Data modelling with dbt
Datalyze turns the raw tables in your warehouse into clean, documented data models built in dbt. We decide what needs modelling, plan the marts around the questions your team asks, and add a semantic layer so people and AI agents get the same answer.
What data modelling involves
- 01Scope
Decide what data needs to be modelled
Not every table deserves a model. We go through your sources and the questions each team needs answered, then agree which data gets modelled.
- 02Plan
Plan the marts and check they cover every use case
Before writing any SQL, we map out every mart and data model, what each one holds, and which questions it serves.
- 03dbt
Build the models in dbt
We build the models in dbt, from staging to marts, with tests, documentation and version control. The logic lives in one place everyone can read, review and change.
- 04Semantic layer
Create a semantic layer
On top of the models we add the context a table name can't carry: what each metric means, the table schema, example queries and the rules behind each definition.
You get a warehouse your team can query without asking an engineer, and definitions every tool and agent agrees on.
What teams say after working with us
Data we've modelled
Buildern couldn't trust its attribution or see which accounts were at risk. We modelled the warehouse in dbt, fixed attribution on top of it, and built an account health score the team now uses to spot churn early.
Read the case studyTermPlus had a half-finished setup and data spread across separate systems. We built the full analytics infrastructure, modelled every source around one user identity, and built the dashboards the team runs on.
Read the case studyThese are real results from teams like yours.
Book a callData modelling questions
How long does data modelling take?
Four to six weeks to a modelled warehouse. We agree what to model and plan the marts first, then build them in dbt and add the semantic layer at the end.
How much does data modelling cost?
It is priced as a project, scoped by how many sources and marts you need. We can also stay on as a monthly retainer to maintain the models as your product changes.
What do you need from us?
Access to your data warehouse and Git, and time with each team that will use the data, so the marts answer their real questions.
Which warehouses and tools do you work with?
We build models in dbt on BigQuery, Snowflake, Databricks or Postgres. If your data is not in a warehouse yet, we set that up first as part of data unification.
What is a semantic layer?
A semantic layer is the written context on top of your models: what each metric means, how it is calculated, how tables join and example queries. It is what makes people, dashboards and AI agents get the same answer.