Skip to content
NEW: Martech, explained →

Martech, explained · CDP architectures · 2:10

Composable CDP, explained

How a composable CDP fits together, and why the wiring between the tools is yours

The short answer

A composable CDP does the same four jobs as a packaged one, collect, unify, segment and activate, split into separate tools around your data warehouse. You own the data and choose each tool. The wiring between them is yours to run, and it is where things break most often, so decide who gets the call at two in the morning.

Questions this answers

How does a composable CDP fit together?

Events arrive in the warehouse through a pipeline, identity rules join the records into one customer, and a reverse ETL tool builds segments and syncs them to email, ads and the website. The results flow back, so the next segment is built on them.

What breaks in a composable CDP?

The wiring. Each line between two tools is a job on its own schedule, and a failed sync can go unnoticed until the morning send goes out on yesterday's segments.

What does a composable CDP cost to run?

Compute and syncs. A segment rebuilt every hour runs warehouse compute and a sync that often bills by the row, every time, so someone has to watch that meter.

Composable or packaged?

Packaged sells convenience, composable sells control. Either way somebody holds the complexity; in a composable CDP it is your team.

Transcript

Every chapter as text. Click a time to play the film from there.

0:00 2 a.m.

At two in the morning, the sync to your email tool fails. Nothing flags it. By nine, the morning send goes out on yesterday's segments.

0:10 Intro

That's a composable CDP failure. To see why, look at how one is built.

Same four jobs, separate tools

A composable CDP does the same four jobs as a packaged one: collect, unify, segment and activate. It just splits them into separate tools around your data warehouse, often from different vendors.

0:31 How it fits together

Events come in through a pipeline and arrive in the warehouse.

Identity rules, in the warehouse or a specialized tool, join an email, a cookie and a device into one customer.

A reverse ETL tool builds segments on top and syncs them out to email, ads and the website.

And the results flow back, so the next segment is built on them.

0:55 The wiring is yours

Every one of those lines is a job somebody runs. Lift them off the diagram and there are seven of them, each on its own schedule, and each one can fail without a sound. In a packaged CDP, the vendor kept that list out of view. Here it's yours, and it's where things break most often.

1:14 Three questions to ask

So the questions change. When a sync fails, how would you know the data behind a segment is stale before you hit send? A segment rebuilt every hour runs warehouse compute and a sync that bills by the row, every time. Who's watching that meter? And an opt-out has to reach every tool the data was copied into, not just one.

1:35 Convenience or control

None of that makes composable the wrong choice. You own the data, and you choose the tools. Packaged sells convenience, composable sells control. Either way, somebody holds the complexity.

1:49 So what is it?

What is a composable CDP, then? The same four jobs, in tools you pick, with the wiring in between as yours to run. Before you build one, decide who gets the call at two in the morning.

See each model step by step at Martech Stack Builder, or build your own.