Pinterest Data Scientist Interview: Stats, ML or Forecasting Round

Last updated: October 2026

Quick Answer

The Pinterest data scientist interview runs a recruiter screen, a SQL and Python technical screen, then a virtual onsite that prep guides and candidates report as about five rounds: coding, an analytical case, a specialty round in statistics, machine learning or forecasting that you choose, and two behavioral rounds. The specialty round is about 45 minutes, per two prep guides. Rehearse the case and specialty rounds aloud in OphyAI Interview Practice, then use OphyAI Interview Copilot on the live video rounds.

Part	Detail
Technical screen	SQL (window functions), some Python or pandas; 45-60 minutes reported
Onsite	About 5 rounds reported; Pinterest’s new grad page says 3-4 technical interviews
Specialty choice	Statistics, machine learning or forecasting (prep guides)

What Pinterest Is Testing in This Round

The specialty round checks depth in the area you picked; the interviewer pushes until you reach your edge. The Pinterest interview guide covers the full loop, pay and the engineering rounds.

What they evaluate, by track:

• Statistics. Experiment design, hypothesis tests (z-tests, t-tests) and causal inference from observational data, such as difference-in-differences or synthetic control (Aced).
• Machine learning. Applying models to practical Pinterest problems, plus fundamentals such as when to use L1 versus L2 regularization (Aced, Prepfully).
• Forecasting. Time-series problems on Pinterest-style metrics (Prepfully); the least documented track.

Red flags:

• Naming a test without its assumptions. An October 2025 L5 candidate said to know “every test down cold” because follow-ups go deep.
• Treating an A/B test as only a p-value: no power, sample size, guardrails or novelty effects.

What to Expect

Early career. Pinterest’s careers blog says ML and data science new grads take CodeSignal, a 30-minute phone screen, a hiring manager review, 3-4 technical interviews, then a 30-minute call with interested hiring managers. Interns take one one-hour technical interview.

Experienced (reported). Aced describes a 45-minute SQL and case screen, then five onsite interviews: coding (SQL plus Python or R), a 60-minute analytical case on a slide deck, the 45-minute specialty round with a teammate in your specialty, a hiring manager round and a cross-functional round. Prepfully reports a one-hour screen and four to six onsite rounds, with the specialty sometimes folded into the case. A PracHub L5 report adds two team-matching rounds; about a month in total.

SQL bar. Aced says the onsite has two SQL questions, the second “significantly more difficult than the first,” and advises spending no more than 20 minutes on the SQL portion. The L5 candidate called the SQL “on the harder side,” opening on a window function.

Example Questions and How to Answer Them

From prep guides and candidate reports, not confirmed Pinterest questions.

1. “Describe an A/B test from start to finish.” (Statistics)

Approach: Hypothesis, randomization unit (user, not session), primary metric and guardrails, power and sample size, run length covering full weeks, then the test and its assumptions. Close with what you do if a win hurts a guardrail.

2. “A feature launched by geography, not randomly. Did it work?” (Statistics, causal)

Approach: Say why before-and-after fails, propose difference-in-differences against comparable regions, check parallel pre-trends, and offer synthetic control when no region matches. The L5 report describes a geo rollout case.

3. “How do you choose between L1 and L2 regularization?” (Machine learning)

Approach: L1 zeroes coefficients and selects features; L2 shrinks them smoothly and copes better with correlated features. Tie it to a Pinterest problem, such as predicting saves, and say how you would validate offline.

4. “Forecast weekly active users for next quarter.” (Forecasting)

Approach: Clarify the horizon, decompose trend and weekly or holiday seasonality, start with a simple baseline, and report intervals. Explain how you would backtest.

5. “WAU is up 5% but email open rates are down 2%. Why?” (Analytical case)

Approach: Aced lists this prompt. Check definitions first, then segment: did new users who rarely open emails drive WAU, or did send volume change?

How to Prepare in 3 Days

• Day 1: map your track. Choose the specialty you can go three follow-ups deep in; explain ten of its topics aloud. See our SQL interview questions.
• Day 2: timed practice. Two SQL problems in 20 minutes, then questions 1, 2 and 5 aloud against a timer. Our data scientist interview questions page has more.
• Day 3: mock and stories. Run a full mock of your specialty and the case in OphyAI Interview Practice, then write six behavioral stories with numbers.

On the live rounds, Interview Copilot suggests answers in real time on Zoom, Microsoft Teams or Google Meet, and Coding Copilot in the desktop app (Premium) snaps the problem window for the coding questions. In the desktop app the answer window is excluded from screen sharing, though not from a webcam or a person in the room.

Start practicing →

Common Mistakes

• Choosing the trendy track. Depth beats fashion; pick the specialty you have shipped work in.
• Thin behavioral prep. The L5 candidate reported 7-8 questions in the cross-functional round and ran out of examples.

Related Guides

• Pinterest interview guide (the full loop and pay)
• Pinterest CodeSignal assessment
• Pinterest system design interview
• Data scientist interview questions

Frequently Asked Questions

Which specialty should I choose for the Pinterest data scientist interview?

Choose the one you can defend under three rounds of follow-ups, usually the one closest to work you have shipped. Statistics is the best documented and overlaps with experiment design across the loop. Machine learning and forecasting are less described, so ask your recruiter what each covers.

How many rounds is the Pinterest data scientist onsite?

Aced and an October 2025 L5 candidate report describe five onsite rounds; Prepfully says four to six. For ML and data science new grads, Pinterest’s careers blog lists 3-4 technical interviews followed by a 30-minute call with hiring managers. Pinterest does not publish an experienced-hire round count, so ask your recruiter.

Is SQL tested in the Pinterest data scientist interview?

Yes, in every candidate report and prep guide we read. Expect SQL in the technical screen and again onsite, with joins, aggregations and window functions such as RANK, plus some Python or pandas. Aced says the second onsite question is much harder, so practice under a 20-minute limit.

How does OphyAI work in a Pinterest data scientist interview?

Rehearse your specialty, the analytical case and behavioral stories in OphyAI Interview Practice, which gives follow-up questions, a transcript and per-answer feedback. On the live video rounds, Interview Copilot suggests answers in real time, and Coding Copilot (Premium) snaps the problem window for coding questions. The desktop overlay is excluded from screen sharing, not from a webcam.

Sources and verification notes

Checked October 2026. Interview Query and 1point3acres blocked fetches and are not used. Third-party and candidate details are labelled in the text.

• Pinterest Careers, ML and data science interns and new grads: first-party; CodeSignal, 30-minute screen, 3-4 technical interviews for new grads, 30-minute hiring manager call.
• Aced (formerly Exponent), Pinterest data scientist interview: third-party; five onsite rounds, 45-minute specialty round, SQL advice, WAU case.
• Prepfully, Pinterest data scientist interview guide: third-party; specialty tracks, L1 versus L2, four to six round onsite.
• PracHub, Pinterest data scientist L5 interview experience: candidate-reported, October 2025.