How to Design Effective A/B Tests for Onboarding

Quick Overview

Evaluates A/B test design for a redesigned onboarding flow in a subscription app. Strong answers define activation, user-level randomization, primary and guardrail metrics, sample size, sequential monitoring, and decisions when activation improves but support burden rises.

How to Design Effective A/B Tests for Onboarding

Company: Netflix

Role: Data Scientist

Category: Analytics & Experimentation

Difficulty: medium

Interview Round: Onsite

##### Scenario Product team plans to launch a redesigned onboarding flow and needs evidence it increases activation. ##### Question Design an A/B test for the new onboarding. State hypothesis, unit of randomization, key metrics, guardrail metrics, and runtime calculation. If early results show uplift but increased support tickets, how would you decide whether to launch? ##### Hints Address sample size, power, sequential checks, and balancing primary vs. secondary metrics.

Quick Answer: Evaluates A/B test design for a redesigned onboarding flow in a subscription app. Strong answers define activation, user-level randomization, primary and guardrail metrics, sample size, sequential monitoring, and decisions when activation improves but support burden rises.

|Home/Analytics & Experimentation/Netflix
Netflix logo
Netflix
Jul 12, 2025, 6:59 PM
mediumData ScientistOnsiteAnalytics & Experimentation
29
0

Design Effective A/B Tests for Onboarding

A consumer subscription app is launching a redesigned onboarding flow for newly registered users. The goal is to increase activation, defined for this prompt as starting to play any title within 7 days of signup unless your organization uses a different definition.

Constraints & Assumptions

  • Users should be randomized at signup and assigned persistently.
  • Define exposure, eligibility, and activation precisely.
  • Include primary metrics, guardrails, power, runtime, monitoring, and decision rules.
  • Handle early mixed results carefully.

Clarifying Questions to Ask Guidance

  • What is the current baseline activation rate?
  • What onboarding steps changed?
  • Are users exposed across multiple devices?
  • What support, retention, or subscription outcomes could be affected?

Part 1 - Experiment Setup

State the hypothesis, unit of randomization, and treatment/control definitions.

What This Part Should Cover Guidance

  • New-user eligibility, sticky user-level randomization, exposure definition, treatment and control flows, and hypothesis.

Part 2 - Metrics

What primary, secondary, and guardrail metrics would you use?

What This Part Should Cover Guidance

  • Activation within 7 days, onboarding completion, first play, subscription conversion, retention, and engagement.
  • Guardrails such as support tickets, cancellations, latency, crashes, user complaints, and long-term retention.

Part 3 - Power and Monitoring

How would you calculate sample size/runtime and monitor the test?

What This Part Should Cover Guidance

  • Baseline rate, MDE, alpha, power, traffic volume, experiment duration, sequential monitoring rules, SRM, and instrumentation checks.

Part 4 - Decision Framework

What would you do if early results show activation uplift but support tickets increase?

What This Part Should Cover Guidance

  • Predefined guardrail thresholds, severity analysis, segment diagnostics, practical significance, iteration, holdout, or ramp decision.

What a Strong Answer Covers Guidance

A strong answer defines the activation metric and exposure cleanly, designs a powered user-level experiment, monitors guardrails, and makes decisions based on both user value and operational risk.

Follow-up Questions Guidance

  • What if activation rises but 30-day retention falls?
  • How would you avoid peeking bias?
  • How would you analyze treatment effects by user segment?
Loading comments...