What is Customer Data Platform?

A customer data platform (CDP) solves one problem: your customer data lives in 20 places and you need it in one place so you can act on it. The CDP ingests data from product analytics, CRM, billing, support, marketing automation, and email tools. It resolves identities across those sources into unified profiles. Then it pushes those profiles out to wherever you need them: ad platforms, email tools, sales tools, dashboards.

The category has three sub-types worth knowing. Segment (now Twilio Segment) is the original event-based CDP that pipes user behavior events to downstream tools via SDKs and APIs. Rudderstack is the open-source alternative with similar architecture. The second sub-type is composable CDP, a label for using your data warehouse (Snowflake, BigQuery) as the storage layer with reverse ETL tools (Hightouch, Census) handling the activation. The third sub-type is operational CDPs like ActionIQ, Lytics, and Tealium, which package storage and activation together for enterprise marketing teams.

For GTM Engineers, the composable CDP pattern has become the default architecture in 2025-2026. You already have a warehouse for analytics. Adding a reverse ETL tool on top gives you 80% of a traditional CDP at 20% of the cost, with full SQL flexibility and no vendor lock-in. The catch: you need a competent data team to maintain the warehouse and the dbt models that build customer profiles. If your data is messy, a composable CDP makes the mess available faster, not cleaner.

Where Segment still wins is event collection at scale. Segment's tracking libraries handle hundreds of millions of events per day with reliability that a self-rolled solution rarely matches. For PLG companies whose growth model depends on capturing every product event, paying $10K-$50K/year for Segment to handle event collection is usually cheaper than building it in-house.

The biggest CDP mistake is buying one before you know what you'd do with unified customer data. A CDP is a delivery mechanism. If your marketing team doesn't have specific use cases (route product-qualified leads to sales within an hour, suppress retargeting ads for customers who already churned, score accounts for expansion likelihood), the CDP becomes another expensive system that ingests data and produces dashboards nobody acts on.

GTM Engineers should evaluate CDPs against three criteria. First, does it solve a use case your business will act on within 90 days of deployment? Second, does it integrate cleanly with both your data sources (CRM, analytics, billing) and your activation tools (ad platforms, sales tools, email)? Third, does the pricing model scale with the value you'll extract, or does it inflate as your event volume grows? Vendors that price on profiles or destinations tend to be more predictable than vendors that price on events or rows of data.