Thoughts and rants from a veteran Martech contractor and interviews with my peers.

Did CEPs just hand the data layer to Databricks?

Three engagement platforms made the same bet in the same week. Whether it pays off is more contingent than the announcements suggest.

In 2019, I had a client running both Segment and Iterable, and the recurring question was where to build the audiences. Segment could build them. It could not act on them. Only Iterable could turn an audience into a journey, so that is where the audiences got built. Segmentation followed the system that could do something with it. Segment lost that job, and before long it was replaced by Hightouch. Sound familiar?

I think about that sequence whenever the industry decides it has finally settled where customer data belongs, because the answer never stays put.

The latest move arrived in

composable cdp

Composable CDP, translated for marketers

Composable CDP hands marketers a finished system and walks away. Four questions to ask before the architecture decision is locked in.

Let's face it, most composable CDP conversations are written for engineers, which makes sense, because engineers are the ones building the thing. The cost of that arrives later, and it arrives for the marketers who run the campaigns on top of the architecture, handed a finished system and told to make it work. I've written before about the inherited costs that creates. This piece is about the part that comes earlier, when the architecture is still a decision and not yet a fact.

The question a marketer should be able to ask in that room is plain. What

customer data platform

Databricks CustomerLake: the data lakehouse is now the CDP

Databricks launched CustomerLake, a lakehouse-native agentic CDP. What it means for composable vendors, Databricks customers, and the next data platform.

Databricks announced CustomerLake at its Data + AI Summit just now, and the rumor is confirmed: it is a customer data platform. A real one. In 2026 no less. Marketer-facing, built natively inside the lakehouse, with agents doing the work that used to need a data team and a few weeks of lead time. The first agentic CDP a lakehouse has built for itself.

Only 16 people guessed right on my LinkedIn poll.

To be fair, I saw it before the keynote. Databricks invited me to a preview and asked what I made of it, which I took as a

customer data platform

A little lift never hurt anyone

Hightouch, Zeotap and mParticle now sell a toggle that lifts your match rate with identifiers you never collected. The gain is real. The work is the consent check nobody does before flipping it. What it does, and what to look at first.

You spend a week defining an audience in your CDP. Lapsed high-value customers, the ones actually worth winning back. You push the segment to Meta feeling good about it. Then the matched figure comes back at around sixty percent of what you sent, and the rest just isn't there. The platform couldn't recognise four in ten of your own customers.

Anyone who has run paid media knows that feeling. It's the match rate gap, and it has a few causes. People sign up with one email and log into Meta with another. Phones get replaced every

composable cdp

The CDP you weren't supposed to buy

Hightouch said the packaged CDP was dead weight. Four years on, Gartner lists it in the CDP Magic Quadrant. The architecture was right. The pitch skipped a step.

In 2022, Hightouch published a piece with a headline built to draw attention: Friends Don't Let Friends Buy a CDP. I remember where I first saw it, neck deep in my client's packaged CDP project. Sure, it was vendor content. Everyone knew it was vendor content. Nevertheless, it got passed around, because the frustration underneath it was real, and most of us had heard it, or even felt it.

The argument was simple enough to fit on a t-shirt. Your data already lives in the warehouse. Why pay a second vendor to copy it into a separate

The Context Audit

A follow-up to the follow-up. My last piece explaining "what the context layers has to do" argued that customer state has to be Queryable, Addressable, Current, and Accountable if agents are going to act on it in real time. It ended pointing at a one-page audit and stopped there.

This is that audit.

The context layer's job is to make the customer state usable when an agent needs to act on it. Whether it's doing that job is not a question anyone can answer from an architecture diagram. Architecture diagrams describe

Latest Posts