Data Scientist Job Description Template (Copy and Paste)
“Did the free-dessert offer keep new subscribers, or did it only push their cancellations back a month?” A meal-kit company that can write that question down is ready to hire a data scientist. A company that can only write “use data to drive growth” is describing a hope, and whoever it hires will spend the first months working out what the question is before any analysis starts.
A data scientist job description should begin with the questions. The role exists to inform decisions under uncertainty: whether a change caused an outcome, what demand will look like next month, which customers behave differently and why, and whether a model would do better than the rule of thumb in use today. The template below is written around those decisions. Fill in the questions, the people who act on the answers and the data that exists today before you list a single library, because those three things decide what kind of data scientist you need.
What Does a Data Scientist Own?
The role owns the analytical answer to a business question. That means framing the question so it can be tested, choosing a method, checking whether the data can support a conclusion, and explaining the result, with its uncertainty, to the people who will act on it. Experiment design, statistical modeling, forecasting and exploratory analysis are the core tools. Machine learning is one of those tools, and plenty of strong data science work never needs it.
The boundary with neighboring roles matters more than the title. A data engineer makes data available and reliable. A machine learning engineer takes a model that has earned its place and builds the system that trains, serves and monitors it. An AI engineer builds product features on top of foundation models. The data scientist decides which questions are worth answering, whether an answer needs a model at all, and what the evidence supports. If one posting asks for all four, it is describing a team. Gun.io’s comparison of data science and data engineering covers where the first of those handoffs usually sits.
| Role | Owns | Typical output |
|---|---|---|
| Data scientist | Question framing, experiments, forecasts, causal and predictive analysis | A recommendation with stated uncertainty, and a prototype model when one is warranted |
| Data analyst | Reporting and descriptive analysis on agreed metrics | Dashboards and recurring reports |
| Data engineer | Pipelines, storage, freshness and data quality | Reliable tables that others can build on |
| Machine learning engineer | Training, serving and monitoring models in production | A model running as a dependable service |
Four Questions to Answer Before You Post
Which decisions will this person inform? “Set the weekly ingredient order for each distribution center” and “decide whether to keep the dessert promotion” call for different people. The first is forecasting with seasonality, holidays and supplier lead times. The second is experiment design and causal inference on a subscriber base where people often choose their own offers, which makes a naive before-and-after comparison misleading.
Who acts on the answer, and how often? A forecast the procurement team uses every Monday has to be reproducible and explained the same way each week. A one-off pricing study needs a defensible method and a clear write-up. If nobody has agreed to act on the work, the role will produce analysis that nobody reads.
What data exists, and who owns it? Suppose order history lives in the subscription platform, failed deliveries in a carrier portal and promotion assignments in a spreadsheet the marketing lead maintains by hand. Expect a data scientist to spend real time joining and questioning those sources, and say so in the posting. If that joining work is large enough to be a job on its own, the first hire may need a data engineer job description instead.
Does the work end in a decision or in a running model? A data scientist can build a churn model and show that it beats the current rule. Scoring it daily, retraining it and monitoring it is production engineering. Decide who owns that step before you hire, so the posting neither promises work that belongs to someone else nor leaves the step without an owner.
Copy-and-Paste Data Scientist Job Description Template
Replace everything in brackets. Delete any line that does not apply instead of leaving a generic version in place; a short, specific posting reads as a real job.
Title and Working Arrangement
Title: [Data Scientist / Senior Data Scientist]
Team and reporting line: [Team name], reporting to [role]
Location and hours: [Remote, hybrid or on-site, with any required time zone overlap]
Engagement: [Full-time employment or contract, and the expected duration if contract]
Summary
[Company] [one sentence on what the company does]. We are hiring a data scientist to help [team or leaders] decide [specific decision, for example “how much of each ingredient to order for each distribution center every week” or “which retention offers to keep”]. Today, [describe how the decision is made now and what happens when it is wrong]. You will work closely with [teams] and report to [role].
Questions You Will Own
- [First question, written as a decision: “Does the first-box discount change twelve-week retention once seasonality is accounted for?”]
- [Second question]
- [Third question, only if there is one. Fewer, clearer questions attract stronger candidates than a long list.]
Responsibilities
- Turn business questions from [teams] into testable hypotheses, and agree in advance what result would change the decision.
- Design experiments, including assignment, sample size and the metric that defines success, and say plainly when an experiment cannot answer the question.
- Build forecasts or statistical models for [use case], and compare each one against the current method or a simple baseline before recommending it.
- Investigate data quality in [named sources], document assumptions, and report where the data limits the conclusion.
- Write up results for [audience], including uncertainty, limitations and the next step you would take.
- Work with [data engineering, ML engineering or product engineering] to hand off any model or metric that has to run on a schedule.
Requirements
- Applied statistics: experiment design, regression, and reasoning about bias, confounding and variance.
- [Python or R] for analysis, written so another person can rerun and review it.
- SQL strong enough to build and check your own datasets in [warehouse or database].
- Experience presenting a result that changed a business decision, or deliberately left it unchanged.
- Clear writing for readers who are not technical.
Nice-to-Haves
- [Forecasting for inventory, demand or capacity]
- [Causal inference methods for observational data, such as difference-in-differences]
- [Domain experience in subscription retention, logistics, healthcare or financial services]
- [Specific tools from your environment, only where they cannot be learned quickly]
How We Work
[Two or three sentences on how decisions get made, who reviews analysis before it reaches leadership, and how often the data scientist meets the people who use the work.]
How to Adapt the Template by Company Stage
The same title covers different jobs at different stages. Adjust the summary, the questions and the requirements together; changing the seniority word in the title does not change the work.
| Stage | What the role usually is | What to change in the template |
|---|---|---|
| Early startup, first data hire | A generalist who also defines metrics and builds basic pipelines | State that the role includes data engineering work, and weight SQL and judgment above advanced modeling |
| Growth company with a data platform | A specialist on a defined problem such as experimentation, forecasting or pricing | Name the problem area and partner teams, and list only the methods that problem needs |
| Enterprise or regulated environment | The analyst of record for decisions that are audited or reviewed | Add documentation, model review and approval requirements, and name who signs off |
At an early-stage company, a first data scientist often has to define what an “active subscriber” is before retention can be measured at all. That is valuable work, and a posting that hides it will attract someone who expects a clean warehouse. A larger company faces the opposite risk: a narrow role written as a generalist one attracts people who will be frustrated by a well-defined scope.
A junior data scientist job description needs a different emphasis again. A junior hire can run sound analyses inside a defined question while a senior reviewer checks the framing and method. If nobody is positioned to give that review, the role is senior regardless of what the title says, and the requirements should reflect it.
What to Look for in Data Scientist Candidates
The most useful evidence is a past analysis the candidate can walk through: the question they were handed, how they reframed it, what the data could and could not support, what they recommended and what happened next. Ask what they would change with another month. A candidate who explains why an experiment came back inconclusive, and what that saved the business from doing, can be more valuable than one with an impressive model and no account of what it changed.
Be wary of a method chosen before the question. Someone who reaches for a gradient-boosted model before asking how subscribers were assigned to the dessert offer may produce a precise answer to the wrong question. Interviews and portfolios give you evidence about how a person reasons; how they work with your data and your decision makers is the real evidence.
Where Data Science Work Meets Production
When an analysis shows a model is worth running every day, the work changes character. Code written to answer a question rarely has the structure, logging and monitoring a production service needs, and nobody should expect it to. One Gun.io brief describes an academic geospatial research system that had been built for a PhD thesis: the code was brittle, had no logging or monitoring, and a visualization took 10 to 15 seconds to load a single view before the architecture was reworked for production use.
Plan for that transition in the job description. Either the data scientist hands work to an engineer, in which case name the role that receives it, or the position includes production responsibilities, in which case list them under requirements and expect a different candidate. Teams that need that second profile often look for machine learning engineers instead of, or alongside, a data scientist.
How Gun.io Thinks About Data Science Fit
At Gun.io, we start with the decision the work is meant to inform, then look at the data behind it and the people who will use the result. Sometimes that points to a data scientist. Sometimes the constraint sits upstream in unreliable data, or downstream in getting a model into production, and the first engineer you need is different from the title on the request. Once the problem is clear, we support data work through staff augmentation, where the engineer works inside your team under your direction, or scoped project delivery, where we take a larger part in organizing a defined piece of work. You can see the engagement models and engineering areas we cover in one place.
We assess technical capability and professional fit through human interviews. Because how someone performs on the actual work is the evidence that counts, selection continues after an engagement begins rather than ending at admission.
Frequently Asked Questions
What Should a Data Scientist Job Description Include?
It should include the decisions the role will inform, the people who act on its work, the data sources available, the methods the problem calls for, and who owns anything that must run in production. Add the working arrangement and how analysis is reviewed. A list of libraries matters less than these, because these are what decide the profile.
Does a Data Scientist Need a PhD?
Most applied business roles do not need one. A PhD signals research training and suits work such as developing new methods or working in a scientific domain, but applied roles depend more on experiment design, sound statistics, SQL and clear communication. Require a graduate degree only when the work itself calls for it.
How Much Machine Learning Should a Data Scientist Know?
Enough to build a model and test it honestly against a simple baseline, because that comparison is part of the job. The more important skill is recognizing when no model is needed, since a seasonal average or a well-designed experiment answers many questions better. Deep production machine learning belongs in a separate posting.
What Is the Difference Between a Data Analyst and a Data Scientist Job Description?
An analyst description centers on reporting and descriptive analysis of agreed metrics. A data scientist description centers on questions without a ready answer, such as causal effects, forecasts and predictions, with the uncertainty stated. Many companies need the analyst first and the data scientist once those metrics are trusted.
How Many Questions Should the Posting List?
Two or three is usually enough, each written as a decision someone will make. A long list suggests the company has not decided what the role is for, and experienced candidates tend to read it that way.
What Drives the Cost of Hiring a Data Scientist?
Cost follows the complexity of the questions, the time the work needs and the experience required to answer them reliably. A role that must also define metrics and build pipelines, or that works with regulated data, needs more experience than one answering a defined question on clean data. Scope the problem first, and the cost conversation follows from it.
Hire for the Question Before the Method
The posting works when a candidate can read it and name the first decision they would help you make. If you are deciding whether your problem calls for a data scientist, an engineer or both, talk to us about the decision you need to make, the data behind it and who will act on the answer.
