What a CDP actually is, and when it is worth having one
CDP became a sales term before it became an understood category. This piece covers what a customer data platform really does, what it does not, and the signal that tells you it is time to have one.
CDP stands for Customer Data Platform. The short definition: a system that collects events from all of your sources, works out which events belong to the same person, and makes that result available for analysis and for activation in other tools.
The definition is simple. What makes it confusing is that nearly every marketing tool now describes itself with similar words. Worth separating.
What a CDP does that the others do not
Three capabilities define the category. A tool missing any of the three is something else — which does not make it worse, only different.
- 01Unified collection: a single entry point for events from the site, the checkout, the CRM, email and ad platforms. The event is defined once, not once per tool.
- 02Identity resolution: the ability to assert that Tuesday’s anonymous visit, Wednesday’s form fill and the following Sunday’s purchase are the same person. This is the hard part, and it is what separates a CDP from an analytics tool.
- 03Activation: pushing the result back into the systems where decisions happen — the ad platform, the CRM, the email automation.
A tool that collects and charts but cannot tie two sessions to one person, and sends nothing anywhere, is an analytics tool. That is not a criticism — it is a different scope.
What a CDP is not
| Category | Question it answers | Where it stops |
|---|---|---|
| Web analytics (GA4) | How many people came, from where, and what they did | Works in aggregates; cannot act on an individual |
| Product analytics (Mixpanel, Amplitude) | How people use my product | Does not integrate sales and media, and sends nothing back |
| CRM (HubSpot, Pipedrive) | What stage is this deal in | Records what the rep types, not anonymous behavior |
| Email automation | Who received, opened and clicked | Only sees its own channel |
| CDP | Who is this person, where did they go, and what now | Does not replace email sending or the checkout |
The signal that it is time
It is not revenue, and it is not tool count. The signal is one specific question you cannot answer in under a day:
"For yesterday’s sale, which ad brought that person — and how long passed between the click and the purchase?"
If answering that means exporting spreadsheets from three places and matching on email by hand, you are already paying the cost — in team hours and in campaigns optimized on incomplete data. A CDP does not conjure that answer out of nothing; it stops destroying information that already exists along the way.
The three most expensive mistakes when adopting one
1. Choosing by catalog size
A 400-integration catalog is impressive on a pricing page and changes nothing if you use six tools. What matters is the quality of the ones you will actually connect: whether the checkout integration deduplicates sales, whether it brings payment status, whether it distinguishes a first purchase from a renewal.
2. Treating identity as a technical detail
A badly designed identity rule produces two opposite and equally bad defects. Too loose and two people sharing a computer become one profile, making the history meaningless. Too strict and the same person becomes four visitors, so the sale is never explained. The rule that works is conservative: an anonymous session never merges into an already-identified lead just because they share a browser.
3. Stopping at collection
Plenty of data projects die with the data beautiful and idle. Collecting everything and returning nothing to the campaign is cost without return. Activation is where the investment pays for itself — and it should be the first thing you switch on, not the last.
CDP and BI: why they are usually separate
Historically the CDP moves data and a BI tool visualizes it. That makes sense in a large company with a warehouse and a dedicated team. In a lean operation the practical result is that you buy two subscriptions, integrate them, and still need someone to maintain the bridge — while the question you had on launch day stays unanswered, because it lived precisely on that bridge.
That is why CrazyLeads merges the layers: sources become queryable tables in the same platform, and a query result becomes a publishable page. It is not a conceptual innovation — it is the removal of an integration step that only existed by historical accident.