Central Analytics, Grade 8 — SQL + data modeling + case, offer

Background

I had 14 months of experience as a Data Scientist at a Big Tech / Manufacturing company when I applied. I hold a diploma in Data Analytics from an AA-rated institute and applied for a Business Analyst role at Flipkart's Central Analytics team, Grade 8, based in Bengaluru.

How I Got the Interview

I got the interview through a referral from a friend. I think my diploma and solid projects — where I solved real problems using data — helped me get shortlisted. There was no back-and-forth with the recruiter before the process started.

I applied on 22 April 2024 and heard back on 4 May — a 12-day turnaround. I did about a week of preparation before each round.

Timeline

Rounds

Round 1 — Data Handling

Format: Video | Duration: ~1 hr 10 min | Interviewer: Senior Business Analyst

Q1 — Data Modeling

Imagine you're the DBA for a food delivery business like Zomato. What information would you capture? Which are dimensions vs. facts? Design the architecture so tables connect to each other.

I started by thinking about universal tables — orders, payments, employees — then added domain-specific ones like food menu and food ratings. For dimensions vs. facts, I reasoned that constant information belonged in one type and transactional in another. I gave each table a unique primary key (order_id, customer_id, address_id, rider_id, cafe_id), added relevant attribute columns, and made the orders table the base/fact table linking everything.

Self-note: I classified customer/employee info as facts and orders/payments as dimensions — conventionally it's the reverse. Orders/payments are facts; customer/employee are dimensions. Worth flagging for future prep.

Query Building

• Q2: Top 10 cities in descending order of total orders placed
• Q3: Customer with the highest returns of biryani in Bangalore
• Q4 (mid-level): Most stressed delivery partner in Bangalore, where stress = total orders delivered ÷ total working hours

I solved Q2 and Q3 with GROUP BY, ORDER BY, RANK/LIMIT, and aggregations. For Q4, I took a small hint to understand the problem, then solved it with window functions.

Q5 — Hard SQL Problem

Correlation between payment status, type of order, and food item — using SQL only.

Where I got stuck: This was the blocker. I tried a simple query, then a CTE with the CORR function, but couldn't structure it properly. The interviewer eventually shared the solution — I struggled to pick the right functions and organise the query logic. They were supportive throughout.

Q6 — Excel

• — VLOOKUP across two sheets
• — Build a pivot table and pull insights from it

I solved both easily — I'm a regular Excel user.

Verdict: Cleared

---

Round 2 — Problem Solving

Format: Video | Duration: ~1 hr | Interviewer: Lead Business Analyst

Q1 — Guesstimate: Toll Station Revenue

Estimate the revenue a toll station makes on the newly opened Bangalore–Chennai expressway.

I structured this top-down: split the 24-hour day into three 8-hour buckets by traffic volume; assumed ~2 minutes per vehicle to pass; estimated vehicles processed per bucket; split traffic into private vs. commercial and applied the relevant toll rate to each across the three time zones; then summed it all up.

Q2 — Optimization: Banana–Camel Problem

A plantation owner must transport 3,000 bananas to a market 1,000 km away using one camel. The camel carries max 1,000 bananas at a time, eats 1 banana per km, can make multiple trips, and can stash bananas at intermediate points. Find the maximum number of bananas deliverable.

Where I got stuck: I initially didn't recognise this as an optimization problem and started solving it manually. The interviewer flagged the constraints, after which I worked through it in an Excel table — testing different load/trip splits (e.g., 1,000 × 3 trips; 1,500/1,000/500) — and converged on 560. The actual answer is 533. The interviewer was impressed with the structured approach despite the off answer.

Q3 — Chess Board Permutation & Combination

A problem was asked here; the exact details aren't in my notes.

Q4 — Probability

Where I got stuck: I struggled greatly with this section — I didn't have the fundamentals solid enough to approach these confidently.

Verdict: Cleared

---

Round 3 — Hiring Manager

Format: Video | Duration: ~1 hr | Interviewer: Senior Manager, Analytics

Q1 — Resume & Project Deep-Dive

I walked through each project briefly, then did a deep-dive into my classification project: a class-imbalanced model predicting customer churn, used to prevent churn via marketing tools and offers. The interviewer probed the use case of each input variable — weather, natural calamities, distance from home, etc. I walked through every variable in full and felt confident here.

Q2 — Stats & ML Fundamentals

• — Confusion matrix: what does a true positive mean in the context of my churn model?
• — Formulas for precision and accuracy

I handled these confidently and on my toes.

Q3 — Business Case: Platform Pricing

As the analytics leader at Flipkart, the CEO asks you to determine — using data — which platform is better priced: Flipkart or Amazon.

My first approach: pull every item listed on both platforms, flag each as expensive, equal, or cheaper, and compute the share of items priced higher on Flipkart. I felt in control. Then the interviewer added a revenue angle — and I went blank. Completely. The interviewer was supportive and hinted toward "average," which eventually unlocked the weighted-average solution. The interview ended there.

Where I got stuck: When the conversation shifted from pure price comparison to a revenue perspective, I froze — fully speechless until the hint. The harder I gripped, the emptier my mind went.

Verdict: Cleared

Final Outcome

I was offered the role on 2 July 2024 — 71 days after applying. The package included a 30+ LPA CTC, ESOPs, and a joining bonus. HR reached out within 3–4 days after each round, with roughly a week between consecutive interviews, and there was a brief package discussion with some light negotiation.

What I'd Do Differently

I'd spend significantly more time strengthening my fundamentals in statistics, probability, and permutations & combinations — I felt underprepared for the probability-heavy questions in Round 2, and I'd practise more complex SQL problems beforehand. In hindsight, the bar was somewhere between medium and hard, harder than I'd initially prepared for, especially on the math and probability side.

Tips for Future Candidates

1. Build strong SQL fundamentals first. Practise a wide variety of problems on platforms like LeetCode, especially medium-to-hard level. Understand how real-world data systems operate and how to query them efficiently.
2. Master logical and analytical puzzles with a structured approach. The banana–camel problem taught me to recognise the shape of the problem (optimization, greedy, dynamic programming) before diving into the math. Test assumptions in Excel or on paper — don't go manual and unstructured.
3. Strengthen your math foundations. Statistics, probability, and permutations & combinations show up, and shallow knowledge crumbles under probing. Invest time here early.
4. Communication and soft skills matter as much as technical chops. Be clear, warm, and collaborative. Demonstrate a learning mindset and openness to feedback. When you get stuck, think out loud and show your reasoning — that counts.