Data Scientist interview
at Netflix.

Tell-me-about-a-time stories on ownership, conflict, and impact — scored against this company's real values.

1. 01 Tell me about a time you delivered candid, evidence-backed pushback to a senior leader whose proposed metric or experiment design you believed was wrong, and how you reached alignment.
2. 02 Describe a high-stakes decision you made and owned end-to-end with little oversight, embodying Netflix's 'context, not control' philosophy rather than waiting for direction.
3. 03 Walk me through an experiment or analysis you got wrong, what you misjudged, and what you'd do differently.
4. 04 Tell me about a time you simplified or killed a metric, model, or process that was adding complexity without adding value, even though others wanted to keep it.

The hard screen for the craft itself — talked through out loud, not whiteboarded in your head.

1. 01 You run an A/B test on a new home-screen ranking model but 20% of members have zero watch time in the window, so which metric and statistical test would give you a stable, decisive read?
2. 02 A trailer-autoplay variant shows higher play-starts but unchanged total watch time, so how do you decide whether this is a genuine UX win or just metric inflation?
3. 03 When testing many ranking variants simultaneously, how would you control the false discovery rate across all the comparisons?
4. 04 Explain how you'd use a difference-in-differences or instrumental-variable approach to estimate the causal impact of a feature you couldn't cleanly randomize.

Open-ended design of the systems this company actually runs, with the interviewer probing your tradeoffs.

1. 01 Design the experimentation platform that lets Netflix run thousands of concurrent A/B tests across personalization, artwork, and UI without tests interfering with one another.
2. 02 Design the metrics and measurement framework to evaluate whether a new recommendation model causally increases long-term member retention rather than short-term engagement.
3. 03 Design an interleaving or quasi-experimental system to compare two home-screen ranking algorithms with high sensitivity at Netflix's scale.
4. 04 Design the causal-inference pipeline for measuring the incremental value of a new original title on subscriptions and retention.