A customer data platform usually gets bought before anyone has agreed on the problem it should solve. That is the pattern I have taken away from every selection process I have been part of. The opening question is almost always which tool to pick. The question that should replace it sounds different: which decision can we not make today because the data sits in five places and none of it lines up.
Start with the second question and you often end up needing a smaller system than expected. Start with the first and you buy features nobody ever switches on.
- Selection starts with use cases, not with a vendor comparison.
- Without a shared definition of customer, a customer data platform only unifies the contradiction.
- Consent belongs in the architecture, not in the final sign-off.
- A narrow first slice produces a showable result faster than a full rollout.
- Sometimes the honest answer is that no separate platform is needed.
Table of contents
Why a customer data platform rarely fails on the tooling
The systems on the market differ less than the vendor decks suggest. Persistent profiles, identity resolution, an interface that lets other tools read the result: everyone does that now. The institute behind the term describes a customer data platform as software that creates and maintains a persistent, unified customer record accessible to other systems. It is a narrow definition, and the narrowness is the point.
The projects I watched from close up never failed on technology. They failed because nobody had decided in advance what counts as a customer. Is it the email address, the account number from the stores, the device in the shop. As long as three departments give three answers, even the best customer data platform produces nothing but a neatly unified contradiction.
The second thing is ownership. Someone always owns the system, and when nobody is named, it becomes IT by default. That is not wrong, but it means maintenance gets prioritised by capacity rather than by what marketing needs next.
What actually carries a customer data platform decision
These days I start with use cases, and with no more than three. Not a list of thirty to choose from later, but the two or three things that should work better next quarter. A welcome flow that knows whether someone already bought in a store. A campaign that stops paying to reach existing customers. A report that does not count the same channel twice.
Those cases tell you which sources genuinely need connecting. It is usually fewer than feared. And the sources tell you which data quality problems have to be fixed first, because a platform does not improve data. It only makes the flaws visible faster.
The third point is consent. In Europe the legal basis is not a footnote, it is the architecture. If permission does not travel with the record, a technically perfect profile is useless, because you are not allowed to activate it. The ground rules sit in Article 6 of the GDPR, and they belong in the selection, not in the final sign-off.
The moment the effort starts paying back
The value does not show up on launch day. It shows up the first time someone asks a question that used to take two weeks of coordination, and the answer arrives that afternoon. Getting there takes longer than any proposal admits.
What shortens the path is a narrow first slice. One market, one channel, one use case. Next to what is technically possible that sounds modest, but it produces an early result you can show. Visibility is the currency that buys the budget for step two.
Proving the value afterwards is its own discipline. If you want a method for that, my piece on marketing mix modelling offers a more honest frame than any dashboard.
What I would do differently
I would decide on the customer data platform later and start the work on definitions earlier. What a customer is, how consent is recorded, who owns the result. Those are not preliminaries. They are the project.
And I would keep one question open: whether a separate platform is needed at all. If the warehouse already exists and activation runs through a handful of channels, a customer data platform is sometimes the expensive answer to a problem a well defined table also solves. Asking costs nothing and occasionally saves a great deal.