Auto Apply to hundreds of jobs in a few hours

Español

日本語

Analytics Engineer skills: what to learn first, and what can wait

Hiring managers for this role aren't asking whether you can write SQL. They're asking whether they can hand you a pile of raw tables and get back models that nobody has to double-check. Learn in this order and you'll spend your evenings on what screens and panels actually test.

996 open jobs

Step one

Gets you the interview

SQL past the basics: window functions, CTEs and deduplicationTake-home tests love a table with duplicate events and late-arriving updates. If you can reach for ROW_NUMBER over a partition, write QUALIFY where the warehouse supports it, and keep a long query readable with named CTEs, you clear the first screen.

dbt models, sources, refs and generic testsdbt shows up in a lot of postings for this title. A public repo with a staging layer, a couple of marts, unique and not_null tests, and a filled-in schema file proves more than a line in your skills section.

Git and the pull request workflowYou'll be reviewed like a software engineer. Branching, writing a useful PR description, resolving a merge conflict without panic: interviewers ask about this because analysts coming over often haven't done it.

Working knowledge of one cloud warehousePick Snowflake, BigQuery, Redshift or Databricks and learn how it charges you, how it stores data and how it handles clustering or partitioning. Knowing why a query scanned the whole table is a strong signal.

Step two

Gets you the offer

Dimensional modelling and choosing a grainPanels hand you a messy source and ask what the fact table should be. Knowing Kimball's facts and dimensions, and saying out loud that a table is one row per order line before you write anything, is what separates you from a strong analyst.

Slowly changing dimensions and dbt snapshotsCustomers change plans and move regions. Being able to explain a type two dimension, set up a snapshot and then join to it correctly for a point-in-time report is a common whiteboard problem.

Incremental models and warehouse costTeams hire this role partly to stop the bill from creeping up. Talk through an incremental strategy, a merge key, and what you'd do when late data lands outside the lookback window.

Writing a metric definition people acceptExpect a scenario where two teams disagree on active users. The winning answer is a written definition, a single model that owns it, and a plan for telling the losing team.

Step three

Gets you promoted

A semantic layer and governed metricsMoving metric logic out of Looker, Tableau or Power BI and into the dbt Semantic Layer or LookML means every dashboard pulls the same number. Owning that migration is senior-level work.

Data contracts and source monitoringMost broken dashboards start upstream, when an app engineer renames a column. Setting up freshness checks, schema tests on sources and an agreement with the product team saves you from being the fire brigade.

Orchestration and CI for the whole projectKnowing how Airflow, Dagster or dbt Cloud jobs run, and building slim CI that only tests changed models, makes the project faster for everyone else.

Setting conventions and reviewing others' modelsA style guide for naming, a folder layout and code review that teaches rather than rewrites. When other people's models start looking like yours, you're doing the next job.

Certificates worth your time

Certificate	Best for	Effort	Worth it?
dbt Analytics Engineering Certification	Analysts moving into their first analytics engineering role	A few weekends if you already use dbt at work, longer if you don't	The one cert that maps straight to the job title. It won't replace a portfolio, but it helps a resume from an analyst background get past a recruiter.
SnowPro Core Certification	People targeting teams that run on Snowflake	A few weeks of evening study	Useful proof you understand warehouses, caching and cost. Worth it only if the postings you want mention Snowflake.
Google Cloud Professional Data Engineer	Engineers on BigQuery-heavy teams who want to lean toward pipelines	A couple of months of steady study	Broader than this role needs. Take it if you're drifting toward data engineering, skip it if you're happy in the modelling layer.

Exam content, formats and renewal rules change, so check dbt Labs, Snowflake or Google Cloud directly before you book a test.

Put it on your resume like this

Weak

Worked on dbt models and helped improve data quality for reporting.

Strong

Rebuilt 40 legacy reporting queries as 12 tested dbt models with incremental loads, cutting nightly warehouse runtime from 3 hours to 25 minutes and ending 2 conflicting revenue numbers.

Questions people ask

Do I need Python to be an analytics engineer?

Not to get hired. SQL and dbt carry most of the job. Python helps later for things like Jinja macros that get complicated, small API loaders and testing scripts, so learn it once your SQL is solid.

Which warehouse should I learn first?

Look at the postings you'd actually apply to and pick whatever shows up most. The ideas carry over between Snowflake, BigQuery and the rest, so the first one is the hard one.

How do I build a dbt portfolio without a job using it?

Load a public dataset into a free warehouse trial or DuckDB, build staging and mart layers with tests and docs, and push it to GitHub. Write a short README that explains your grain choices, because that's what reviewers read.

Is Looker or Tableau worth learning for this role?

Enough to see how your models get used. You're not hired to build dashboards, but knowing why a BI tool struggles with a wide, unmodelled table makes your marts better.

Got step one? Start applying. HeroApply matches you to roles that fit.

Start your trial