Hire Data Scientists Who Answer Business Questions

How to vet a data scientist

Good data scientists answer business questions with evidence, not just build models. Ask candidates to walk through a project from the original question to the decision it changed: how they explored and cleaned the data, which methods they chose, how they validated results and how they presented findings to people without statistical training. Projects that ended with a clear recommendation matter more than impressive algorithms.

Check fundamentals: SQL fluency, statistics such as hypothesis testing and confidence intervals, experiment design, and Python or R with pandas and visualization libraries. For predictive work, look for awareness of overfitting, data leakage and honest evaluation. Our guide to predictive analytics outlines the kind of work many business data scientists do.

Communication is often the deciding skill. Ask candidates to explain a past analysis in two minutes to a non-technical manager. Clear, honest explanations, including uncertainty and limitations, build the trust that turns analysis into action. Ask them to also explain what the analysis could not prove.

• Strong SQL and data cleaning skills.
• Solid statistics and experiment design.
• Validates models honestly and avoids leakage.
• Communicates findings clearly with visuals.
• Connects analysis to business decisions.
• Documents work so others can reproduce it.
• Comfortable with BI tools for sharing results.

Interview questions we use for data scientists

We combine a short data exercise with discussion of past projects. Candidates explore a realistic dataset, explain what they would check first and propose how they would answer a business question, including what they would not conclude from the data and which extra data they would request.

Strong answers question the data before trusting it, choose simple methods when they suffice and describe results with appropriate uncertainty. We value candidates who say when an A/B test or more data is needed rather than overstating what a dataset can prove.

• Sales dropped last month. How would you investigate why?
• How would you design an A/B test for a pricing change?
• What is the difference between correlation and causation in a real example?
• How do you handle missing or inconsistent data?
• How would you validate a forecast before the business relies on it?
• How do you present uncertain results to executives?

Data scientist or machine learning engineer?

Data scientists focus on questions, analysis, experiments and models that inform decisions. Machine learning engineers focus on building and running models as reliable production software. Many organizations need both: data scientists to discover what works and engineers to deploy it at scale. Hiring the wrong profile for the job is a common reason data initiatives stall.

If your goal is insight, reporting and experimentation, start with a data scientist, supported by our data analytics services where infrastructure is missing. If your goal is a model serving predictions inside a product, our machine learning engineers are usually the better fit.