Data Analyst Interview Questions India 2026: SQL, Excel, Python, and Case Studies

Foundation: GROUP BY + HAVING, subqueries, CASE WHEN for conditional bucketing, NULL handling (IS NULL, COALESCE), date functions (MONTH(), YEAR(), DATEDIFF, DATE_TRUNC, EXTRACT).

Intermediate: CTEs (WITH clause for readable multi-step queries), JOINs across multiple tables, window functions (ROW_NUMBER for deduplication, LAG for period-over-period).

Advanced (top-tier product companies): complex window functions, query optimisation reasoning, writing against denormalized data warehouse schemas.

Most commonly asked data analyst SQL question in India: 'Find the top 3 products by revenue in each product category.' SELECT * FROM (SELECT product, category, SUM(revenue) AS total_revenue, RANK() OVER (PARTITION BY category ORDER BY SUM(revenue) DESC) AS rnk FROM sales GROUP BY product, category) t WHERE rnk <= 3.

Common pattern: 'Find the month-over-month retention rate.' Requires counting users active in both month N-1 and month N divided by users active in month N-1. Written as a self-join on activity tables filtered by month.

Python skills tested at product companies: Pandas: pd.readcsv(), DataFrame filtering (df[df['col'] > val]), groupby + aggregation (df.groupby('city')['revenue'].sum().resetindex()), merge/join (pd.merge(df1, df2, on='id', how='left')), apply for custom transformations, missing value handling (fillna, dropna, isnull().sum()), datetime parsing (pd.to_datetime, dt.month, dt.year).

Visualisation: matplotlib (line, bar, scatter), seaborn (heatmaps for correlation matrices, box plots for distribution comparison, histograms). Most analyst interviews test analysis, not visualisation, in the main rounds; visualisation appears in take-home assignments.

Excel/Google Sheets (consulting, banking, FMCG analytics): VLOOKUP and XLOOKUP (=XLOOKUP(lookup, searchrange, returnrange)), SUMIF/SUMIFS and COUNTIF/COUNTIFS, pivot tables, INDEX MATCH (more flexible than VLOOKUP; looks in any direction), conditional formatting.

Toolkit by role type: product company data analyst = Python + SQL; consulting/banking/FMCG analytics = Excel + SQL.

Metrics definition: 'How would you measure the success of Swiggy's new grocery delivery feature?' Structure: (1) North Star metric (gross orders or GMV for grocery). (2) Supporting metrics (order conversion rate, reorder rate within 30 days, average order value, delivery time satisfaction). (3) Guardrail metrics (grocery growth should not harm existing restaurant order GMV by more than X%).

Root cause analysis: 'Our weekly active users dropped 15% last week. How would you investigate?' Approach: (1) Segment (all users or a specific cohort? all geographies or one city? all platforms or just iOS? all entry points or just organic?). (2) Check external events. (3) Check internal changes (product release, marketing change, pricing change). (4) Look at leading indicators (sessions, visit duration, feature usage) to find where in the funnel the drop occurs.

Experiment design: 'Design an A/B test for a new checkout flow.' Expected: hypothesis, primary metric (conversion rate), sample size via power analysis (alpha 5%, power 80%, minimum detectable effect), randomisation unit (user vs session), experiment duration, success criteria.

Distributions: normal distribution (68-95-99.7 rule), skewed distributions (positive skew: median < mean; negative skew: median > mean), when to use mean vs median (median when skewed or has outliers; mean when approximately normal).

Hypothesis testing: null hypothesis (H0: no difference), alternative (H1: difference exists), p-value (probability of result at least this extreme given H0 is true), significance level (alpha 0.05), Type I error (false positive: reject H0 when true), Type II error (false negative: fail to reject H0 when false), power (probability of correctly rejecting a false H0, typically 80%).

A/B testing: t-test for continuous metrics (revenue per user, session duration), chi-squared for proportions (conversion rate). Minimum sample size depends on baseline conversion rate, minimum detectable effect, alpha, and power.

Correlation vs causation: ice cream sales and drowning rates are correlated (both caused by summer heat); confounding variables are the standard explanation. Evaluators check whether you would draw incorrect conclusions from a correlation.