Snap Data Scientist Interview Questions

Snap·Data Scientist·Technical Phone Screen

Jul 2026

Snapchat data scientist interview that went deep on Random Forest internals. Not a surface-level ML chat, they wanted you to actually know why things work, not just that they work. Pretty rigorous for a single technical round.

Walk through every source of randomness in a Random Forest and explain how each one affects model bias and variance.
Given a dataset with 100k rows, 100 features, and a 5% positive rate, how would you choose n_estimators, max_depth, and max_features? Specifically explain how max_features controls tree correlation.
Compare out-of-bag error to k-fold cross-validation. Under what conditions might they give different results, and why?
Why does impurity-based feature importance systematically favor continuous or high-cardinality features? How would you fix it?
What strategies exist for handling class imbalance in a Random Forest, and what are the downstream effects on probability calibration and decision thresholds?

“This tripped me up more than I expected.”

View Post

Snap·Data Scientist·Technical Phone Screen

Jun 2026

Snapchat data scientist interview with a meaty experiment design question centered on a spam filter rollout. The whole thing was one long multi-part problem and it went pretty deep into stats and causal inference territory.

You're launching a stricter spam filter and need to evaluate its effect on friend request dynamics. Walk through your full experiment design: hypotheses, randomization, metrics, power analysis, and how you'd handle incomplete spam labels.

“This was basically five questions in a trenchcoat.”

View Post

Snap·Data Scientist·Technical Phone Screen

Jun 2026

Snapchat data scientist interview with a heavy SQL and Python focus, all centered around a user events table. The questions were legitimately hard and the median calculation one in particular made me feel like I had no idea what I was doing for a solid minute.

Using a user_events table, write a SQL query to compute 7-day retention (inclusive) for users whose first visit was on 2023-04-01.
Write a SQL query returning the top 3 pages ranked by distinct purchasing users over the last month.
Find the median number of daily events per active user in April 2023 using SQL.
Given a Pandas DataFrame matching the user_events schema, write Python code to produce an hourly time-series of clicks per page for the past 24 hours.

“Went straight for a CTE to isolate first-visit users, then joined back on the same table filtering for events within 7 days.”

View Post

Snap·Data Scientist·Onsite - Behavioral / Leadership

Jun 2026

Snapchat onsite for a product data science role, behavioral and leadership heavy. The whole thing was framed around a cross-functional product launch scenario where you're driving outcomes without owning anyone's roadmap, which sounds manageable until you're actually in the room.

Walk me through your background and why it makes you a strong fit for a product data science role.
Tell me about a time you got a team or partner org to move in a direction you wanted, without having any formal authority over them.
What's the hardest piece of feedback you've received, and what did you actually do with it?
When leadership, data signals, and UX research are all pointing in different directions, how do you figure out what to do?

“I had a version of this answer ready but I think I leaned too hard into the technical side and not enough on the product judgment angle.”

View Post

Snap·Data Scientist·Technical Phone Screen

Jun 2026

Snapchat Data Scientist interview with a pandas wrangling task on story engagement data. Pretty straightforward if you know your groupby patterns, but the filtering step tripped me up a bit.

Given a story engagement dataset, use pandas to find users who have posted at least three group stories, and separately aggregate total views and likes broken down by story type.

“I started with the aggregation part first because it felt easier, groupby story_type and sum views and likes, that was fine.”

View Post

Snap·Data Scientist·Technical Phone Screen

Jun 2026

Snapchat data scientist interview that went deep on a churn prediction pipeline scenario. The whole session basically revolved around one big multi-part question covering the full ML lifecycle, which felt intense but at least kept things focused.

For a weekly churn prediction pipeline running on 10M users, walk through how you'd approach feature engineering, model selection, and hyperparameter tuning.
Why would you choose Gradient Boosted Trees over Logistic Regression for this kind of churn problem?
Describe two approaches for explaining model predictions to stakeholders who don't have a technical background.
Recall drops 15% week-over-week on your churn model. What's your debugging process?

“I started with feature categories (recency, frequency, engagement decay) and mentioned handling class imbalance with SMOTE vs just reweighting the loss.”

View Post

Snap·Data Scientist·Onsite - Behavioral / Leadership

Jun 2026

Snapchat data scientist behavioral round, pretty much one big meaty question about influencing a senior stakeholder when your A/B test data isn't giving you clean answers. Left feeling like I either nailed it or completely overexplained everything.

Tell me about a time you had to convince a senior leader from another team to change a launch plan when your A/B test results were ambiguous. Walk through the stakes, your hypothesis, how you set success metrics and guardrails, how you navigated disagreement, what artifacts you produced, the trade-offs you surfaced, the outcome, and what you'd do differently.

“This question is basically five questions stapled together and they want all of it.”

View Post

Snap·Data Scientist·Technical Phone Screen

Jun 2026

Snapchat data scientist interview with a meaty A/B testing design question. The whole session was basically one long scenario, but it went several layers deep, so there was a lot to keep track of.

Design an A/B test to measure the impact of a new recommendation algorithm on GMV. What primary and guardrail metrics would you track, and why?
How would you calculate the required sample size for this test, given a 3% baseline conversion rate, a 7% relative lift target, alpha of 0.05, and 80% power?
How would you handle the novelty effect and potential seasonality imbalances between test and control groups?
If the primary metric shows no significant change but secondary engagement metrics improve, how do you interpret and communicate those results?

“I went with GMV per user as primary and latency plus user retention as guardrails.”

View Post

Snap·Data Scientist·Technical Phone Screen

May 2026

Snapchat data scientist interview that went deep on A/B testing, and I mean really deep. One question, but it sprawled into six or seven distinct sub-problems covering everything from sample size math to bot filtering. Left feeling like I'd half-answered a take-home assignment out loud.

Walk through an end-to-end A/B test for adding a home-page banner: randomization unit, metric selection, sample size calculation, analysis approach, diagnostic checks, and launch decision criteria.

“This started as one question and quietly became six.”

View Post

Snap·Data Scientist·Technical Phone Screen

May 2026

Snapchat data scientist technical screen, pretty much a logistic regression gauntlet from start to finish. They went way deeper on the math than I expected, like full gradient and Hessian derivations, not just 'explain logistic regression at a high level.' Walked out unsure if I'd nailed it or completely embarrassed myself.

Write out the sigmoid function and the Bernoulli log-likelihood for binary logistic regression. Then derive the gradient and Hessian with respect to the coefficients under L2 regularization, and explain why the objective is convex.
With a single-feature model where the intercept is -1.2 and the coefficient is 0.8, compute the predicted probability for x = 2.0. Also report the odds and the odds ratio for a one-unit increase in x.
If the positive class occurs only 2% of the time and a false negative is ten times more costly than a false positive, what is the Bayes-optimal decision threshold? How would you calibrate model probabilities, and what are the tradeoffs between Platt scaling and isotonic regression?
What are the numerically stable forms for log(sigmoid(z)) and log(1 minus sigmoid(z)), and why do the naive expressions cause problems?
How does severe class imbalance affect maximum likelihood estimates in logistic regression, and what regularization or reweighting strategies would you apply? Back it up analytically.

“This is where I started sweating.”

View Post