Data scientist is a title that collides with two others, and confusing them leads to expensive mishires. A data engineer builds the pipelines that move and store data. A machine learning engineer puts models into production reliably. A data scientist analyzes data to answer questions and drive decisions, and their real product is not a model or a pipeline but an insight that changes what the business does. The most common mistake is hiring a data scientist and expecting production engineering, or hiring one who builds elegant analyses that never actually influence a decision. Hire for impact on decisions, because that is the job.
Key Takeaways
- Data scientist, data engineer, and ML engineer are three different roles.
- A data scientist turns data into insights that drive decisions.
- The best ones are measured by decisions changed, not models or analyses produced.
- Screen for business impact and communication, not only statistical technique.
Three Roles, One Confusing Title
Sort out which you need before interviewing. A data engineer owns the infrastructure that moves and stores data reliably (how to hire a data engineer). An ML engineer takes models and makes them run in production at scale (how to hire a machine learning engineer). A data scientist sits between the data and the decisions, analyzing, running experiments, building models to answer questions, and communicating what it means so the business acts differently. Startups often say data scientist when they actually need a data engineer to build their data foundation, or an ML engineer to ship a model, and the mismatch wastes months. Decide what problem you are solving, then hire the matching role.
Measured by Decisions, Not Models
The trait that separates a strong data scientist from a merely technical one is impact. A weak data scientist produces sophisticated analyses and elegant models that are technically impressive and change nothing, because they never connect to a real decision or get communicated in a way anyone acts on. A strong one starts from the question that matters, does the analysis that answers it, and communicates the result so clearly that the business actually decides differently. Statistical skill is table stakes; the differentiator is whether their work moves decisions. So screen for impact: ask about analyses that actually changed what a company did, not the most technically complex thing they built.
| Weak data scientist | Strong data scientist |
|---|---|
| Elegant analyses, no impact | Analyses that change decisions |
| Starts from technique | Starts from the business question |
| Communicates in jargon | Communicates so people act |
| Measured by models built | Measured by decisions changed |
A Concrete Version
Ask a candidate to describe a project where their analysis changed what the company did. A strong data scientist tells a clear story: here was the business question, here is what I found in the data, here is how I communicated it, and here is the decision that changed as a result. The impact is the point of the story. A weaker candidate describes the most technically sophisticated thing they built, the fancy model, the complex method, without a clear line to a decision it influenced, which reveals that they optimize for technical impressiveness over business impact. That difference tells you whether they will drive decisions at your company or produce analyses nobody uses.
The Honest Counterpoint
Impact-first does not mean statistical rigor is optional, and a data scientist who communicates well but analyzes sloppily is dangerous, because confident wrong conclusions drive bad decisions. The point is not to trade technique for communication but to require both: sound analysis and the ability to turn it into a decision. There are also genuinely research-heavy roles, in some companies, where deep methodological depth is the point and immediate business impact is a longer game; those are real but rarer than most startups need. For the typical startup, hire the data scientist whose rigor and communication combine to change decisions, and be wary of both the impressive analyst who changes nothing and the persuasive one whose analysis does not hold up.
Cost and Sourcing
A senior data scientist in the US commonly runs $150 an hour or more, higher with strong domain or ML depth. Nearshore in Latin America, the same seniority lands around $60 to $95 an hour, with the overlap that helps because data science is tightly coupled to the business questions it serves (staff augmentation for AI and ML teams). Screen for decisions changed and clear communication alongside sound technique, decide first whether you actually need a data scientist rather than a data or ML engineer, and hold the bar with a rigorous vetting process. See available engineers.
Frequently Asked Questions
What is the difference between a data scientist, data engineer, and ML engineer?
A data engineer builds data pipelines and storage. An ML engineer puts models into production reliably. A data scientist analyzes data to answer questions and drive decisions. Startups often confuse them and mishire; decide which problem you are solving first.
What should I test when hiring a data scientist?
Business impact and communication alongside statistical rigor. Ask for a project where their analysis actually changed what a company did, and look for a clear line from question to insight to decision, rather than the most technically complex thing they built.
Why measure a data scientist by decisions rather than models?
Because their real product is an insight that changes what the business does. A sophisticated analysis that never influences a decision has no value, so the differentiator is impact, not technical impressiveness.
How much does a data scientist cost?
In the US, commonly $150 an hour or more for a senior. Nearshore in Latin America, around $60 to $95 an hour at the same seniority.
The Bottom Line
Hiring a data scientist starts with not confusing the role with data engineer or ML engineer, then screening for the thing that actually matters: impact on decisions. The best data scientists combine sound analysis with the ability to communicate it so the business acts differently, and they are measured by decisions changed rather than models built. Require both rigor and communication, decide first whether a data scientist is even the role you need, and you avoid the common mishire of an impressive analyst whose work changes nothing.
Roberto Espinoza is CEO of Ruzora, which helps US startups hire pre-vetted senior LATAM engineers, with a vetted shortlist in 72 hours. See available engineers.
