One candidate reports Oliver Wyman Data Scientist interviews covering SQL, credit-risk case work, and a partner round. Another describes an Excel and presentation assessment day. Both accounts stress business communication.
$165K
Avg. Base Comp
$203K
Avg. Total Comp
5 rounds
Typical Rounds
2-3 months
Process Length
Two candidates describe quite different Oliver Wyman Data Scientist processes, so prepare for a range rather than a single script. Both ended in a rejection, and both accounts stress explaining the analysis in business terms.
One candidate reports five rounds over two to three months. It started with an HR conversation, then a SQL-focused technical round with a Lead Data Scientist covering CTEs and window functions at a medium level. Next came a credit-risk modeling take-home that was discussed in detail, and a second credit-risk case. The final round was with a partner and included a project discussion and a puzzle. This candidate was asked how to weigh precision against recall in a loan approval model. They were told in feedback that their communication was not clear enough, even though they solved the puzzle.
A second candidate describes an assessment day. It opened with exploratory data analysis in Excel on a dataset handed out that morning, followed by a group discussion of findings. Candidates also gave a presentation on a topic assigned weeks earlier, and the day closed with statistical reasoning questions in Excel. The panel cared more about business implications than model internals.
To prepare, practice lending-style tradeoffs, such as how to weigh precision against recall in a loan approval model. Rehearse SQL window functions and finish each analysis with a plain recommendation a non-technical stakeholder could act on.
Synthesized from 2 candidate reports by our editorial team.
Had an interview recently?
Share your experience. Unlock the full guide.
Real interview reports from people who went through the Oliver Wyman process.
The assessment day opened with an exploratory data analysis task in Excel on a dataset we were handed that morning. Once everyone had worked through it, we gathered to talk through our findings as a group, which felt less like a test and more like a stakeholder meeting where you have to defend what you noticed. The bigger piece was a presentation. We had been given the topic a few weeks in advance, so I prepared slides covering my analysis and an ML model I had built. What the panel seemed most interested in was not the model's internals but whether I could translate the results into business implications, the way you would when explaining to a non-technical stakeholder what to do next. I would spend more time on that framing than I did on the technical depth. The day ended with a set of analytical questions, again in Excel, that leaned on core statistical reasoning rather than advanced formulas. I came away feeling the process rewarded clear thinking about the business problem as much as technical execution. I did not receive an offer.
Prep tip from this candidate
Prepare your presentation so every finding ends with a concrete business implication, since that connection was emphasized over model details. Also drill statistical reasoning in Excel, because the closing questions tested it directly.
Share your own interview experience to unlock all reports, or subscribe for full access.
Sourced from candidate reports and verified by our team.
Topics based on recent interview experiences.
Featured question at Oliver Wyman
Suppose your default risk model has high recall but low precision. What business implications might this have for a mortgage bank?
| Question | |
|---|---|
| Text Editor With OOP | |
| 2nd Highest Salary | |
| Employee Salaries | |
| Size of Joins | |
| Bagging vs Boosting | |
| Hurdles In Data Projects | |
| P-value to a Layman | |
| Type-ahead Search | |
| RAG Strict Source Control | |
| Three Zebras | |
| Precision and Recall | |
| Production Model Monitoring | |
| Target Indices | |
| Fine-Tuning VS RAG | |
| Car Recommendation Architecture | |
| Transformer Encoder Layer | |
| Data Preparation for Imbalanced Data | |
| Assumptions of Linear Regression | |
| Duplicate Rows | |
| Type I and II Errors | |
| One-Hot Encoder | |
| Swap Variables | |
| Model Deployment Preparation | |
| SARIMA in Retail Forecasting | |
| Solo ML Deployment | |
| Overfit Avoidance | |
| Multicollinearity in Regression | |
| Credit Card Fraud Model | |
| Explaining Linear Regression to Different Audiences |
Synthesized from candidate reports. Individual experiences may vary.
One candidate reports the process opened with an HR call featuring standard behavioral questions, such as why they wanted to change jobs. Expect to explain your motivation and background clearly, since this may be a short screen before technical rounds.
The same candidate reports a second round with a Lead Data Scientist. It covered SQL (CTEs and window functions at a medium level), some foundational data science questions, and a discussion of a past project. Be ready to walk through your own work in detail.
This candidate reports a credit-risk modeling problem completed without internet access, followed by a detailed discussion with the interviewer. Practice building and defending a risk model and explaining each modeling choice.
A fourth round, as that candidate reports it, was another credit-risk case. It asked how to weigh precision against recall in a loan approval model. Be ready to tie that tradeoff to lending decisions.
The final round reported by this candidate involved a partner, a project discussion, and a puzzle. Feedback said communication was not clear enough, even though the puzzle was solved. Practice narrating your reasoning aloud and tying it to a decision.
A different candidate describes a day built around Excel exploratory analysis of a same-day dataset, a group discussion of findings, and a presentation prepared in advance. It ended with Excel-based statistical questions. The panel reportedly valued business implications over model internals.
