Data Analyst Interview Questions India 2026 — SQL, Python, Stats & Case Studies

SQL is asked in virtually every data analyst interview in India. The questions that actually differentiate candidates: 1. Window functions: ROW_NUMBER(), RANK(), DENSE_RANK(), LAG(), LEAD(), PARTITION BY. These are asked at every product company. Example: 'Find the 2nd highest salary per department.' SELECT * FROM (SELECT *, DENSE_RANK() OVER (PARTITION BY dept ORDER BY salary DESC) as rk FROM employees) t WHERE rk = 2; 2. Self joins: 'Find all employees who earn more than their manager.' 3. Aggregation + HAVING: 'Find departments with more than 5 employees earning above ₹10L.' 4. Common Table Expressions (CTEs): Readable alternative to subqueries. Interviewers at Flipkart and Amazon specifically look for CTE usage as a signal of SQL maturity. 5. Query optimisation: 'How would you optimise a slow query?' — Cover indexing (B-tree vs hash), explain plan, avoiding SELECT *, avoiding functions on indexed columns in WHERE clause.

Product companies (Flipkart, Swiggy, Meesho) typically test Python/Pandas for data manipulation: Must-know Pandas operations: • read_csv(), head(), describe(), info(), value_counts() • groupby() + agg() — 'Find average order value per city' • merge() — equivalent of SQL joins • pivot_table() — aggregate with multiple dimensions • apply() + lambda — custom transformations • Handle missing values: fillna(), dropna(), isnull() Typical question: 'Given a dataframe of orders, find the top 5 customers by total spend in the last 30 days, excluding cancelled orders.' Expected: filter, groupby, sort, head(5) — all in 5–10 lines of clean Pandas.

Goldman Sachs, JPMorgan, Amazon, and Flipkart analytics roles test statistics: Core topics: • Probability: Bayes theorem, conditional probability, distributions (normal, binomial, Poisson) • Hypothesis testing: p-value, Type I/II error, confidence intervals • A/B testing: How to set up, minimum sample size calculation, multiple testing correction • Regression: Linear regression assumptions (LINEARITY), R-squared interpretation, multicollinearity • Central Limit Theorem: Why it matters for sampling-based analysis A/B testing question (very common): 'We ran an A/B test for 2 weeks. Variant B has a 3% higher conversion rate. Is this statistically significant? How do you decide?' Expected answer: Define significance threshold (p<0.05), calculate minimum sample size before starting, check if duration was sufficient, use t-test or chi-squared, report confidence interval not just p-value.

At product companies (Swiggy, Zomato, Flipkart, Meesho), expect business case questions: 'Swiggy orders dropped 15% last Tuesday between 7-9 PM. How would you diagnose the issue?' Framework: 1. Confirm the data: Is the drop real? Check for data pipeline issues first. 2. Narrow the scope: Is it all cities or specific ones? All restaurants or specific cuisines? Specific device types? 3. External vs internal: Weather event? Competitor promotion? App crash? Server downtime? 4. Hypothesis tree: For each dimension, form a hypothesis and identify the data that would confirm/deny it. 5. Recommendation: What immediate action and what monitoring to put in place. This type of question tests structured thinking, not SQL knowledge — practice narrating your diagnostic process out loud.