Most organisations already collect plenty of data: sales from a till or online shop, customers in a CRM, costs in accounting software, and a few important spreadsheets someone updates by hand. The problem is rarely a lack of data. It is that the data sits in different places, in different shapes, and nobody fully trusts the numbers.
An analytics engineer fixes that. The role sits between raw data and the people who make decisions, turning messy sources into clean, tested, documented data that reports, dashboards and data science can be built on.
What an analytics engineer actually does
- Brings data together
- From different systems into one place, usually a cloud warehouse such as Snowflake or BigQuery.
- Models it
- Into facts and dimensions, so "revenue", "customer" and "order" mean the same thing everywhere.
- Tests it
- So errors are caught before anyone makes a decision on bad numbers.
- Automates it
- So data refreshes on a schedule without copying and pasting.
- Documents it
- So the business is not dependent on one person's memory.
Analytics engineer, data analyst or data engineer?
Data engineer
Pipes- Main focus
- infrastructure and moving large volumes of data reliably
- Typical output
- ingestion pipelines, platforms
- Python, Spark, orchestration, cloud services
Analytics engineer
Models- Main focus
- transforming data into trustworthy, reusable models
- Typical output
- data models, tests, documentation, metrics
- SQL, dbt, data modelling, testing
Data analyst
Answers- Main focus
- answering business questions with data
- Typical output
- analyses, dashboards, recommendations
- SQL, BI tools, statistics, storytelling
In smaller organisations, one person often covers all three, and the lines between the roles are blurring as AI tools take on more of the routine work.
The core skills
- SQL
- Joins, window functions and CTEs. Start with the window functions guide.
- Data modelling
- Especially the star schema.
- A transformation framework
- Such as dbt.
- A cloud warehouse
- Such as Snowflake.
- Python
- For automation and the tasks SQL is not suited to.
- BI tools
- Such as Power BI, and enough DAX to build reliable measures.
- Software habits
- Git, code review, testing and CI.
- Communication
- Translating business questions into data definitions, and back again.
Signs your organisation needs one
- Reports take hours to prepare and are often late.
- Different teams quote different numbers for the same metric.
- Only one person knows how the key spreadsheet or query works.
- You have data in several tools but no single view of the business.
- Dashboards are built, then abandoned because nobody trusts them.
What a first project looks like
- 1
Start from the decisions
Agree which decisions the data must support, and list the data that already exists.
- 2
Model the key sources
Build a simple, documented model that connects them, with one agreed definition per metric.
- 3
Add tests
Check keys, missing values and freshness on every run.
- 4
Automate the reporting
Put Power BI, Excel or Google Sheets on top, refreshing on a schedule.
The result is a set of numbers that update themselves and that everyone agrees on.
Check your understanding
1/3Which role focuses on turning raw data into trustworthy, reusable models?
Data engineers move data, analysts answer questions; analytics engineers model, test and document the data in between.
2/3Which of these is a typical sign an organisation needs analytics engineering?
Conflicting numbers usually mean definitions live in many places instead of one tested model.
3/3Which skill is most central to the role?
SQL and modelling are the core; Python, BI tools and Git complete the toolkit.
You got 0 out of 3.
Ready to build the skills? Start with your first dbt project and data quality tests every pipeline needs, then practise with the interview prep.
Alessandro Ecclesie AgazziWritten by Alessandro Ecclesie Agazzi, analytics engineer in London. Updated 30 September 2026. About me