A data analyst answers questions from data you already trust, a data engineer builds the pipelines that make the data trustworthy, and a data scientist answers the questions that need statistics. An ML engineer comes later still, putting models into production. The titles overlap in job ads, which is how a team can bring in a data scientist first and watch them spend months cleaning spreadsheets. This guide explains what each role does, what each one needs before they can be useful, and how to choose your first.
The four data roles at a glance
| Role | What they do | What they need first | A first deliverable you could write down |
|---|---|---|---|
| Data or BI analyst | Turns business questions into reports and dashboards, in SQL and a BI tool such as Power BI, Tableau or Looker | Data in one place, and someone who owns the questions | One KPI dashboard rebuilt and reconciled to the source data, with every metric defined |
| Data or analytics engineer | Builds the pipelines that bring data into your warehouse, and the models on top of it | Access to the source systems and a warehouse, even a small one | One data source loaded on a schedule, with tests and documentation |
| Data scientist | Forecasts, experiments, segmentation, and the analysis behind a decision | Clean, modeled data and a decision that depends on the answer | An analysis of one business question, with the method, the result and its limits |
| ML engineer | Puts machine-learning models into production and keeps them working | A model worth running and the infrastructure to run it | One model moved from a notebook into a scheduled pipeline, with monitoring |
What a data analyst does
An analyst sits closest to the business. Someone asks "why did churn go up in March?" or "which channel brings our best customers?", and the analyst finds the answer in the data, then builds the report or dashboard so the question does not have to be asked again.
The work that matters most is unglamorous: agreeing what each metric means. Two dashboards that disagree about "active customers" do more damage than no dashboard at all, so a good analyst writes the definitions down and reconciles the numbers to the source before anyone relies on them.
An analyst needs the data to already be in one place they can query. If it is not, they will spend their time exporting and stitching files by hand, and the reports will break every time a source changes.
What a data engineer does
A data engineer builds the plumbing. Your data lives in your product database, your CRM, your billing system, your ad platforms and a dozen spreadsheets. The engineer writes the pipelines that load it into one warehouse on a schedule, tests that it arrived intact, and models it into tables an analyst can use without knowing where each column came from.
The title "analytics engineer" usually means the modeling half of that job: turning raw loaded tables into clean, documented ones. For a small team, one person often does both.
A data engineer is the right first person when the honest answer to "where does this number come from?" is "it depends who you ask."
What a data scientist does
A data scientist answers the questions that need statistics rather than counting. Will this price change lose customers? Did the new onboarding flow actually work, or was it a good month? Which customers are likely to leave, and why?
That work depends on clean, modeled data. Point a data scientist at raw exports and most of their time goes on cleaning, which is a data engineer's job done slowly.
What an ML engineer does
An ML engineer takes a model that works in a notebook and makes it work in production: training pipelines, serving, and monitoring the inputs and outputs so you know when the model starts to drift. If nobody has built a model worth running yet, you do not need one.
Which data role to bring in first
- Your data is already in one warehouse and you need answers from it: start with an analyst.
- Your data is spread across tools and nobody trusts the numbers: start with a data engineer. Every later role depends on that foundation.
- You have clean data and a decision that depends on a forecast or an experiment: a data scientist.
- You have a model that needs to run reliably: an ML engineer.
Most teams that are new to data need an engineer or an analyst first, and are tempted by a data scientist. The order matters because each role stands on the one before it.
Mid-level or senior
Choose senior if nobody on your team will set direction or review the work: a senior person has owned this kind of work alone before. Choose mid-level if someone senior on your team, or a clear product lead, will set direction and review it. Your first data person is often the only one, which usually means senior.
Giving a contractor access to your data
Data roles work inside your systems, so plan access before the first day:
- Give the access the work needs, and no more. A read-only warehouse role covers most analyst work.
- Regulated data needs written terms first. Under our agreements you do not give a contractor protected health information, payment-card data or similarly regulated personal data unless we have agreed in writing how it will be handled, and we sign a data processing addendum where the law requires one (MSA §8).
- Location matters. We never permit access from a country under comprehensive U.S. sanctions, or from one the U.S. Department of Justice lists as a country of concern for sensitive personal data (MSA §8).
- Remove access when the engagement ends. You hold the keys, so you remove them. A laptop or license you issued is returned or shut off (SOW §2).
Frequently asked questions
Is a data engineer the same as a data scientist?
No. A data engineer builds the pipelines and models that make data reliable; a data scientist uses that data to answer statistical questions. A data scientist without a data engineer usually ends up doing the engineering, slowly.
Do I need a data engineer or a data analyst?
If your data is already in one place and you need reports from it, an analyst. If the first problem is getting the data into one place and agreeing on the numbers, a data engineer.
When should I bring in an ML engineer?
When you have a model that works and needs to run in production. Before that, an ML engineer has nothing to put into production.
Can a contractor work in our data warehouse?
Yes, with the access you give them. They work in your systems, you control the access, and every contractor is bound by written confidentiality terms (MSA §6).
Where to go from here
sourceBOLD engages mid-level and senior data analysts, data engineers, data scientists and ML engineers across the Americas, month to month, as their contractor of record. Data roles shows typical starting prices for each, and how we vet candidates covers what happens before you meet anyone.