Data Scientist

Also known as: Data Science Specialist, Applied Scientist, Research Scientist, Decision Scientist, Machine Learning Scientist.

What does a Data Scientist do?

A Data Scientist is a specialist who takes a vague business question ("why are customers leaving", "how much to stock", "who should get credit"), turns it into a data problem and finds an answer the business can rely on. Their tools are statistics, experiments and machine learning models, and the result is a decision someone makes differently than they would have without it. Typical tasks:

• define the problem with the business owner and decide how success will be measured;
• explore the data and check whether the needed signal is there;
• build forecasting, churn, scoring and demand models;
• design experiments and estimate effects when a clean A/B test is not possible;
• explain the result to those who will act on it and be honest about uncertainty.

When does a company need a Data Scientist?

A company needs a Data Scientist when it has accumulated data and faces costly decisions currently made on intuition: pricing, purchasing, credit limits, customer retention. The first Data Scientist usually arrives after a warehouse is in place and analytics exists, and the business has questions a dashboard cannot answer. Another reason is a company that wants to find out whether investing in ML is worth it before hiring a whole team. The role will not work if the data is scattered and dirty: the Data Scientist will spend the first six months fixing pipelines, and it is more honest to hire a Data Engineer first.

Data Scientist vs ML Engineer: what's the difference?

A Data Scientist differs from an ML Engineer in being responsible for the right answer rather than for a working system. A Data Scientist decides what problem to give a model, on what data, and how to tell whether it helps the business; an ML Engineer takes that result and turns it into a service that works under load. A Data Scientist overlaps with a Product Analyst on experiments, but the analyst answers the product team's questions about user behaviour, while a Data Scientist more often builds models for business decisions.

Why is it hard to hire a Data Scientist?

Data Scientists are hard to hire because people doing completely different jobs work under this title, and a CV barely helps to tell them apart.

• Some Data Scientists are product analysts in practice, others are ML researchers, others write production code. Searching by title brings in all of them.
• Strong statistics and the ability to talk to the business rarely come together: many candidates build accurate models nobody uses.
• The impact of the work is hard to see: "built a churn model" does not say whether a single decision changed.
• Experienced Data Scientists with domain expertise, for example in credit risk or pricing, are few, and they are usually not looking.

What to decide before hiring a Data Scientist?

• Which decisions should this person's work change? Without a specific decision, a Data Scientist drifts into research with nowhere to apply the results.
• Do you need analytics and experiments, or models in production? These are different profiles: the first is closer to an analyst, the second to an ML Engineer.
• Who will take models to production? If nobody, you need a Data Scientist with engineering skills, and there are fewer of them.
• How much does the domain matter? In credit risk, insurance or healthcare, industry experience can outweigh knowledge of the latest models.

Where to find a Data Scientist?

Data Scientists most often come from fintech, banks, e-commerce, marketplaces, retail and product companies with a strong data culture; some come from academia, physics and mathematics. A strong profile signal is a story about which decision changed thanks to a model or an experiment, not just which algorithm was used. NEWHR finds such candidates by the results of their work, not by applications to a job post.

How to assess a Data Scientist in an interview?

Data Scientists are assessed with a case on the company's data and a review of a past project. In the case, the candidate frames the problem, chooses an approach, decides how to measure success and explains the result the way they would to a business owner. On the company side, the Head of Data and the person who will use the results take part.

Strong signals:

• first asks which decision is behind the question, and only then chooses a method;
• names the limits of their conclusions and what could have distorted them;
• can give an example where a simple model turned out more useful than a complex one.

Red flags:

• immediately proposes a neural network without looking at the data;
• judges success only by a model metric, with no link to the business;
• cannot explain their result without jargon.

How long does it take to hire a Data Scientist?

We present the first candidates for the Data Scientist role within 5–7 days of the brief.

Related roles

• ML Engineer — takes models to production and owns how they work under load.
• Product Analyst — answers the product team's questions about user behaviour and experiments.
• Data Engineer — builds the pipelines and warehouse without which a Data Scientist works with dirty data.
• Quantitative Researcher — applies similar methods to market data, where a mistake costs money immediately.
• CDO (Chief Data Officer) — owns the company's data strategy and leads the data teams.