Customer Data Platform
A customer data platform stitches the records of a single customer into one identity across every system that touches them. The problem it exists to solve is older than the category: the same person is a lead in the marketing tool, a contact in the CRM, and an account number in the billing system, and nothing connects the three. Each department's numbers add up on their own and disagree with each other's. The CDP category promises to fix this with software. The join itself, deciding which records are the same person and which department's field wins when they disagree, is the work, and no license includes it.
How it actually works
Identity resolution is the core mechanism. The platform ingests records from each source system and matches them on whatever keys exist: email, phone, device, account number. Deterministic matching links records that share a key. Probabilistic matching scores likely links where no key is shared. The output is one profile per customer with every interaction, purchase, and support contact attached to it.
A revenue question such as which campaigns produce customers who renew joins marketing touch data to sales outcomes to finance's revenue records. In most companies that join does not exist. Marketing reports cost per lead from its system, sales reports pipeline from the CRM, finance reports revenue from billing, and each number is computed over a different population of records. The three departments are answering different questions and using the same words for them.
This is also where the category's failure mode sits. A CDP implemented as another marketing tool ingests the marketing sources, builds profiles for ad targeting, and never touches sales or finance data. The company then owns four unreconciled views of the customer instead of three. The platform only pays for itself when it carries the identity every department resolves to, which is an organizational agreement before it is a piece of software. Two numbers frame the purchase decision. The value of the join is the budget currently allocated on unjoined data multiplied by the misallocation rate that unjoined data produces, and that product is zero for a CDP that never carries sales and finance records. The total cost of ownership runs well past the license: implementation, the integration build, and the standing stewardship of match rules and field precedence are recurring costs the license line does not show, the same pattern that holds across the revenue-operations stack.
In practice
The Potbelly work depended on this join existing. Finding five times more efficient spend at a level budget required connecting what marketing spent to what the business collected, across systems that did not share a customer record. The pattern repeats in most diagnostics we run: teams optimizing each system in isolation, because the join that would show the whole picture was never built. The enrichment data feeding the profiles carries its own error rate. We have used Demandbase, the top tier in B2B and ABM, and helped clients implement 6Sense, Apollo, and Audience Labs directly into their CDP or CRM; they support re-marketing and customer segmentation when they are accurate. In one of our old offices, the IP address was categorized in Demandbase as The University of Utah, because the university had a satellite campus in the same parking lot. That noise informs conversion rates, channel referrals, split tests, and which ads get the credit ([bot and AI traffic](/insights/marketing-measurement#bot-and-ai-traffic)).
Where we come in
We build the identity layer inside our Revenue Systems Architecture practice: one pipeline from your data sources to the decisions your team makes, for businesses from new startups to the Fortune 15, including Digital Realty and Equinix. The Potbelly engagement ran on this kind of architecture work before a dollar of budget was added. We demonstrate ROI before each build, and the systems are portable, so the identity you assemble stays with you if you change vendors.
Start a Revenue Health Pre-Assessment →See it in action
Related terms
- Identity resolution
- Matching records across systems to a single customer. Deterministic matching uses shared keys; probabilistic matching scores likely links without them.
- Single source of truth
- The one agreed record a value resolves to. Without it, each department computes its own version, and reconciliation happens only when a discrepancy forces it.
- First-party data
- Data your customers gave you directly. As third-party tracking degrades, it is the raw material the whole platform runs on.
- Customer 360
- The assembled profile: every interaction, purchase, and support contact for one customer in one place. The value comes from the joins between systems rather than from the profile view itself.
- Data silo
- A system whose records never reconcile with the rest of the business. Every department tool becomes one by default.