Uber Business Analyst Interview: SQL, Business Insights, and Strategy Output Full Assessment

Background

Let me share my situation first: 2 years of business analysis experience, previously at an O2O company as a business analyst, mainly doing user behavior analysis and business strategy output. Uber's Business Analyst position has always been my target — after all, Uber is a data benchmark in the ride-sharing and mobility space, with access to massive amounts of real business data. For an analyst, it's practically paradise. This year I finally got an interview opportunity. Although I didn't pass in the end, the entire interview process taught me a lot, and I'm sharing it with everyone.

The interview process was three rounds: Round 1 focused on SQL + Python, Round 2 on business case + strategy, and Round 3 on data-driven decision making. Each round lasted about 1-1.5 hours — quite intense. Let me break it down in detail.

Interview Process Recap

Round 1: SQL + Python

Round 1 was a technical interview with a senior analyst from the data analytics team. No small talk — straight into SQL.

First SQL question: Given an orders table (order_id, user_id, order_time, amount, status), find the cumulative spending and order count for each user in the last 30 days, keeping only the top 100 users by spending. This wasn't too hard — I quickly wrote it using window functions. But the interviewer immediately added a condition: If users have refunded orders, the refund amount should be deducted from cumulative spending. How would you modify the query? I added a CASE WHEN condition on status to handle refunds.

The second SQL question was more challenging: Given a user behavior table (user_id, event_type, event_time, page_id), find the number of sessions per user (session definition: a gap of more than 30 minutes between two events counts as a new session), and the average duration per session. This tested session segmentation logic. I used the LAG function to calculate the time difference between adjacent events, then used cumulative sum to label session numbers, and finally counted sessions and durations. The interviewer said the logic was correct but asked me to optimize the SQL performance. I suggested using partitioned tables and indexes to speed up queries.

For the Python section, he asked several data processing questions: How do you handle missing values in pandas? How do you create pivot tables? How do you merge two DataFrames? These were basic operations that I answered smoothly. But his final question was more advanced: If the data is too large for pandas to handle, what would you do? I said we could consider using Dask or PySpark for big data processing, or use chunked reading. He nodded and didn't push further.

Overall, Round 1 went okay. I did well on SQL and was average on Python. The interviewer said my SQL foundation was solid but suggested I learn more about big data processing tools.

Round 2: Business Case + Strategy

Round 2 was the business interview with the data lead for the business line. This was the hardest round because it wasn't testing technical skills but business understanding and strategy output capability.

First case: Uber's ride repeat rate has been declining for 3 consecutive months. How would you analyze the reasons? I broke it down from several dimensions: first check if it's an overall decline or specific to certain cities/categories (dimensional breakdown), then look at user segmentation (new vs. existing users, high-frequency vs. low-frequency), then external factors (competitor promotions, seasonality), and finally internal factors (ride experience, driver quality, pricing changes). The interviewer appreciated my breakdown approach but followed up: If you find that existing user repeat rates are declining but new user repeat rates are unchanged, how would you analyze further? I said we could look at changes in existing users' ride frequency, average fare, category preferences, and whether they're churning to competitors.

The second case was more open-ended: Uber wants to enter a new city. How would you evaluate whether it's worth entering? I analyzed from five dimensions: market size, competitive landscape, supply side (number and quality of drivers), demand side (user density and spending capacity), and operational costs. The interviewer asked: What if data is insufficient? I said we could use proxy indicators and analogy methods — for example, using city GDP and population density to estimate market size, and using data from similar cities for comparison. He seemed satisfied with this answer.

For the strategy output section, he asked: If your analysis shows that a category's GMV growth is mainly driven by subsidies, and removing subsidies would cause GMV to drop significantly, what would you recommend? I said first look at the ROI of subsidies — if ROI is positive, subsidies are effective and can continue but need optimization; if ROI is negative, gradually reduce subsidies while improving natural conversion rates. He followed up: How do you improve natural conversion rates? I answered from three angles: supply optimization (improving driver quality), demand matching (optimizing recommendation algorithms), and user experience (reducing wait times).

Round 3: Data-Driven Decision Making

Round 3 was with a Director, focusing more on data-driven decision-making mindset.

He started with a macro question: What do you think is the greatest value of a business analyst? I answered "turning data into decisions" — analysts aren't just making reports and pulling numbers, but more importantly discovering business insights from data and driving decision implementation. He followed up: Have you ever encountered a situation where data analysis and business decisions were inconsistent? How did you handle it? I shared an experience where my analysis showed a campaign had negative ROI, but the business team insisted on running it because they believed it had brand value. My approach was to quantify brand value (using brand search index and NPS), then re-evaluate together with the business team, ultimately reaching consensus.

He also asked a very practical question: If your analysis conclusion contradicts your boss's intuition, how do you handle it? I thought for a moment and said: first ensure my analysis is correct, then use data and logic to persuade, not authority or emotion. If the boss still insists, suggest running a small-scale experiment to validate. He seemed to appreciate this answer.

Finally, he asked about my understanding of Uber's business: What is Uber's core competitive moat? I answered from three angles: network effects (two-sided network of drivers and riders), data accumulation (massive transaction and behavior data), and operational capability (efficient dispatch system). He added one point: Uber's moat also lies in organizational capability and company culture — something I hadn't deeply considered before.

Key Questions Summary

Round 1:

1. Find top 100 users by spending in the last 30 days (SQL)

2. How to deduct refund amounts from cumulative spending in SQL?

3. Calculate user session count and average session duration (SQL)

4. How to handle missing values in pandas? How to create pivot tables?

5. What to do when data is too large for pandas?

Round 2:

1. Ride repeat rate declining for 3 months — how to analyze reasons?

2. Existing user repeat rate declining but new users unchanged — how to analyze further?

3. How to evaluate whether a new city is worth entering?

4. How to analyze when data is insufficient?

5. GMV growth driven by subsidies — what to recommend?

Round 3:

1. What is the greatest value of a business analyst?

2. How to handle inconsistency between data analysis and business decisions?

3. Analysis conclusion contradicts boss's intuition — what to do?

4. What is Uber's core competitive moat?

5. What are your biggest strengths and weaknesses as an analyst?

Key Takeaways

1. SQL is fundamental — it must be rock solid. Uber's SQL interview isn't simple CRUD — it tests window functions, session segmentation, multi-table joins, and other advanced operations. I recommend practicing LeetCode SQL problems and specialized SQL exercises.

2. Business cases need a breakdown framework. Don't jump to answers — break down the problem first. Dimensional breakdown, user segmentation, internal/external analysis — these frameworks should be second nature.

3. Strategy output needs to be a closed loop. Analysis is just the first step. What matters more is the complete loop from analysis to strategy to implementation. Interviewers will push until you give actionable recommendations.

4. Data-driven doesn't mean data-only. Round 3 tests data-driven decision-making mindset, not data dogmatism. Understand the business, understand people — data is just a tool to aid decisions.

5. Deeply understand the company's business. Before the interview, thoroughly research the target company's business model, core metrics, and competitive landscape — these will come up in Rounds 2 and 3.

FAQ

Q: Does the Uber Business Analyst interview test Python?

A: Round 1 does, but mainly pandas basics — no algorithm problems. SQL is the focus.

Q: Can I pass without tech industry experience?

A: It's difficult but not impossible. I recommend preparing at least 1-2 data analysis projects where you can clearly explain your analytical approach and conclusions.

Q: What SQL environment is used for the interview?

A: Online SQL editor — the specific platform varies. I recommend being familiar with various SQL syntaxes.

Q: Do business cases have standard answers?

A: No — they test analytical framework and logic. Process matters more than conclusion.

Q: How long until interview results come out?

A: Usually 3-7 days after each round. The entire process takes 2-4 weeks.