Design an experiment to launch fractional shares

Quick Overview

This question evaluates a candidate's competency in experimentation design, product analytics, statistical power and sample-size calculation, causal inference, and risk identification within a brokerage product context.

Design an experiment to launch fractional shares

Company: Robinhood

Role: Data Scientist

Category: Analytics & Experimentation

Difficulty: easy

Interview Round: Onsite

## Context You work on a brokerage/investing app. The team is considering launching **fractional share trading** (users can buy/sell fractions of a stock/ETF instead of whole shares). ## Task 1. **Explain the expected benefits** of fractional shares for the business and for users. 2. Propose an **experimentation plan** to evaluate the launch: - What is the **primary success metric**? - What **diagnostic metrics** help explain movement? - What **guardrail metrics** ensure you don’t harm users or the business? - Define the **unit of randomization** (user/account/household/etc.) and the treatment/control experience. - Call out key risks: interference/network effects, novelty effects, seasonality, and any selection bias. 3. **Sample size / power** - Show how you would estimate the **required sample size** for the primary metric (state assumptions such as baseline rate/variance, MDE, \(\alpha\), and power). - If you need to **adjust the sample size mid-experiment**, explain a statistically valid approach. 4. **Constraint scenario** - If the required sample size is **too large** (you don’t have enough traffic or time), what do you do? Provide at least 3 practical options (design, metrics, variance reduction, ramp strategy, or alternative inference methods) and discuss tradeoffs. ## Output Provide a structured written plan, including metric definitions, assumptions used in power/MDE calculations, and a final recommendation on whether/how to launch.

Quick Answer: This question evaluates a candidate's competency in experimentation design, product analytics, statistical power and sample-size calculation, causal inference, and risk identification within a brokerage product context.

|Home/Analytics & Experimentation/Robinhood
Robinhood logo
Robinhood
Oct 1, 2025, 12:00 AM
easyData ScientistOnsiteAnalytics & Experimentation
9
0

Context

You work on a brokerage/investing app. The team is considering launching fractional share trading (users can buy/sell fractions of a stock/ETF instead of whole shares).

Task

  1. Explain the expected benefits of fractional shares for the business and for users.
  2. Propose an experimentation plan to evaluate the launch:
    • What is the primary success metric ?
    • What diagnostic metrics help explain movement?
    • What guardrail metrics ensure you don’t harm users or the business?
    • Define the unit of randomization (user/account/household/etc.) and the treatment/control experience.
    • Call out key risks: interference/network effects, novelty effects, seasonality, and any selection bias.
  3. Sample size / power
    • Show how you would estimate the required sample size for the primary metric (state assumptions such as baseline rate/variance, MDE, α\alpha , and power).
    • If you need to adjust the sample size mid-experiment , explain a statistically valid approach.
  4. Constraint scenario
    • If the required sample size is too large (you don’t have enough traffic or time), what do you do? Provide at least 3 practical options (design, metrics, variance reduction, ramp strategy, or alternative inference methods) and discuss tradeoffs.

Output

Provide a structured written plan, including metric definitions, assumptions used in power/MDE calculations, and a final recommendation on whether/how to launch.

Loading comments...