Airbnb — Marketplace — Product Sense & Metrics

Role context: Data Scientist, Marketplace Data Science · Est. study time: 35 min · 4 practice questions

How to prepare for this role

Airbnb Marketplace DS builds the inference frameworks and intelligence to grow a two-sided marketplace. It's the most causal-inference and experimentation-heavy of the three tracks, judged on thinking at the market level, reasoning about supply and demand, and designing experiments that survive marketplace interference.

Airbnb data science is analytics, experimentation, and causal-inference heavy. On this track the day-to-day is: measure and grow supply and demand per market, understand how their balance drives bookings, and build the experiment designs that other DS teams rely on to measure marketplace changes honestly. The defining instinct this role must have: a marketplace change makes a user-level A/B test biased, because treated and control participants compete for the same finite supply and demand.

Where to spend your prep time:

  • Marketplace product sense (this article) — supply/demand balance, why you think in markets not users, and what "unmet demand" means.
  • Causal inference and interference — cluster randomization by market, listing-side vs guest-side designs, switchbacks, shadow-price corrections, and the bias/variance trade-off. See the A/B Test & Causal Inference section.
  • SQL and communication — be fluent in SQL, and practice explaining a market-level effect (and why a naive A/B would have misled) to a non-technical stakeholder.

The through-line: think in markets, reason about supply and demand, and measure marketplace changes without being fooled by interference. That matters more than any method.

What Airbnb Marketplace actually is

Airbnb is a two-sided marketplace: hosts supply listings, guests supply demand, and the platform's job is to balance and match them, market by market. A "market" is roughly a location and a time — a city on given dates. Health is local: New York in June and a quiet town in November are different marketplaces with different supply/demand balance, and a global average hides both.

Two facts define the work:

  • Supply and demand must balance locally. Too little supply in a hot market means guests can't book (unmet demand); too much supply in a cold market means hosts get no bookings and churn. Marketplace DS measures and grows both sides and the matching between them, per market.
  • Interference is the signature technical challenge. Because buyers and sellers in a market are connected, an intervention on one participant affects others. So a Bernoulli (user-level) A/B test violates the stable-unit-treatment-value assumption and gives a biased total effect — the treated units steal bookings from control units in the same market. This is the core of Airbnb's public experimentation research, which studies listing-side vs guest-side randomization, cluster randomization by market, and shadow-price corrections. For a Marketplace DS, respecting interference is the job.