Create a Data Product
There are three ways to create a data product. Pick the one that matches what you already have and how hands-on you want to be.
| Path | Best when… | Trade-off |
|---|---|---|
| Create data product manually (Contract First) | You are modelling a new or planned data product, or you have no connector yet. | Most control, most typing. You describe the data product by hand. |
| Import asset as data product (Data First) | The data already exists in a connected platform (Snowflake, Databricks, BigQuery, …). | Fastest start. Generates an output port and a draft data contract from the asset, which you then refine. |
| Build with an AI agent (Code First) Preview | Your team builds data products as code (dbt, Databricks) and uses a coding agent. | Generates the pipeline and the metadata from a data contract, but needs a configured Data Product Builder. |
All three produce the same thing: a data product with one or more output ports, each specified by a data contract. You can switch approaches later — for example, start manually and let an agent implement the pipeline.
Recommended for your first one: create from an asset if you have a connector, otherwise create manually.