CDP architecture guide
Composable or enterprise CDP? Start with the operating problem.
The best CDP is not the one with the longest feature list. It is the one the organisation can implement, operate and connect to a valuable customer decision.
The architecture choice follows the use case
A large enterprise CDP can be the right answer when a business needs streaming personalisation across owned channels, extensive journey orchestration and hundreds of prebuilt connectors.
A composable approach can be stronger when the data warehouse is already central, paid-media activation is the immediate priority and the organisation wants to keep ownership of its architecture and logic.
When enterprise breadth earns its cost
- Real-time personalisation across web, app, service and owned messaging is essential.
- A broad catalogue of prebuilt source and destination connectors will genuinely be used.
- Dedicated data engineering and marketing-operations teams will own the platform.
- Global procurement, certification and governance requirements dominate the decision.
- The CDP is part of a wider customer-experience architecture.
When a focused composable model is stronger
- Trusted customer data already lives in CRM, cloud or warehouse systems.
- The first decisions are suppression, lifecycle activation or better acquisition seeds.
- The business needs paid-media and clean-room destinations more than journey orchestration.
- A managed operating model would remove a capability gap.
- Clear scope, implementation speed and transparent cost matter.
The operating model is part of the product
A CDP only creates value while someone is maintaining identity rules, audience definitions, permissions, destinations and measurement. A broad licence without an operating team can become expensive shelfware.
Fractyl can provide that operation as a managed service, or be licensed to the agency or client team already accountable for the work.
Questions to use in a shortlist
- Which customer decision must improve first?
- Which sources and destinations are genuinely required?
- Who will own identity and audience logic after launch?
- What data must remain in the current architecture?
- How will the first use case be measured?
- What is the three-year licence, implementation and operating cost?