Pinterest Data Scientist Interview Questions

Pinterest·Data Scientist·Technical Phone Screen

Jul 2026

Pinterest DS interview with a meaty product analytics case built around a homepage carousel launch that showed mixed signals in post-launch metrics. The whole thing felt like a real incident debrief more than a textbook question, which I appreciated but also wasn't fully ready for.

A new horizontal recommendations carousel was launched on the home page. Post-launch, the treatment group shows a drop in home-page CTR while app-wide DAU and total time spent stay flat. How would you define success metrics and guardrails for this surface?
Walk through a step-by-step diagnosis plan for why home-page CTR dropped while DAU and time spent remained unchanged. What logs or queries would you actually run?
What segmentation cuts would you use to localize the CTR drop? Name at least several specific dimensions.
Based on your diagnosis, what would you recommend as next steps? How do you decide between rolling back the feature versus iterating on it?

“I jumped straight to CTR and kind of forgot to think about what 'good' CTR even means for a carousel.”

View Post

Pinterest·Data Scientist·Technical Phone Screen

Jul 2026

Pinterest data scientist interview that was basically a pandas deep-dive. Four tasks, all coding, all interconnected, and the nested dict stuff at the end was where I started to feel the pressure.

Given a DataFrame where each row has a list of event dicts and a nested attrs dict, explode the events into a long-format table with one row per event, extracting fields like event_type, event_ts, amount, and nested attrs fields. No row-level for-loops allowed.
Aggregate the exploded event table into a per-user features table using groupby with named aggregations. For fields like version and beta flag, take the value associated with the user's most recent event.
From the per-user features table, compute a conversion rate by user segment defined by version, beta flag, and dark flag. Return one row per segment with columns for user count, purchaser count, and conversion rate.
Write a robust helper function that safely extracts nested keys from the attrs dict column using vectorized operations. Handle missing keys and None values, and explain in comments why your approach avoids SettingWithCopy issues and preserves vectorization.

“The explode part I got pretty fast.”

View Post

Pinterest·Data Scientist·Technical Phone Screen

Jul 2026

Pinterest DS interview that was basically a gauntlet of SQL and pandas tasks built around a shopping engagement schema. No behavioral stuff, just pure technical work. The questions were detailed enough that I kept second-guessing whether I was overcomplicating things.

Write a single SQL query returning daily shopping engagement metrics for each date in a 7-day window, including DAU, click counts, average positive stay time, and a rolling 7-day unique user count. Deduplicate clicks within 5 minutes per user-pin pair using window functions, and use a generated dates CTE to fill in days with zero activity.
Compute next-day retention: among users who had at least one shopping click on 2025-08-31, what percentage also had a shopping click on 2025-09-01?
For the 7-day window ending 2025-09-01, return each user's top 2 pins by deduplicated shopping click count, with ties broken by total positive stay time and then by smallest pin ID.
In pandas, map product category codes to names using a provided dictionary so that any missing or unrecognized codes fall back to 'Unknown', replace negative stay_time_sec values with NaN, fill remaining NaN stay times with 0 for aggregation while still excluding zeros from averages, sort the DataFrame by user ID ascending then event timestamp and stay time descending, and finally compute each user's top product category by total positive stay time over the last 7 days with alphabetical tie-breaking.

“This one took me a while to structure.”

View Post

Pinterest·Data Scientist·Technical Phone Screen

Jul 2026

Pinterest DS interview with a fairly involved SQL question. Single problem but they packed a lot of requirements into it, more than I expected for a phone screen.

Given three tables (events, users, and pin_classification), write a single SQL query to compute click-through rate by pin format for new US users, where 'new' means the user signed up within 30 days before the event date.

“The filtering logic is where I almost tripped up.”

View Post

Pinterest·Data Scientist·Technical Phone Screen

Jul 2026

Pinterest data science interview with a graph-based coding problem. The question was more algorithmic than I expected for a DS role, felt like it belonged in a software engineering loop.

Given the membership lists of all Pinterest boards, write a function to determine whether two pins are connected through any chain of shared boards. Follow-up: what is the minimum number of boards needed to connect them?

“I stared at this for a bit before realizing it's a graph reachability problem.”

View Post

Pinterest·Data Scientist·Technical Phone Screen

Jul 2026

Pinterest data science interview with a pandas-heavy coding question. The problem looked straightforward but had a bunch of data quality landmines buried in it that you had to catch on your own.

Given a DataFrame with user engagement data on pins, filter to video pins (handling case variations and typos like 'vedio'), normalize and map categories using a provided dictionary, exclude invalid time values, and return the canonical category with the highest average time spent along with that average rounded to two decimals.

“The typo thing ('vedio') is the kind of thing you either catch or you don't.”

View Post

Pinterest·Data Scientist·Technical Phone Screen

Jul 2026

Pinterest data scientist interview with a system design coding question that was heavier on CS fundamentals than I expected for the role. The thread-safety follow-up is where things got interesting.

Implement an LRU cache with O(1) get and put operations, then walk through how you'd make it thread-safe for concurrent access.

“The data structure part I knew cold.”

View Post

Pinterest·Data Scientist·Technical Phone Screen

Jul 2026

Pinterest data scientist coding screen, one question the whole time. Pretty focused on pandas manipulation with some edge case gotchas baked in. Left feeling okay about it but not confident.

Given a pandas DataFrame with pin engagement data and a dict mapping category IDs to names, write Python to find the category with the highest average time spent among video pins only. Filter out null category IDs and non-positive time values, break ties by lexicographic order on category name, and return a tuple of (category_name, avg_time) rounded to two decimals. Solution should run in O(n) time with O(k) extra space for k distinct categories.

“I started grouping by category_id right away and almost forgot to filter the nulls before the join with the category_map dict.”

View Post

Pinterest·Data Scientist·Onsite - Behavioral / Leadership

Jul 2026

Pinterest data scientist loop, behavioral heavy with a technical depth probe baked in. Four main prompts, all requiring concrete numbers and follow-up resilience. The prep bar here is high.

Tell me about a time you disagreed with a product manager's prioritization decision that was based on shaky data. How did you push back and influence the outcome without having direct authority?
Describe a project where success metrics were undefined at the start and requirements kept shifting. How did you settle on a north-star metric, what guardrails did you put around it, and how did you explain your trade-offs to senior leadership?
Walk me through a project where you owned the full analysis from data pull to final recommendation. Be ready for follow-up questions on model assumptions, how you checked data quality, how you handled missing values, and why you didn't just use a simpler approach.
Tell me about a time you ran out of strong examples during an interview or intense questioning session. How did you stay composed, and what did you change going forward?

“This one tripped me up because I kept wanting to frame it as 'I was right, they were wrong' and that reads badly.”

View Post

Pinterest·Data Scientist·Technical Phone Screen

Jul 2026

Pinterest Data Scientist interview with a system design flavored coding question around streaming data. The problem was well-scoped but the follow-up on distribution is where things got interesting.

Design a data structure that continuously ingests a stream of integers and can return the top-K largest values seen so far at any point. How would you scale this for very large K or across distributed data streams?

“Started with a min-heap of size K which felt right, and they seemed fine with that.”

View Post