EXL Service Data Scientist Interview: Questions & Prep (2026)

EXL Service is an analytics and digital solutions company known for its work in insurance, banking, healthcare, and retail analytics. With 398 open Data Scientist roles as of July 2026, EXL is one of the most active hirers in this space across India right now.

The interview process typically spans 3-4 rounds: a recruiter screening call, one or two technical rounds covering statistics, SQL, and machine learning, a case study or take-home assignment, and a final manager or HR discussion. Candidates report that EXL places strong emphasis on domain-applied analytics, so knowing how data science connects to business outcomes matters as much as coding skill.

Salaries at EXL for Data Scientists align with broader market bands. Based on knok jobradar data, Data Scientist roles in India range from 8-16 LPA at entry level (0-2 years), 18-30 LPA at mid-level (3-5 years), 30-48 LPA at senior level (6-9 years), and 45-70+ LPA at Lead or Principal level.

Technical and Statistics

1. Explain the difference between bagging and boosting. Which would you use when, and why?
2. How do you handle class imbalance in a classification problem? Walk through at least two approaches.
3. A model has high training accuracy but poor test accuracy. What do you diagnose and how do you fix it?
4. Write a SQL query to find the top 3 customers by revenue per region without using a subquery.
5. How would you evaluate a regression model beyond R-squared? What metrics matter for a business report?

Machine Learning and Deployment

1. Describe a model you built end-to-end. What data pipeline did you set up, and how did you handle missing values?
2. How do you monitor a machine learning model after deployment? What signals tell you it is drifting?
3. EXL works heavily in insurance. How would you build a claims fraud detection model? Walk through your approach.

Domain and Business

1. A business stakeholder says your model predictions are wrong. How do you respond and what do you investigate first?
2. How would you explain a complex model result (like an XGBoost output) to a non-technical client?
3. Tell me about a time you found an insight in data that changed a business decision.
4. EXL often works with US-based clients in insurance and banking. How do you adapt your communication style for cross-functional or offshore collaboration?

Q: Describe a model you built end-to-end.

*Situation:* At my previous company, the marketing team was losing customers silently. Churn was being noticed only after it happened.

*Task:* I was asked to build a model to predict which customers were likely to churn in the coming month so the retention team could act early.

*Action:* I pulled historical transactional data from the data warehouse, handled missing values using median imputation for numerical columns and mode for categorical ones, and engineered features like 'days since last login' and 'average order drop over recent months.' I trained an XGBoost classifier, tuned it with cross-validation, and presented precision-recall curves to the product lead since churn recall mattered more than raw accuracy. I also built a simple weekly report in a dashboard so the retention team could act without needing to query the model directly.

*Result:* The model flagged at-risk customers with strong recall on the holdout set. The retention team ran a targeted campaign and reported a measurable improvement in retention within two months. EXL clients in insurance or banking face similar needs, so I am comfortable adapting this pattern.

---

Q: How do you handle class imbalance?

*Situation:* I was building a fraud detection model for a fintech client where genuine fraud cases made up a very small fraction of all transactions.

*Task:* A standard classifier was predicting 'not fraud' for almost everything and still showing high accuracy, which was useless for the business.

*Action:* I used three approaches in combination: SMOTE to oversample the minority class in training data, class-weight adjustment in the model so the classifier penalised missing a fraud case more heavily, and switched from accuracy to F1-score and AUC-PR as my evaluation metrics. I also consulted the business team on the cost of a false negative (missed fraud) versus a false positive (blocked legitimate transaction) to set the right threshold.

*Result:* The final model caught a much higher share of actual fraud cases with an acceptable false positive rate. The client approved it for production after a two-week shadow period.

---

Q: Tell me about a time you found an insight that changed a business decision.

*Situation:* A retail client assumed their discount campaigns were driving revenue growth based on raw sales numbers.

*Task:* I was asked to validate whether the promotions were actually profitable.

*Action:* I ran a cohort analysis comparing customers who received discounts versus those who did not, controlling for seasonality. I built a simple incremental lift analysis and found that a large portion of discounted purchases were from customers who would have bought anyway, meaning the discounts were eroding margin without driving new demand.

*Result:* The client paused the blanket discount program and moved to a targeted approach based on predicted purchase propensity. Turning a data finding into a cost-saving recommendation is exactly what EXL clients value.

STAR for experience questions: Structure every behavioral answer as Situation, Task, Action, Result. Keep Situation and Task brief (two to three sentences total) and spend most of your time on Action and Result. Interviewers at EXL want to see how you think, not just what happened.

For technical questions: State your approach first, then walk through your reasoning step by step. If you are unsure of a formula, say so and reason from first principles. Interviewers typically value structured thinking over a memorised answer.

For case study or domain questions: Start by clarifying the business problem. Then define success metrics before you touch any model. EXL is a client-facing company, so showing you can connect a model to a business outcome is as important as the technical approach itself.

For 'how would you explain this' questions: Pick an analogy the audience would recognise, avoid jargon, and end with the business implication. Saying 'this feature was the biggest driver of the prediction' is more useful to a client than citing a SHAP value.

EXL operates in domain-heavy industries like insurance, banking, and healthcare, so interviewers are looking for more than clean Python notebooks.

Applied thinking over theory. EXL wants candidates who can take a messy business problem, pick the right tool, and explain the result in plain language. Pure algorithmic knowledge without business context will not impress.

Communication skills. Because EXL works with external clients, your ability to explain a model to a non-technical stakeholder is tested explicitly. Expect at least one question where you have to simplify a concept.

SQL and data handling. Candidates report that SQL questions are common, even at senior levels. Know window functions, joins, and how to write performant queries on large tables.

Ownership mentality. EXL values candidates who take projects from raw data to business recommendation without waiting to be told each step. Reference examples where you drove a project end-to-end.

Domain familiarity. If you are applying to a team that works in insurance or banking, read up on how data science is used there: claims prediction, credit scoring, risk segmentation. You do not need to be a domain expert, but showing awareness helps.

Week 1: Technical foundations

Revise core statistics: probability, hypothesis testing, distributions, and A/B testing concepts. Practice SQL on platforms like HackerRank or LeetCode (medium difficulty). Review the math behind logistic regression, decision trees, and gradient boosting.

Week 2: Machine learning and case studies

Pick two or three projects from your own portfolio and practice narrating them in STAR format. Prepare to explain every modelling choice you made. Read up on EXL's publicly available case studies and annual reports to understand which industries they serve.

Week 3: Mock interviews and domain prep

Do at least two full mock interviews where you talk through a problem out loud. Practice explaining a model result to a non-technical audience. If your target role is in insurance or banking, review publicly available material on fraud detection, claims modelling, or credit risk.

Before the interview

Prepare three to four strong project examples that cover different scenarios: a model you built from scratch, a time you fixed a failing model, a time your analysis changed a business decision, and a time you worked with a difficult stakeholder. Have your GitHub or portfolio ready to share if asked.

If you are still searching for openings, knok checks 150+ job sites nightly, applies to roles matching your resume, and messages HR for you. EXL had 398 open Data Scientist roles as of July 2026.

Memorising algorithms without understanding the 'why.' EXL interviewers typically ask follow-up questions. If you cannot explain why you chose XGBoost over a linear model for a specific problem, the answer falls apart.

Ignoring business context. Answering 'I would maximise accuracy' without discussing what accuracy means in that domain is a common mistake. Always tie your technical choices to the business outcome.

Weak SQL preparation. Many candidates over-prepare Python and under-prepare SQL. EXL roles often involve data pipelines and reporting, so SQL gaps are noticed quickly.

Vague STAR answers. Saying 'I improved the model significantly' without a concrete result is less convincing than saying 'the model flagged at-risk customers with strong recall on the holdout set.' Use specific, observable outcomes.

Not asking questions. EXL interviewers often read a lack of questions as low interest. Prepare two or three thoughtful questions about the team's client base, data stack, or how model results are presented to clients.

Skipping the communication step. If asked to explain a result and you jump straight to technical details, you miss the point. EXL is a client-facing company. Lead with the business finding, then offer to go deeper if needed.