if Dasher activation < 60% of target or restaurant activation < 70% of target, delay launch.

2. Soft Launch: Demand Generation & Early Experience

Test the full loop with limited users and geographies. Core metrics: order volume (outcome), first-order conversion rate (leading), delivery time P90 (outcome), and cancellation rate (guardrail). Guardrails: if P90 delivery time > 45 min or cancellation rate > 5%, pause and fix ops before scaling.

3. Full Ramp: Scaling & Marketplace Health

Scale demand and supply while monitoring marketplace balance and unit economics. Core metrics: weekly order growth (outcome), Dasher utilization (leading), customer retention (outcome), and contribution margin per order (guardrail). Guardrails: if Dasher utilization < 60% or contribution margin < -$2 per order, roll back marketing spend and re-evaluate.

4. Define Leading vs. Outcome Metrics & Thresholds

For each phase, explicitly label which metrics are leading (predictive, actionable) vs. outcome (lagging, business results). Set guardrail thresholds as absolute values or relative to mature markets, with clear rollback/pause triggers. Example: leading = Dasher acceptance rate; outcome = order completion rate; guardrail = acceptance rate < 70% triggers pause.

5. Establish Review Cadence & Escalation

Specify how often metrics are reviewed (daily during soft launch, weekly during ramp) and who owns the decision to pause/rollback. Include a plan for iterating thresholds as data accumulates. This shows operational rigor and cross-functional awareness.

Key Points to Mention

  • Precise metric definitions with formulas (e.g., delivery time P90 = 90th percentile of order delivery duration from order placement to drop-off).
  • Leading indicators (e.g., Dasher sign-up completion rate, restaurant activation rate) vs. outcome metrics (e.g., order volume, retention).
  • Guardrail thresholds tied to customer experience (delivery time, cancellation rate) and unit economics (contribution margin per order).
  • Phase-specific goals: pre-launch = readiness, soft launch = learning and fixing, full ramp = scaling efficiently.
  • Rollback/pause triggers with clear ownership and review cadence (e.g., daily standups during soft launch).
  • Benchmarking against mature markets and adjusting for new-market volatility (e.g., set thresholds 20% looser initially).

AI-generated suggestions, not part of the candidate's original notes. May be inaccurate — verify before relying on them.

Q2

Design a geo-based A/B or staggered rollout to estimate the causal impact of the city launch on orders and ETAs. Address seasonality, matched-control zone selection, spillover effects across adjacent zones, novelty effects, and minimum detectable effect assumptions. What decision rules would you present to leadership after two weeks?

A/B Testing & ExperimentationProduct Analytics & MetricsAdaptability & Ambiguity
Author's notes

This is the part that really got me.

Create a free account to read the full note

AI HintsAI Generated

Suggested Approach

Frame the problem as a quasi-experiment where the city launch is the treatment, and use a matched control zone (or synthetic control) to estimate the counterfactual. Address each threat to validity (seasonality, spillover, novelty) with specific design and analysis choices, and propose decision rules that balance statistical rigor with business practicality.

Pro tip: Emphasize that with only two weeks of data, you'll likely need to rely on pre-registered decision rules and guardrail metrics rather than definitive causal claims; propose a sequential testing framework or Bayesian approach to make interim decisions while controlling error rates.

1. Define treatment and control zones

Identify the launch city as the treatment zone and select matched control zones using pre-launch data on orders, ETAs, demographics, and seasonality. Use propensity score matching or synthetic control methods to create a comparable counterfactual.

2. Address seasonality and time effects

Use difference-in-differences (DiD) with pre-period data to adjust for common seasonal trends. Include time fixed effects and possibly day-of-week/hour controls, and test for parallel trends pre-launch.

3. Mitigate spillover and novelty effects

Define a buffer zone around the treatment city to exclude adjacent areas from control, and monitor for spillover via cross-border orders. For novelty, plan to analyze early vs. later weeks and consider a longer horizon if possible.

4. Power analysis and MDE

Conduct a power analysis using historical variance to determine the minimum detectable effect (MDE) given the sample size and two-week duration. State assumptions (e.g., baseline orders, variance, desired power) and discuss whether the MDE is practically relevant.

5. Decision rules for leadership

Propose pre-registered decision rules based on primary metrics (orders, ETAs) and guardrails (e.g., cancellation rates). For two weeks, suggest interim readouts with confidence intervals and a recommendation to continue, expand, or halt based on whether effects exceed MDE and guardrails are not violated.

Key Points to Mention

  • Difference-in-differences (DiD) or synthetic control methods to estimate causal impact.
  • Matched control selection using pre-launch data and propensity scores.
  • Spillover mitigation via buffer zones and monitoring cross-zone effects.
  • Novelty effect analysis by comparing early vs. later weeks and using longer-term trends.
  • Power analysis and MDE calculation with explicit assumptions.
  • Pre-registered decision rules and guardrail metrics for leadership.

AI-generated suggestions, not part of the candidate's original notes. May be inaccurate — verify before relying on them.