Data scientist job description template: the kind of data science, the decisions it feeds and the evidence to ask for
On this page
"Data scientist: use machine learning and statistics to extract insights from data" covers an experimentation specialist who designs A/B tests for a product team, a modeler who ships a fraud score into production, a decision scientist who forecasts demand for finance, and a research scientist who publishes. Each reads the posting looking for the same thing: what decisions or products their work will change, what data exists, who builds the pipelines, and how models reach users. A useful data scientist job description answers those and asks for evidence of finished work rather than a list of libraries. Below are the seat types, a copy-ready template, the education and classification lines with sources, screening and scorecard rows, and the mistakes that turn data science postings into keyword lists. For the first call, use the data scientist phone screen questions.
Name the seat first
| Seat | What the work looks like | Must-have that separates it |
|---|---|---|
| Product or experimentation data scientist | Test design, power and sample size, metric definitions, readouts to product teams | Has designed and read out experiments and changed a launch decision with the result |
| Machine learning or modeling data scientist | Feature work, training, validation, monitoring a model that users or systems rely on | Has shipped a model to production and handled drift or a performance drop |
| Decision or business data scientist | Forecasts, causal estimates, pricing or capacity analysis for leaders | Has built a forecast or estimate that leadership used and checked it against what happened |
| Research scientist | New methods, papers, prototypes handed to engineering | Has publications or method work, plus evidence of transfer into a product |
| Domain data scientist (health, credit, insurance) | Models under regulatory or fairness review, documentation, validation with risk teams | Has taken a model through review and documented its limits |
Then settle the facts: the decisions or products the work feeds, the data available and its condition, whether data engineers build the pipelines, how a model reaches production and who owns it there, the main language and platform, and who the data scientist reports to. If the role is mostly dashboards and reporting, start from the data analyst job description template; if it is mostly pipelines, the data engineer job description template.
Education and classification lines, with sources
Not legal advice. The points below were checked against the BLS Occupational Outlook Handbook and Department of Labor fact sheets as of October 2026. State overtime rules can be stricter; confirm classification with counsel.
Degree wording. BLS says data scientists typically need at least a bachelor's degree, that some employers require or prefer a master's or doctoral degree, and lists duties such as creating, validating, testing and updating algorithms and models (BLS Occupational Outlook Handbook, last modified August 27, 2026). Require an advanced degree only for research seats; for applied seats, write "or equivalent experience" and mean it.
Computer employee exemption. It covers employees whose primary duty is applying systems analysis techniques, or designing, developing, documenting, analyzing, creating, testing or modifying computer systems or programs, paid on a salary basis at the standard level or at least $27.63 an hour (DOL Fact Sheet 17E). The Department of Labor lists the standard level as $684 a week. A data scientist who builds production models may fit; one who mostly runs reports may not.
Learned professional exemption. It requires a primary duty of work requiring advanced knowledge in a field of science or learning customarily acquired by a prolonged course of specialized intellectual instruction, plus the salary test (DOL Fact Sheet 17D). Data science is not named in the regulation, so decide on the actual duties, not the title.
The data scientist job description template
[Senior] Data Scientist, [Experimentation / Machine Learning /
Decision Science / Research] — [Product area or business unit]
[Company], [city, state] — [On-site / Hybrid: days / Remote within:
states]
Pay: $[min]–$[max] per year, plus [bonus / equity, if any].
[Benefits summary, if required.] [Exempt, after review.]
SUMMARY
Your work will shape [decision or product], used by [who]. You
will work with [data source(s)], which [data engineers maintain /
you will help shape]. Models reach production through [process],
and [you / ML engineers] own them after launch. You will report to
[title] and work with [product, engineering, finance, operations].
WHAT YOU WILL DO
- [Design experiments, choose metrics and read out results.]
- [Build, validate and monitor models for (use case).]
- [Forecast or estimate (quantity) and track accuracy.]
- Explain results and their uncertainty to [audience] in writing.
- Document data, assumptions and limits so others can reuse work.
YOU MUST HAVE
- [N] years of applied statistics or machine learning, with work
you can walk through end to end.
- Working skill in [language] and [query language], or similar.
- [Production: has shipped and maintained a model.]
- [Experimentation: has designed tests and handled pitfalls such
as low power or multiple comparisons.]
- Bachelor's degree in a quantitative field or equivalent
experience. [Research seats only: PhD or equivalent research.]
NICE TO HAVE
- Experience with [platform], [domain], [method].
If you meet the must-haves and none of these, please apply.
HOW WE HIRE
[Recruiter screen; conversation with (lead) about a project you
choose; a take-home or live exercise of at most (N) hours using
data we provide; meeting (product / engineering) partners.]
[EEO STATEMENT]
Need an adjustment to apply or interview? Contact [address].Pay and EEO
Post the approved range and check pay transparency laws by state. BLS publishes pay data for data scientists in the Handbook profile linked above; this page does not quote figures. Close with a short EEO statement and an accommodation contact.
From must-haves to screening questions and scorecard rows
| Must-have | Screening question | Scorecard competency | Evidence of a strong answer |
|---|---|---|---|
| End-to-end ownership | "Pick one project. What decision did it change, and how do you know?" | Business impact | Names the decision, the alternative and the measured outcome |
| Statistical judgment | "Tell me about a result that looked significant but was not. What gave it away?" | Statistical rigor | Leakage, peeking, multiple comparisons or bias, found and explained |
| Modeling (if production) | "How did you know your model was still working three months after launch?" | Production readiness | Monitoring, a drift signal, a retraining or rollback decision |
| Data quality | "What was wrong with the data on your last project, and what did you do about it?" | Data judgment | Specific issues, checks they built, a conversation with the data owner |
| Communication | "How did you explain uncertainty to someone who wanted a single number?" | Stakeholder communication | Ranges, plain language, a recommendation despite uncertainty |
The later rounds are in the interview guide for data scientists, and the first months in the 30-60-90 day plan for data scientists. If the seat is closer to deploying models than to analysis, the machine learning engineer screening questions fit better.
Common mistakes in data scientist job descriptions
A list of every library
Ten frameworks in the must-haves filter for keyword matching, not judgment. Name the main language and platform and accept similar ones.
No word on the data
Candidates want to know whether the data exists, is clean and has an owner. Saying "we are early; you will help define logging" filters for the right person.
A PhD by default
Unless the seat is research, an advanced degree requirement excludes strong applied candidates.
Hiding that it is mostly reporting
If most of the week is dashboards, call it an analyst role. Data scientists who expect modeling leave quickly.
No path to production
If no one deploys models, say so. A modeler who learns this in month two will not stay.
Before you publish
- The seat, the decisions or products it feeds, and the data situation are in the summary.
- The path from model or analysis to use is described, with owners.
- Degree wording matches the seat, with equivalent experience accepted where it fits.
- Tools are limited to the weekly ones, and classification has been reviewed.
- The pay range, EEO statement and accommodation contact close the posting.
Questions people ask
What is the difference between a data scientist and a data analyst?
Analysts mostly describe what happened and why, using queries, dashboards and reports. Data scientists more often build models, design experiments or estimate effects that drive a decision or a product feature. Titles overlap heavily, so the job description should say which outputs the role produces.
Should a data scientist job require a PhD?
Only for research seats that publish or invent methods. BLS says data scientists typically need at least a bachelor's degree and that some employers require or prefer a master's or doctoral degree. For applied seats, ask for evidence of the work, such as a model in production or an experiment that changed a decision, and accept equivalent experience.
Is a data scientist exempt from overtime?
Often, but the title does not decide it. Some data scientists fit the computer employee exemption, which covers systems analysis and programming duties and can be met by a salary at the standard level or an hourly rate of at least $27.63, and others fit the learned professional exemption. Both turn on the actual primary duty. Review the seat and check state law.
Which tools should the posting list?
List the language and the data platform the person will use every week, and say whether similar tools are acceptable. A long list of libraries filters out good candidates who use a different but equivalent stack.