Deloitte Data Scientist Interview Loop
Deloitte hires data scientists into consulting teams. The loop checks the technical work and whether you can sit in front of a client. For experienced hires it is a recruiter call, a hiring manager conversation and a technical step. Then fit interviews with a manager and with a senior manager or partner. The technical step is one of three things. A take-home modeling task, an online test, or a live interview with the team on ML, statistics, SQL and Python. Campus candidates get the consulting format instead. That is a behavioral interview and a case, then a final round with a group case.
What decides it
Explaining technical work to a non-technical audience. The fit rounds with senior people are not a formality. They decide whether you can be put in front of a client.
Practice by round
SQL on project and timesheet data: over-budget projects, active projects by industry, project duration by practice, average bill rate and budget range by practice. Deloitte’s SQL questions are plain, and they expect them right.
Harder SQL like top clients by revenue in a year, and Python for a small modeling task. The take-home, when there is one, is a classification problem with a notebook and a short write-up, sometimes presented back to the team.
| Question | Difficulty |
|---|---|
| Difference Between Times | Medium |
| Fastest Hometowns | Medium |
| Election Results | Medium |
| Maximum of Two Numbers | Medium |
| Top 3 Year Month Sales | Medium |
| Population Density | Medium |
| Top Clients by Revenue in 2025 | Medium |
| Minimum Number of Platforms | Hard |
The client case: where client churn comes from, and the data projects that take a business question through to a recommendation. In the live version you may get a dashboard and be asked what is wrong with it. Then you explain that to someone who does not code.
| Case study | Difficulty |
|---|---|
| Where Client Churn Comes From | Hard |
| Data project | Difficulty |
|---|---|
| Predicting Price | Easy |
| Terrorism Hotspots | Medium |
| Modelling Churn in Energy Company | Hard |
| Financial Health Check | Hard |
Model and statistics fundamentals asked plainly: reading a ROC curve, how k-means works, probability versus likelihood, posterior probability, skewed data and a normality test. Increasingly also how retrieval-augmented generation works and when an LLM can judge another model.
| Question | Difficulty |
|---|---|
| Correlation Coefficient | Easy |
| K-means Algorithm | Medium |
| Shapiro-Wilk Test | Medium |
| Skewed Data | Medium |
| DMatrix in XGBoost | Medium |
| Probability vs Likelihood | Medium |
| Posterior Probability | Medium |
| ROC Curve | Hard |