What a CDP does
In a complete data stack, customer information arrives from dozens of sources: web events, mobile app, email engagement, transactional purchases, support tickets, CRM updates. Each source has its own identifier and its own format.
A CDP unifies these into a single profile per customer. The profile becomes the queryable substrate for:
- Audience building, “all customers who purchased in the last 30 days and opened the last newsletter”
- Lifecycle triggers, “send win-back email when a customer hasn’t purchased in 60 days”
- Ad platform syncing, push audience segments to Meta Custom Audiences, Google Customer Match
- Personalisation, surface the right product on the homepage based on past behavior
- Analytics enrichment, feed unified profiles into BI tools and AI agents
Without a CDP, this work fragments. Email lives in Klaviyo, ads live in Meta, web events live in GA4, transactions live in Shopify, and nothing reconciles.
Traditional vs composable CDP
Two architectures coexist in 2026:
Traditional (“packaged”) CDP, Segment, mParticle, Tealium. The CDP owns the data: receives events, stores profiles, activates them.
Composable CDP, Hightouch, Census, Rudderstack (in its “reverse ETL” mode). The data lives in your warehouse. The CDP layer just orchestrates activation FROM the warehouse to destination tools.
Composable has won most new deployments because:
- Data stays in one place (the warehouse), eliminating sync drift
- Easier to integrate with semantic layers and AI agents
- Better privacy posture (less data movement)
- Lower vendor lock-in
A new stack in 2026 typically uses composable. Existing stacks often run traditional + composable side-by-side during migration.
What a CDP is NOT
- Not a CRM. CRMs (Salesforce, HubSpot) are sales-team tools for managing relationships. CDPs are infrastructure for customer-data activation across all channels.
- Not a data warehouse. Warehouses store and query. CDPs activate (push the data into other tools).
- Not an analytics tool. CDPs feed analytics. They’re not where dashboards live.
The roles complement. They don’t replace each other.
CDP and identity
Identity resolution is the CDP’s central job. The same person might appear as:
- An anonymous cookie ID on the website
- A hashed email in Klaviyo
- A logged-in user ID in the app
- A customer record in Shopify
The CDP stitches these into one profile. Without that stitching, “your customers” is several different lists in several different tools, and every campaign you run is missing context.
Common mistakes
- Treating the CDP as a data dumping ground. Without an activation use case, the CDP is just extra plumbing.
- Skipping identity resolution rigor. Loose matching (e.g. matching on first name) creates merged profiles that aren’t actually the same person.
- Deploying a CDP before you have first-party tracking. Garbage in, garbage out.
FAQ about CDP (Customer Data Platform)
What is a CDP (Customer Data Platform)?
A CDP unifies customer data from every source (web, mobile, email, CRM, transactions) into a single queryable profile per person. It is the infrastructure behind audience building, lifecycle triggers, and ad platform syncing.
What is the difference between a CDP and a CRM?
CRMs are human-operated systems for sales and support teams to manage individual relationships. CDPs are infrastructure for activating customer data across all channels. They complement, not replace, each other.
Traditional vs composable CDP, which should I choose?
For new stacks in 2026, composable (Hightouch, Census, Rudderstack) usually wins. The data lives in your warehouse. The CDP just orchestrates activation. Less sync drift, better AI integration, lower vendor lock-in.