TL;DR
Data science interviews test five areas: Statistics (hypothesis testing, probability),Machine Learning (algorithms, evaluation), SQL (joins, window functions),Coding (Python/pandas), and Business Cases (metrics, experimentation). The best candidates explain the "why" behind their answers-not just the "what."
The Five Types of Data Science Questions
Data science interviews are uniquely broad. You need to demonstrate both technical depth and business thinking.
| Type | What It Tests | Example |
|---|---|---|
| Statistics | Probability, hypothesis testing, experimental design | "How would you design an A/B test?" |
| Machine Learning | Algorithms, model selection, evaluation | "Explain the bias-variance tradeoff" |
| SQL | Data manipulation, query writing | "Find the second-highest salary" |
| Coding | Python, pandas, data wrangling | "Clean and transform this dataset" |
| Case Study | Business thinking, metrics, communication | "How would you measure the success of Reels?" |
Statistics & Probability Questions
These questions test your foundational understanding. Many candidates memorize formulas but can't explain the intuition.
Probability
- What's the probability of getting at least one 6 when rolling a die 4 times?
- You have two coins: one fair, one always heads. You pick one randomly and flip it 3 times, getting 3 heads. What's the probability it's the fair coin?
- Explain Bayes' theorem. Give a real-world example.
- What's the expected number of coin flips to get two heads in a row?
- In a room of 23 people, what's the probability two share a birthday?
Statistics & Hypothesis Testing
- Explain the difference between Type I and Type II errors. Which is worse?
- What's a p-value? What does p = 0.05 actually mean?
- When would you use a t-test vs. a z-test?
- Explain the Central Limit Theorem. Why does it matter?
- What's the difference between correlation and causation? Give an example.
- What is statistical power? How do you increase it?
- Explain confidence intervals. What does 95% confidence actually mean?
Sample Answer: Explaining p-values:
Wrong answer: "p-value is the probability the null hypothesis is true."
Right answer: "The p-value is the probability of observing data at least as extreme as what we observed, assuming the null hypothesis is true. A p-value of 0.05 means there's a 5% chance of seeing this result (or more extreme) if there's actually no effect. It does NOT tell us the probability the null is true-that requires Bayesian analysis."
Common Mistake
Many candidates incorrectly state that "p-value is the probability the null hypothesis is true." This fundamental misunderstanding can cost you the interview.
Machine Learning Questions
Interviewers want to know that you understand algorithms deeply-not just how to import sklearn.
Fundamentals
- Explain the bias-variance tradeoff.
- What is regularization? Explain L1 vs L2.
- How do you handle overfitting?
- What's the difference between bagging and boosting?
- Explain gradient descent. What's the role of learning rate?
- What's cross-validation? When would you use k-fold vs. leave-one-out?
Algorithms
- Explain how a decision tree works. How does it decide where to split?
- What's the difference between Random Forest and Gradient Boosting?
- How does logistic regression work? What assumptions does it make?
- Explain SVMs. What is the kernel trick?
- How does k-means clustering work? What are its limitations?
- Explain PCA. When would you use it?
Model Evaluation
- When would you use accuracy vs. AUC-ROC vs. F1?
- Explain precision vs. recall. When does each matter more?
- How do you evaluate a recommendation system?
- What metrics would you use for an imbalanced classification problem?
Framework: Explaining ML Algorithms
1. Intuition: What's the core idea in plain English?
2. How it works: Brief description of the algorithm
3. Assumptions: What conditions need to be true?
4. Pros/Cons: When to use it, when not to
5. Practical tips: Hyperparameters, common issues
SQL Questions
SQL is tested in almost every data science interview. Practice writing queries without auto-complete.
- Find the second-highest salary from an employees table.
- Write a query to find duplicate emails in a users table.
- Calculate a 7-day rolling average of daily active users.
- Find customers who made purchases in January but not February.
- Rank employees by salary within each department.
- Find the most popular product in each category.
- Calculate month-over-month growth in revenue.
- Write a query to find gaps in sequential IDs.
SQL Pattern: Window Functions
Window functions are crucial for data science SQL interviews:
ROW_NUMBER(): Assign unique ranks
RANK() / DENSE_RANK(): Handle ties differently
LAG() / LEAD(): Access previous/next rows
SUM() OVER(): Running totals
AVG() OVER(ROWS BETWEEN): Moving averages
Python/Coding Questions
Expect to write Python code for data manipulation, sometimes on a whiteboard.
- Write a function to calculate the moving average of a list.
- Given a pandas DataFrame, clean and transform messy data.
- Implement a function to calculate cosine similarity between two vectors.
- Write code to handle missing values in a dataset (explain your strategy).
- Implement k-means clustering from scratch.
Business Case Study Questions
These questions test whether you can apply data science to real business problems. Structure your answer clearly-interviewers are evaluating your thinking process.
- How would you measure the success of Instagram Reels?
- Netflix wants to test a new recommendation algorithm. How would you design the experiment?
- Uber notices ride cancellations increased 10% last week. How would you investigate?
- How would you build a model to predict customer churn?
- Your A/B test shows the new feature increases engagement but decreases revenue. What do you do?
Framework: Metrics Case Studies
1. Clarify the goal: What are we optimizing for? Who's the user?
2. Define North Star metric: One metric that captures success
3. Supporting metrics: Guardrails and diagnostic metrics
4. Potential issues: Gaming, Goodhart's Law, unintended consequences
5. Tradeoffs: Short-term vs long-term, user vs business
How to Practice
Data science interviews require broad preparation across multiple domains.
- SQL: Practice on LeetCode, StrataScratch, or Mode Analytics
- Statistics: Review a good stats textbook (like "Naked Statistics")
- ML: Be able to explain algorithms, not just use them
- Coding: Practice pandas operations without auto-complete
- Case studies: Study metrics for products you use daily
The hardest part of data science interviews is communicating technical concepts clearly. You might understand the bias-variance tradeoff perfectly, but can you explain it to a non-technical PM? MORT's Interview Practice helps you practice explaining these concepts out loud-which is very different from knowing them in your head.
Company-Specific Tips
- Google: Heavy on statistics and experimentation. Know A/B testing deeply.
- Meta: Strong SQL focus. Practice complex analytical queries.
- Netflix: Product sense and metrics. Think about recommendation systems.
- Amazon: Leadership principles + case studies. Prepare stories.
- Startups: More practical, end-to-end projects. Show you can ship.
Practice data science interviews with AI
MORT's Interview Practice includes data science-specific questions covering statistics, ML concepts, and case studies. Get feedback on how clearly you explain technical concepts.