Reference architecture layers
- Collection — Web SDK, Mobile SDK, server APIs and source connectors.
- Foundation — XDM schemas, datasets, Identity Service, Real-Time Customer Profile.
- Intelligence — segmentation, Query Service and analytics in CJA.
- Activation — Real-Time CDP destinations, Journey Optimizer, Target and partner tools.
- Governance — sandboxes, roles, labels, policies, consent and monitoring.
Architecture best practices
- Use development and production sandboxes with a promotion process.
- Agree the identity strategy before building schemas.
- Apply least-privilege access and audit changes.
- Design for consent from the start rather than retrofitting it.
- Plan capacity: profile counts, event volumes and audience limits.
- Keep secrets for APIs in a secure vault; never in code or tag managers.
Enterprise integration patterns
Common patterns include CRM-to-AEP batch syncs, event streaming from Kafka or cloud pub/sub into the streaming API, warehouse integration with Snowflake, BigQuery or Databricks, and server-to-server event forwarding. See data engineering for pipeline design.
Who this guide is for
This guide is written for beginners, students and career switchers who want a clear mental model, and for developers, data engineers, implementation specialists, consultants, architects, marketers, agencies and enterprise teams who need practical, step-by-step detail they can apply on real projects.
Frequently asked questions
How many sandboxes should we use?
At minimum one production and one development sandbox. Larger programmes add sandboxes for testing, regions or business units, depending on licensing.
Last updated: