An analytics engineer builds and maintains the tested, documented data models that everyone else's numbers come from; a data analyst uses those models to answer questions. The analyst asks "why did revenue fall in March?"; the analytics engineer makes sure there is one agreed definition of revenue, that it reconciles to the books, and that the table behind it is fresh and correct every morning. In the 13 analytics engineer postings we counted on 29 September 2026, all 13 asked for SQL, dbt and a cloud warehouse. That is the job in one line.
The title grew up around dbt, which let analysts write transformations as version-controlled SQL with tests, the way software engineers write code. Once the modelling layer became code, someone had to own it as an engineer would: tests, reviews, documentation, deployments. That person is the analytics engineer.
What does an analytics engineer do day to day?
Roughly four things:
- Modelling. Turning raw tables from the product database, the CRM and the billing system into clean, dimensional models: orders, customers, subscriptions, with the joins and edge cases handled once, in one place.
- Testing and documentation. Every model gets tests (no duplicate keys, no nulls where there should be none, totals that match the source) and a description a new analyst can read.
- Metrics. Agreeing definitions with the people who use them. What counts as an active customer, when revenue is recognised, which refunds are excluded. Then encoding those definitions so every dashboard uses the same one.
- Operating it. Freshness checks, broken-pipeline alerts, warehouse cost, and a change process so that renaming a column does not silently break the board report.
The measure of success is dull and important: when two teams quote a number, it is the same number.
How is an analytics engineer different from a data analyst?
The analyst works on top of the models; the analytics engineer works on the models. The ads show the difference clearly, even at small sizes. From the same 29 September fetch (3,610 postings from Hacker News "Who is hiring?", Remotive and Arbeitnow):
| What the ad mentions | Analytics engineer (13) | Data analyst (10) |
|---|---|---|
| SQL | 13 | 8 |
| dbt | 13 | 3 |
| Snowflake, BigQuery or Redshift | 13 | 3 |
| Python | 9 | 7 |
| Airflow, Dagster or Prefect | 7 | 0 |
| Data pipelines or ETL | 6 | 0 |
| CI/CD | 5 | 0 |
| Stakeholders or cross-functional work | 11 | 3 |
| Stated years of experience (median) | 11 ads, median 5 | 4 ads, median 5 |
Three things stand out. SQL, dbt and the warehouse appear in every analytics engineer ad, against a minority of analyst ads. Orchestration, pipelines and CI/CD, which are engineering vocabulary, appear only on the analytics engineer side. And the analytics engineer ads mention stakeholders more often, not less: 11 of 13. Owning the definitions means negotiating them.
Python is roughly level between the two (9 of 13 and 7 of 10), so it is not what separates them. Neither role reads as entry level in these ads: where years were stated, the median was five for both, though only four analyst ads stated any.
The small print: 13 and 10 are small numbers, and we give counts rather than percentages for that reason. The sources are startup- and Europe-heavy (11 of the 13 analytics engineer postings came from Arbeitnow), some Arbeitnow postings are duplicated, and Australia barely appears. The counts file is public, with the method in it.
Where does AI fit into analytics engineering?
In two places, and neither replaces the modelling.
AI over the metrics layer. Companies want people to ask questions of their data in plain language. That only works if the model answers from agreed definitions; otherwise it invents a plausible revenue figure. A well-built metrics layer is what makes an AI assistant over company data trustworthy, which makes the analytics engineer more useful, not less.
AI in the workflow. Models can draft SQL, write model documentation and suggest tests. The analytics engineer's job is to review that output as they would a junior colleague's, because a wrong join that runs without error is the most expensive kind of mistake in this work.
In these ads, AI is not yet a headline requirement for the role: of the 13, one mentioned LLMs and three mentioned agents. The core ask is still SQL, dbt and the warehouse.
How do you become an analytics engineer?
The usual route is from analysis, and the gap is engineering habit rather than new maths:
- From data analyst: you already write SQL and know what the business asks. Add dbt, testing, version control and a warehouse, and learn to think in reusable models rather than one-off queries.
- From finance or operations reporting: you already know what "reconciles to the books" means, which many engineers do not. Add SQL depth and dbt.
- From software or data engineering: you have the engineering habits; the gap is dimensional modelling and the patience to agree definitions with people.
If you are starting from zero, begin with SQL and stay there longer than feels necessary. Window functions, joins that do not duplicate rows and reading a query plan are the difference between an analyst and an analytics engineer in an interview.
How do you prove you can do the job?
With a small but real analytics project in a public repository: raw data loaded into a warehouse, a dbt project with staging and mart layers, a test on every model, documentation generated from it, and one metric (monthly recurring revenue is a good one) defined, tested and reconciled to a source total. Add a scheduled run and a freshness check, and write a short README explaining one definition you had to decide and why. That is closer to the job than any certificate, and an interviewer can open it.
Where do you start this week?
Take a public e-commerce dataset, load it into a free warehouse tier or a local database, and write three models: customers, orders and a daily revenue table. Add two tests to each. Then write down, in one paragraph, what "revenue" means in your model and which rows you excluded. If the SQL was the hard part, spend a month on SQL. If the definitions were the hard part, you are thinking like an analytics engineer already.
Which Square 1 programme fits?
The Analytics Engineer Bootcamp is twelve weeks, live on Zoom with one instructor, about 15 hours a week. Its six blocks each end in a deployed project and a gate: a reporting layer that is correct within a time budget, a tested and documented model layer, a metrics layer that reconciles revenue to the cent, an AI assistant that cites the metric definition it used, the analytics platform with monitoring and a change process, and an employer brief in the final block alongside a hiring sprint. The recorded viva asks you to defend a number all the way to its source. The entry bar is some SQL and spreadsheet fluency. SQL and Data for AI is an on-demand course, recorded by an instructor and graded by Nova, the AI tutor, if you want the SQL first; the Data Engineering with AI Bootcamp is the route if pipelines and platforms interest you more than metrics. All are taking waitlist places today. The free data science skill check takes about three minutes if you want to see where you stand.