Best Practices

There are two proven ways to introduce data contracts with the Data Contract CLI. Pick the one that matches your situation.

Data-first approach

Create a data contract based on the actual data. This is the fastest way to get started and to get feedback from data consumers.

Use an existing physical schema (e.g. SQL DDL) as a starting point to define your logical data model in the contract. Right after the import, double-check that the actual data meets the imported model:
```
datacontract import sql --source ddl.sql
datacontract test
```
Add quality checks and additional type constraints one by one, making sure the data still adheres to the contract:
```
datacontract test
```
Validate that the datacontract.yaml is correctly formatted and adheres to the Open Data Contract Standard:
```
datacontract lint
```
Set up a CI pipeline that runs daily for continuous quality checks. Use the ci command for CI-optimized output (GitHub Actions annotations and step summary, Azure DevOps annotations). You can also report results to tools like Entropy Data:
```
datacontract ci --publish https://api.entropy-data.com/api/test-results
```
See Test your contract → Scheduling and CI/CD.

Create a data contract based on the requirements from use cases, before the data product exists.

Start with a datacontract.yaml template:
```
datacontract init
```
Create the model and quality guarantees based on your business requirements. Fill in the terms, descriptions, etc., then validate the format:
```
datacontract lint
```

Use the export functions to start building the providing data product as well as the integration into consuming data products:

# data provider
datacontract export dbt-models
# data consumer
datacontract export dbt-sources
datacontract export dbt-staging-sql

Test that your data product implementation adheres to the contract:
```
datacontract test
```