Change data capture (CDC) is a widely used technique to connect database tables to data streams but has drawbacks related to exposing internal data models to downstream consumers, which can lead to system failures. The evolution of this pattern involves using first-class data products and data contracts to decouple internal models from external data products. This approach allows for the creation of reliable data streams that can be consumed by various applications, whether operational or analytical. A data product formalizes responsibilities and includes a data contract, which defines schema, metadata, and dedicated ownership, ensuring the data remains trustworthy and easy to use. The post discusses different techniques for building such data products, including the outbox pattern and utilizing Apache Flink SQL to handle data from multiple sources. Confluent's Data Portal facilitates the discovery and management of streaming data products, enhancing collaboration and data governance. The post emphasizes the benefits of a stream-first approach to data products, which enables both real-time and batch processing while maintaining high data quality and interoperability.