Interview experience
ZS Associates Interview Questions and Answers (2026)
Real ZS Associates interview questions - candidate interview experiences and HR round prep, in one place.
Overview
Section titled “Overview”ZS Associates hires on structured problem-solving, not DSA - a quant-heavy aptitude test, a 30-35 minute case interview with guesstimates, and a track-dependent analytics/technical round set it apart from a typical IT-services loop.
ZS Associates interview process at a glance
Section titled “ZS Associates interview process at a glance”| Round | Duration | What they test |
|---|---|---|
| Online aptitude assessment | 50-75 min (varies) | Quant, logical reasoning, data interpretation, verbal |
| Behavioural / video assessment (some drives) | 15-30 min | Situational judgement, communication |
| Case interview | 30-35 min prep + discussion | Structured problem solving, guesstimates, assumptions |
| Technical / analytics interview (track dependent) | 30-45 min | SQL, Python or Excel, statistics, data interpretation |
| Fit / HR interview | 20-30 min | Why ZS, resume deep-dive, behavioural stories |
Common HR questions asked at ZS Associates
Section titled “Common HR questions asked at ZS Associates”- Tell me about yourself?
- Why ZS Associates, and why analytics or consulting?
- Tell me about a time you used data to change someone’s mind?
- Walk me through a project on your resume and the assumptions you had to make?
Sample answer frameworks for each of these are on the ZS Associates HR interview questions page.
Common technical interview questions and answers
Section titled “Common technical interview questions and answers”Q: How would you structure a guesstimate, for example the annual market size for a diabetes drug in India?
Work top-down from population and state every assumption out loud. Start with roughly 1.4 billion people in India, take the adult share at about 65 percent, giving around 910 million adults. Apply a diabetes prevalence of roughly 10 percent for about 90 million diabetics. Then narrow with funnel steps: perhaps 50 percent are diagnosed, of whom around 60 percent are on regular medication, and your drug’s therapeutic class might address 30 percent of those - about 8 million patients. Multiply by annual therapy cost, say 6,000 rupees, for roughly 48 billion rupees of addressable market, then apply a realistic market share. Interviewers care far more about the clean structure and the sanity check at the end than about your exact prevalence figure.
Q: Write a SQL query to find the Nth highest salary.
The most readable version uses DENSE_RANK: SELECT DISTINCT salary FROM (SELECT salary, DENSE_RANK() OVER (ORDER BY salary DESC) AS rnk FROM Employee) t WHERE rnk = N. DENSE_RANK is the right function because ties should share a rank without skipping the next value - ROW_NUMBER would give two employees on the same salary different ranks, and RANK would skip N entirely after a tie. Without window functions you can use LIMIT with OFFSET on a DISTINCT salary list, or a correlated subquery counting how many distinct salaries are strictly greater. Always mention what should happen when fewer than N distinct salaries exist: the window version returns no rows rather than an error.
Q: What is the difference between RANK, DENSE_RANK, and ROW_NUMBER?
All three assign a number over an ordered partition, but they differ on ties. ROW_NUMBER always gives distinct sequential numbers, so tied rows get arbitrary but different values - use it for deduplication, such as keeping one row per customer. RANK gives tied rows the same number and then skips, so two rows at rank 1 are followed by rank 3. DENSE_RANK gives tied rows the same number and does not skip, so the sequence continues at 2. For an analytics question like the top three products per region, DENSE_RANK is usually what a business stakeholder means, because it includes every product genuinely tied for third.
Q: What does a p-value actually mean, and what does it not mean?
A p-value is the probability of observing data at least as extreme as yours if the null hypothesis were true. It is not the probability that the null hypothesis is true, and it is not the probability your result happened by chance. A p-value of 0.03 against a 0.05 threshold means you reject the null, but it says nothing about effect size - with a large enough sample, a commercially meaningless 0.1 percent lift becomes statistically significant. That is why ZS-style analytics answers should pair significance with a confidence interval on the effect and a judgement about whether the magnitude is worth acting on.
Q: What is the difference between a Type I and a Type II error, and how do you control them?
A Type I error is a false positive - rejecting a true null hypothesis - and its rate is alpha, the significance level you choose, typically 0.05. A Type II error is a false negative, failing to reject a false null; its rate is beta, and power is 1 minus beta, conventionally targeted at 0.80. The two trade off: tightening alpha to 0.01 reduces false positives but raises the chance of missing a real effect. The only way to reduce both at once is a larger sample. In a commercial context the choice depends on cost - launching an ineffective campaign versus missing an effective one - so state which error is more expensive before picking a threshold.
Q: A pharmaceutical client’s brand has lost 15 percent of sales in one quarter. How would you structure the diagnosis?
First clarify the number: 15 percent versus what baseline, in units or in revenue, and across which geographies and channels - a revenue drop with flat units is a price or mix problem, not a demand problem. Then decompose sales into volume times price, and volume into number of prescribers times prescriptions per prescriber times units per prescription. Segment each factor by region, specialty, and payer to see whether the loss is broad or concentrated in a few accounts, which points at very different causes. External hypotheses come next: a competitor launch, a generic entry, a formulary or reimbursement change, or a supply disruption. Close with a recommendation tied to whichever branch the data actually supports, plus what you would measure to confirm it.
Q: How would you explain the difference between correlation and causation to a client?
Correlation means two variables move together; causation means changing one produces a change in the other. They diverge for three reasons: a confounder driving both, reverse causality, or coincidence in a small sample. The classic commercial trap is finding that high-detailing physicians write more prescriptions and concluding that detailing works - when the sales force was targeting high-volume prescribers in the first place, so prescription volume caused the detailing. Establishing causation needs a randomised controlled test, or a quasi-experimental method such as difference-in-differences with a matched control group when randomisation is not possible.
Q: How would you handle missing values and outliers in a dataset before analysis?
First diagnose why the data is missing, because the mechanism dictates the fix: missing completely at random can be dropped with little bias, missing at random can be imputed conditionally on other variables, and missing not at random - where the reason relates to the missing value itself, such as high earners skipping an income field - biases any simple imputation. Options range from listwise deletion to mean or median imputation, to model-based imputation like KNN or MICE, and adding an explicit missing-indicator column often preserves signal. For outliers, distinguish data-entry errors, which should be corrected or removed, from genuine extreme values, which usually should not - a handful of very high prescribers are often the most commercially important accounts. Winsorising, or a log transform, keeps them in the analysis while limiting their leverage.
Frequently asked questions about ZS Associates interviews
Section titled “Frequently asked questions about ZS Associates interviews”What is the ZS Associates interview process like?
ZS Associates fresher hiring typically runs in four stages: 1. Online aptitude assessment - quantitative aptitude, logical reasoning, data interpretation and verbal ability, tightly timed. 2. Behavioural / video assessment - recorded situational responses, on many but not all drives. 3. Case interview - a business case you analyse on your own for roughly 30-35 minutes, then discuss with the panel, often with guesstimates and puzzles. 4. Fit / HR interview - why ZS, resume deep-dive and behavioural questions. Analytics and technology tracks add SQL, Python or Excel questions.
What questions are asked in ZS Associates interviews?
ZS interviews centre on structured problem solving rather than heavy DSA: business cases in pharma, healthcare and commercial analytics, guesstimates (market sizing), puzzles, data interpretation, statistics fundamentals, and SQL for the analytics and technology tracks (joins, subqueries, nth-highest salary, window functions). Behavioural questions such as ‘why ZS’ and resume-based STAR stories appear in almost every round.
How many rounds are there in the ZS Associates interview?
Most candidate reports describe three to five rounds: an online aptitude test, an optional behavioural or video assessment, one or two case / technical interviews, and a fit interview with a manager or recruiter. On-campus drives often compress the interviews into a single day. Round counts differ between the Decision Analytics and Business Technology Solutions tracks, so check that cycle’s placement email.
What is the ZS Associates case interview like?
The case interview is the round that distinguishes ZS. You are usually given a business problem, often set in pharmaceuticals, healthcare or commercial analytics, and about 30-35 minutes to work through it alone before presenting your structure, assumptions and recommendation to the panel. Interviewers reward candidates who decide for themselves what to analyse next, state assumptions explicitly, and back conclusions with numbers.