Interview experience
EXL Interview Questions and Answers (2026)
Overview
Section titled “Overview”EXL runs a single online assessment into a role-dependent split: engineering candidates get a DSA/CS-fundamentals loop, while analytics/BPM candidates (EXL’s larger core business) get SQL, Excel, statistics, and case-style guesstimates.
EXL interview process at a glance
Section titled “EXL interview process at a glance”| Round | Duration | What they test |
|---|---|---|
| Online Assessment | 45-60 min | ~40 MCQs on aptitude, reasoning, verbal ability (+ CS fundamentals or SQL/Python/stats depending on role) |
| Technical Interview 1 | 20-30 min | Programming basics/DSA (engineering) or SQL/Excel/stats (analytics) |
| Technical/Case Interview 2 | 30-45 min | Project walkthrough or live demo, guesstimates, case studies |
| Managerial round (some drives) | 20-30 min | Team fit, business-line understanding, role alignment |
| HR Round | 15-20 min | Fit, motivation, why EXL/why analytics |
Online Assessment
Section titled “Online Assessment”A single timed test (~40 MCQs, 45-60 minutes) covering quantitative aptitude, logical reasoning, and verbal ability for all applicants, with the technical component branching by role - CS fundamentals for engineering roles, SQL/Python/statistics for analytics roles.
Common questions
- Quantitative aptitude - percentages, ratios, time-speed-distance
- Logical reasoning and data sufficiency
- SQL/Python MCQs or basic statistics questions (analytics-track applicants)
- Basic CS fundamentals MCQs - OS, DBMS, networks (engineering-track applicants)
Technical Interview 1
Section titled “Technical Interview 1”Branches hard by track. Engineering candidates get language basics and simple DSA; analytics candidates get SQL, Excel, and statistics fundamentals.
Common questions
- Basic DSA - arrays, strings, sorting, searching (engineering track)
- Write a SQL query using joins and GROUP BY (analytics track)
- Explain a statistical concept you’ve used in a project - correlation, hypothesis testing, distributions (analytics track)
- Core language fundamentals in whichever language your resume lists
Technical/Case Interview 2
Section titled “Technical/Case Interview 2”A project walkthrough or live demo, paired with guesstimates and case-style reasoning - most pronounced for analytics/BPM roles, where this round often decides the offer.
Common questions
- Walk us through a project you built, including a live demo if asked
- Estimate the number of X in a given city or market (guesstimate)
- A case-style question built around a past data project: what formulas or approach you used and why
- Explain a business-process or data-quality problem you solved and how you validated the fix
Round-by-round breakdowns are on the EXL interview experience page.
Managerial round (some drives)
Section titled “Managerial round (some drives)”Where run as a separate stage, this checks fit for the specific team and business line rather than testing new technical material - understanding of the role, past project experience, and expectation alignment.
Common questions
- What do you understand about the team/business line you’re being considered for?
- Walk me through a project outcome you’re proud of and your specific role in it
- Are your expectations on role scope and location aligned with what we’re offering?
HR Round
Section titled “HR Round”The closing 15-20 minute conversation on fit, motivation, and compensation. Interviewers explicitly probe for a genuine “why EXL, why analytics/BPM” answer rather than a generic one.
Common questions
- Tell me about yourself
- Why EXL, and why the analytics/BPM industry specifically?
- Compensation expectations and notice period
- Are you comfortable with the process/BPM nature of some roles versus pure product engineering?
Sample answer frameworks for each of these are on the EXL HR interview questions page.
Common technical interview questions and answers
Section titled “Common technical interview questions and answers”Q: Write a SQL query using a join and GROUP BY to find total claim amount per policy type.
A typical answer is SELECT p.policy_type, COUNT(c.claim_id) AS claim_count, SUM(c.amount) AS total_amount FROM policies p LEFT JOIN claims c ON c.policy_id = p.policy_id GROUP BY p.policy_type ORDER BY total_amount DESC; LEFT JOIN matters here because policy types with zero claims still need a row - an INNER JOIN would silently drop them. Note that COUNT of a column ignores NULLs, so a policy type with no claims correctly shows zero, whereas COUNT(*) would wrongly count the NULL-padded row as one. Every non-aggregated column in the SELECT list must appear in GROUP BY, which PostgreSQL enforces strictly and MySQL enforces when ONLY_FULL_GROUP_BY is on.
Q: What is the difference between WHERE and HAVING?
WHERE filters individual rows before grouping happens, so it cannot reference an aggregate function. HAVING filters the groups that GROUP BY produced, so aggregates are allowed there. In a claims query you would use WHERE to keep only claims filed this year, and HAVING to keep only policy types whose total claim amount exceeds a threshold. Putting a plain row condition in HAVING still returns the right answer but does more work, because rows are grouped first and then thrown away; putting an aggregate in WHERE is simply invalid. The logical order of evaluation is FROM, WHERE, GROUP BY, HAVING, SELECT, ORDER BY, LIMIT - which also explains why a SELECT alias cannot be used in WHERE but can be used in ORDER BY.
Q: What does a correlation coefficient of 0.8 actually tell you?
Pearson’s r of 0.8 means a strong positive linear relationship: as one variable rises the other tends to rise, and about 64 percent of the variance in one is explained by the other, since r squared is 0.64. It says nothing about causation - a lurking third variable or reverse causation produces the same number. It also captures only linear structure: a perfect parabola can have r near zero, which is why you plot the data before trusting the coefficient. Pearson’s r is sensitive to outliers, so one extreme claim value can move it substantially; Spearman rank correlation is the robust alternative when the relationship is monotonic but not linear.
Q: What is a p-value, and what does a significance level of 0.05 mean?
A p-value is the probability of observing data at least as extreme as what you got, assuming the null hypothesis is true. If it falls below the chosen significance level - conventionally 0.05 - you reject the null and call the result statistically significant. It is not the probability that the null hypothesis is true, and it is not the probability that your finding is a fluke; that is the most common mistake candidates make. A 0.05 threshold means accepting a 5 percent chance of a Type I error, rejecting a null that was actually true, and testing many hypotheses at once inflates that risk unless you correct for it. Statistical significance is also not practical significance: a huge sample can make a commercially meaningless difference significant.
Q: What is the difference between VLOOKUP and INDEX-MATCH in Excel?
VLOOKUP searches for a value in the leftmost column of a range and returns a value from a column a fixed number of positions to the right, so it cannot look leftward and it breaks when someone inserts a column, because the column index is hard-coded. INDEX-MATCH splits the job: MATCH finds the row position of the lookup value in one column, and INDEX returns the value at that position from any other column, in either direction. Because both arguments are references to real columns rather than an offset number, INDEX-MATCH survives column insertion and is generally faster on large sheets, since it scans two columns rather than the whole table. Always pass 0 or FALSE as the final argument for an exact match; in newer Excel versions XLOOKUP replaces both.
Q: How would you approach a guesstimate such as the number of two-wheelers in Bengaluru?
Structure beats precision - the interviewer is checking whether you decompose cleanly and state assumptions out loud. Start from a population base of roughly 13 million, convert to households using an average size of about 4, giving around 3.25 million households. Segment by income - say 20 percent low, 60 percent middle, 20 percent high - and assign two-wheeler ownership rates such as 0.3, 1.2 and 0.8 per household, since high-income households often substitute cars. That gives roughly 0.2 plus 2.3 plus 0.5 million, about 3 million, and you then add a commercial-fleet allowance for delivery riders. Close by sanity-checking against a known reference point and naming the assumption your estimate is most sensitive to.
Q: What is the time complexity of binary search, and when can you not use it?
Binary search runs in O(log n) time and O(1) space iteratively, because each comparison halves the remaining search space - a million sorted elements need at most about 20 comparisons. It requires the data to be sorted and randomly accessible by index, so it works on an array but not on a linked list, where reaching the midpoint is already O(n). If the array is unsorted and you only need one lookup, sorting first at O(n log n) is worse than a single linear scan at O(n); sorting pays off only when many searches follow. Two classic bugs are computing the midpoint as (low + high) / 2, which can overflow - use low + (high - low) / 2 - and getting the loop-termination condition wrong so it never converges.
Q: What is the Central Limit Theorem and why does it matter in analytics work?
The Central Limit Theorem states that the sampling distribution of the sample mean approaches a normal distribution as sample size grows, regardless of the shape of the underlying population, provided that population has finite variance. Its standard deviation, the standard error, is the population standard deviation divided by the square root of n - so quadrupling the sample size only halves the error. This is why confidence intervals and t-tests work on skewed real-world data such as claim amounts or call-handling times, which are nowhere near normal themselves. A rule of thumb is that n of about 30 suffices for mildly skewed data, though heavily skewed distributions need considerably more.
Frequently asked questions about EXL interviews
Section titled “Frequently asked questions about EXL interviews”What is the EXL interview process for freshers?
EXL typically runs 3-4 rounds: 1. Online Assessment (45-60 minutes) - around 40 MCQs on quantitative aptitude, logical reasoning, verbal ability, and sometimes CS fundamentals or SQL/Python/statistics depending on the role. 2. A first Technical Interview (20-30 minutes) - basics of your programming language plus simple DSA for engineering roles, or SQL/Excel/stats for analytics roles. 3. A second Technical/Case Interview (30-45 minutes) - project walkthrough or live demo, along with guesstimates, case studies, and quant-reasoning questions, especially for analytics/BPM roles. 4. HR Round (15-20 minutes) - fit, motivation, and culture questions. Some drives combine the two technical rounds into one.
What questions are asked in EXL interviews?
For engineering roles, expect basic DSA (arrays, strings, sorting, searching, stacks, queues) and CS fundamentals (OS, DBMS, computer networks, OOPs). For analyst/BPM-analytics roles, EXL leans heavily on SQL, Excel formulas, statistics, and past-project or guesstimate/case-study questions, plus a direct ‘why EXL, why analytics/BPM’ question that interviewers use to filter out candidates who seem unsure about the industry.
How many rounds are there in the EXL interview?
Most candidates go through 3-4 rounds: an online aptitude/technical assessment, one or two technical or case interviews, a managerial round for team/role fit, and a final HR round. Attrition is steep at the technical/case round - candidate reports describe large drop-offs between the written test and the interview stage.
How should I prepare for EXL interviews?
Practice basic-to-medium DSA and brush up OS/DBMS/networks fundamentals if you’re applying for a tech role. For analytics/BPM roles, get comfortable with SQL joins/aggregations, Excel formulas, and structured case/guesstimate thinking (market sizing, profitability-style questions), and prepare a clear, specific answer for ‘why EXL’ and ‘why analytics/BPM’ rather than a generic response - interviewers explicitly probe for genuine interest in the industry.
Is EXL’s process different for engineering roles versus analyst/BPM-analytics roles?
Yes. EXL hires for both software engineering roles (product/platform teams) and analytics/BPM roles (its larger, core business - insurance and healthcare analytics, business process management). Engineering roles run a more DSA/CS-fundamentals-heavy loop closer to a typical product company. Analyst/BPM roles lean on SQL, Excel, statistics, guesstimates, and case-style discussion of past data projects, with less emphasis on classic coding-interview DSA.
What does EXL’s managerial round check for?
The managerial round (where run as a separate stage) focuses on overall fit for the specific team, understanding of the business line you’re joining, past project experience, and role-expectation alignment - it’s less about testing new material and more about confirming you and the team are a match before the closing HR/compensation conversation.