How to become an Analytics Engineer
Most analytics engineers were data analysts who got tired of rewriting the same messy query every Monday. The job sits between the data engineers who land raw data in the warehouse and the analysts who build dashboards on top of it. You own the clean, tested tables in the middle, and everybody downstream trusts your numbers or doesn't.
The layer you'd own
Raw data arrives in Snowflake, BigQuery, Redshift or Databricks through a loader like Fivetran or Airbyte. It's a mess. Order records get updated in place, customer IDs change format halfway through the history, and the billing system and the product database disagree about what counts as an active account. Your job is to turn that into a small set of well-named tables that answer the business's questions the same way every time.
Most of the work happens in dbt. You write SQL models, stack them into staging, intermediate and mart layers, add tests for uniqueness and nulls, and document every column so an analyst in finance can use it without pinging you. Changes go through git, a pull request and a CI run, the same way a software engineer ships code. That's the real difference from an analyst role. You're held to engineering habits, not just correct answers.
The part people don't expect is how much of it is arguing about definitions. Revenue, churn and active user each have several versions inside most companies, and somebody has to pick one, write it down and defend it in a meeting with the head of sales. That somebody is usually you. If you like settling those fights with a clear model and a paragraph of documentation, you'll like this job. If you'd rather just answer the question in front of you and move on, stay an analyst.
The routes people actually take
Grow out of a data analyst seat
This is the route most people take. Start by moving your own recurring queries into dbt at your current job, even if nobody asked. Add tests, write the docs, open pull requests. After a few months you've got a real project to point to, and internal moves into analytics engineering are far easier than cold applications because your manager has watched you do the work.
Shift over from data engineering
Data engineers already know pipelines, orchestration in Airflow or Dagster, and warehouse performance. What they usually lack is business context. Spend time with the analysts, learn how the sales and finance teams read a dashboard, and practise dimensional modelling with Kimball's fact and dimension tables. Pitch yourself as the engineer who also knows what the numbers mean.
Come in from a BI developer role
If you've built Looker, Tableau or Power BI dashboards and fought with the semantic layer, you're closer than you think. LookML and dbt share the same logic of defining things once. Move your calculated fields upstream into the warehouse, learn git properly, and describe the switch as moving business logic out of the dashboard tool and into tested code.
Build a public project from scratch
Without a data job, this is the slow route, but it does work. Load a public dataset into a free warehouse tier, build a full dbt project with staging and mart layers, add tests and a docs site, and put it on GitHub with a short README explaining your modelling choices. Hiring managers read the README more carefully than the SQL.
Words that show up in analytics engineer postings
Applicant tracking systems and recruiters both scan for these exact terms, so if you've done the work, use their words rather than your own paraphrase.
Turning analyst work into an engineering résumé line
Created SQL queries and dashboards for the marketing team.
Rebuilt marketing attribution as 14 tested dbt models in Snowflake, replacing 40 ad hoc queries and cutting the weekly report refresh from 3 hours to 12 minutes.
What wears analytics engineers down
Nobody notices your work when it's right. They notice first thing in the morning when the revenue dashboard shows a gap because an upstream team renamed a column overnight. You'll spend a lot of time on that kind of break, and it feels thankless. Source freshness checks and contracts with the engineering teams help, but they never make it go away.
The other grind is the request queue. Once people learn you can add a field to the mart, they'll ask for one every day. Saying no, or saying not yet, is part of the job. The analytics engineers who last are the ones who push back on one-off columns and keep the models small enough that someone else could maintain them.
What Analytics Engineer postings ask for
Hiring the most
- Boeing109
- Anduril Industries22
- blueorigin.wd5/BlueOrigin12
- boeing.wd1/EXTERNAL_CAREERS10
- aero.wd5/External9
Remote
8% of openings are fully remote.
Posted pay
$151,000 – $222,495
Typical range in the 28 of the newest 60 postings that list pay.
Questions people ask
Do I need a computer science degree to become an analytics engineer?
No. Plenty come from economics, business, science or no degree at all, usually by way of an analyst job. What hiring managers check is whether you write clean SQL, understand data modelling and work comfortably in git. A degree helps you get the first analyst job, not this one.
How much Python does an analytics engineer need?
Less than a data engineer. SQL carries most of the job. You'll want enough Python to write a small script, read an Airflow DAG, or use a dbt macro that calls out to it. Some teams lean heavier, so read the posting, but strong SQL beats average Python every time.
Is the dbt certification worth getting?
It's a decent signal if you're coming in cold with no dbt project at work, since it shows you've learned the tool's conventions. If you already have a public project or real dbt work to show, spend the time on that instead. A good repo is more convincing than a badge.
Ready to apply?
Tell HeroApply you want Analytics Engineer roles. It finds the openings and applies for you each day.