How to Crack the Data Analyst Interview (End to End)

The data analyst interview in India is not one conversation - it is a funnel of five to six stages, and each one is engineered to fail you on a different skill. Freshers lose because they prepare for "an interview" as a single event, then get knocked out at the SQL test or the case round they never rehearsed. This guide walks the full funnel as it actually runs in 2026 - at product companies like Swiggy, Zomato, Flipkart, PhonePe and CRED, and at the large service firms (TCS, Accenture, Wipro, Cognizant, Capgemini) - and tells you exactly what each stage tests and how to clear it.

A typical India funnel looks like this:

Resume screen → online SQL/aptitude test → technical round (SQL + Power BI/Excel) → case study or take-home → behavioural → HR.

Product companies lean harder on the case study; service companies lean harder on the aptitude test and breadth of tools. Either way, treat each stage as a separate skill to be drilled, not a single hurdle.

Stage 1 - Resume screening

Before a human reads your resume, a recruiter scans it for 6-8 seconds and an ATS keyword-matches it against the job description. You pass by being concrete and keyword-aligned.

Prepare: Lead every project bullet with an outcome and a number. Mirror the exact tools named in the JD (SQL, Power BI, Excel, Python, data cleaning) so the ATS matches you. Add a live portfolio link a recruiter can click.

What works

• Project lines with numbers: "Built a Power BI dashboard on 2 lakh UPI transactions; surfaced a 22% fraud spike in tier-2 cities."
• The right keywords mapped to the JD: SQL, Power BI, Python, Excel, data cleaning, dashboards, DAX.
• A linked, live portfolio. A recruiter who can click into a working project trusts you instantly.

What fails

• "Responsible for data analysis tasks." Vague and forgettable.
• No links, no GitHub, no dashboard.
• A wall of certifications with zero projects behind them.

Stage 2 - The online SQL / aptitude test

Most companies now filter with an automated test before a human spends time on you - on HackerRank, HackerEarth, AMCAT, or an in-house platform. It usually mixes timed SQL problems with quantitative aptitude and logical reasoning, especially at service firms.

Prepare: For SQL, drill 30-40 timed problems on joins, GROUP BY/HAVING, subqueries and window functions until syntax is automatic - the clock, not the difficulty, is what fails people. For aptitude, revise percentages, ratios, averages, permutations and data interpretation; free question banks from any CAT/placement prep cover the exact pattern. Read every prompt twice; the trap is usually a missed condition, not hard logic.

Stage 3 - The technical round (SQL + Power BI/Excel)

This is the highest-leverage human stage. You'll share a screen and solve SQL live, then often defend a Power BI or Excel build. It eliminates the most candidates, so over-invest here.

SQL: the patterns that come up

Aggregation with filtering - top products by revenue per city:

SELECT city, product, SUM(amount) AS revenue
FROM   orders
WHERE  order_date >= '2026-01-01'
GROUP  BY city, product
HAVING SUM(amount) > 50000
ORDER  BY revenue DESC;

Joins - know INNER, LEFT, and the classic "find customers with no orders":

SELECT c.customer_id, c.name
FROM   customers c
LEFT   JOIN orders o ON o.customer_id = c.customer_id
WHERE  o.order_id IS NULL;

Window functions - rank stores within each city, and find the top earner per city:

SELECT city, store_id, revenue
FROM (
  SELECT city, store_id, revenue,
         RANK() OVER (PARTITION BY city ORDER BY revenue DESC) AS city_rank
  FROM   store_sales
) ranked
WHERE city_rank = 1;

Running totals and month-on-month - a frequent follow-up:

SELECT month,
       SUM(revenue) AS monthly_revenue,
       SUM(SUM(revenue)) OVER (ORDER BY month) AS running_total
FROM   sales
GROUP  BY month
ORDER  BY month;

Self-joins, date logic, and "find the second-highest salary" are also classics. The single most common trap is the difference between WHERE and HAVING (row filter vs. group filter), followed by confusing JOIN types. Know these cold.

How to pass: Think out loud. Interviewers care more about your reasoning than a perfect first query. Say "I'll group by city, then filter the groups with HAVING" before you type. If you blank, restate the question - it buys time and shows structure.

Power BI / Excel: be ready to write a measure

If you list Power BI, expect to explain a DAX measure and the difference between a calculated column and a measure. A common ask is a month-over-month growth measure:

Revenue MoM % =
VAR CurrentRevenue = SUM ( Sales[Amount] )
VAR PreviousRevenue =
    CALCULATE (
        SUM ( Sales[Amount] ),
        DATEADD ( 'Date'[Date], -1, MONTH )
    )
RETURN
    DIVIDE ( CurrentRevenue - PreviousRevenue, PreviousRevenue )

Be ready to explain why DIVIDE is safer than / (it returns blank instead of erroring on divide-by-zero), and why a measure recalculates per filter context while a calculated column is fixed at refresh. For Excel, expect VLOOKUP/XLOOKUP, INDEX-MATCH, SUMIFS, and pivot tables.

Stage 4 - The case study or take-home

Now they test whether you can think like an analyst, not just query like one. A typical product-company prompt: "Daily active users on our food-delivery app dropped 15% last week. How would you investigate?"

Structure your answer

1. Clarify: Which segment - new or returning users? Which cities? Is the 15% vs. last week or last year?
2. Hypothesize: App bug or release? A competitor's discount? Festival season ended? A payment-gateway failure?
3. Find the data: Which tables and metrics would confirm or kill each hypothesis.
4. Quantify and recommend: "I'd segment the drop by city and platform first, because that isolates a tech issue from a market issue."

The structure is the answer. Random guessing fails; a clear framework passes even if you don't reach a final number.

Guesstimates

Service firms and analytics consultancies love a guesstimate, e.g. "How many Swiggy orders are placed in Bengaluru per day?" Don't guess a number - build it: state Bengaluru's population (~1.4 crore), estimate the share who order online, average orders per active user per week, then divide to a daily figure and sanity-check it. They grade the path, not the answer.

For take-homes

If they hand you a CSV (say, cab-trip data), don't just chart it. Clean it, state your assumptions, show one or two real insights, and end with a recommendation. A clear README beats a fancy notebook.

Stage 5 - The behavioural interview

This stage checks whether you can communicate and fit the team. Indian hiring managers consistently flag communication as the gap that separates technically-okay freshers from hireable ones.

Present your portfolio in 60-90 seconds (STAR + business question)

The single biggest win here is a crisp project pitch. Don't narrate the tools - frame the business question, then walk Situation → Task → Action → Result, ending on a number:

"Kirana stores were losing stock to expiry but had no visibility (Situation). I was asked to find which categories drove the waste (Task). I cleaned 50k rows of POS data in SQL, built a Power BI dashboard with a DAX measure for waste %, and segmented by category and store (Action). It showed dairy and bakery drove 60% of waste, so reorder cycles could be tightened - projected ~12% lower spoilage (Result)."

That's about 75 seconds. Have two stories ready: one technical win, one about handling difficulty (a messy dataset, a tight deadline, conflicting stakeholder asks).

Questions to expect

• "Walk me through your most recent project."
• "A stakeholder disagrees with your analysis. What do you do?"
• "How do you prioritise when three people want dashboards by Friday?"
• "Why data analytics?"

Red flags that get candidates rejected

• Reading SQL syntax off the question but unable to explain why a join behaves that way.
• Jumping straight to a number in a case without clarifying the question.
• Calling everything "my project" but unable to explain a single decision in it (a sign it was copied).
• Trashing a past manager or team.
• Zero questions for the interviewer - it reads as disinterest.

Smart questions to ask the interviewer

• "What does the data stack look like - warehouse, BI tool, how queries get to production?"
• "What would success in this role look like in the first 90 days?"
• "Who consumes these dashboards, and how are analyst priorities decided?"

Stage 6 - The HR round

Mostly a fit-and-logistics check, but people still trip here. Prepare: a one-line "why this company," an honest reason for leaving (or for choosing analytics, as a fresher), a researched salary range for your city and band, and your notice period. Be cooperative on numbers but don't undersell - quote a range, not a single figure.

Putting it together: a 4-week plan

• Week 1: SQL drills daily - joins, aggregations, window functions; do them timed to prep for the online test.
• Week 2: Two end-to-end projects on Indian datasets (UPI, kirana, cabs), each with a README and one DAX measure.
• Week 3: Case + guesstimate practice - solve five business problems out loud using the framework above.
• Week 4: Mock behavioural and HR rounds; record your 90-second project pitch and rewatch it.

Why it is worth the effort

A strong, structured candidate in 2026 commonly lands ₹4-7 LPA as a fresher, with Bengaluru and Hyderabad paying above the national average, and job-switchers in the first three years routinely seeing 40-60% hikes (Scaler, upGrad). The funnel is the one gate between you and that band. Prepare for each stage as a separate skill, and you'll clear the whole pipeline.

FAQ

How to prepare for a data analyst interview?
Prepare per stage, not as one event. Drill SQL daily (joins, GROUP BY/HAVING, window functions), revise aptitude for the online test, practise case studies and guesstimates out loud, build two portfolio projects on Indian datasets, and rehearse a 90-second STAR pitch for each project.

What SQL is asked in data analyst interviews?
Joins (INNER/LEFT, and "rows with no match"), aggregation with GROUP BY and HAVING, subqueries, and window functions like RANK(), ROW_NUMBER() and running totals. The most-tested concept is WHERE vs. HAVING, plus the difference between join types. Classics include second-highest value and top-N per group.

How many rounds are there in a data analyst interview in India?
Usually five to six: resume screen, an online SQL/aptitude test, a technical round (SQL + Power BI/Excel), a case study or take-home, a behavioural round, and an HR round. Product companies weight the case study; service companies weight the aptitude test and tool breadth.

Do I need Python to get a data analyst job in India?
For most entry-level roles, strong SQL plus Power BI or Excel is enough to clear the funnel. Python (pandas) is a strong plus and is expected more often at product companies and for analytics-heavy roles, but it rarely gates a fresher offer on its own.

How long does the interview process take?
Typically one to three weeks end to end. Service firms can compress all rounds into a single day during campus or walk-in drives; product companies usually spread rounds across one to two weeks, with the take-home adding a few days.

What is the most common reason candidates get rejected?
Weak communication and unstructured thinking - being unable to explain why a query works, jumping to a case answer without clarifying, or not being able to defend a project's decisions. Strong reasoning out loud beats a silent correct answer.

Related: Power BI Portfolio Projects That Get You Hired · Build your skill score

Don't just read. Prove your skill on DevWithData.