What should a CDP talk to - and why it’s time to rethink the conversation...

The term “CDP” gets tossed around like sweeties at a trade show, but most Marketing Operations teams are still chewing over the basics: What should our CDP actually talk to? And just as critically - why?

At a glance, this seems straightforward: plug your CDP into your MarTech stack, ingest some data, spit out some personalisation, and boom - instant ROI. But scratch the surface and it gets complex. Because not every connection is valuable. Not every data source deserves a seat at the table. And not every tool in your stack should be whispering sweet nothings into your CDP.

Let’s unpack the real questions behind CDP-to-MarTech integration - and why getting this right is the difference between customer intelligence and expensive confusion.

Not all tools deserve a relationship with your CDP

Your CDP is not a dumping ground. It’s not there to hoard data like some digital dragon. Its job is to unify, segment, and activate person-level data that drives outcomes.

So ask yourself: does this system contain identifiable, actionable data that can be used to segment or personalise? If the answer is “no,” it doesn’t belong. Aggregated analytics? Leave it out. Anonymised behavioural data with no linkage to a known profile?

Thanks, but no thanks.

This isn’t about data volume. It’s about data purpose.

Understanding the three core CDP touchpoints

Let’s get into the mechanics. A CDP should typically be doing one (or more) of these three things with your MarTech stack:

• Push data out to marketing platforms (email, adtech, mobile push, etc.)

• Respond to data requests from those systems when they need profile enrichment or audience segmentation

• Receive data from MarTech tools - either via push (e.g., webhook, stream) or pull (scheduled syncs)

Every integration should serve a use case. Too many CDPs get bloated with unnecessary connections “just in case.” That’s not integration; that’s insecurity.

Systems of record vs. systems of confusion

Not all data is created equal. Your CDP should integrate with systems of record - places where critical customer truth lives. Think CRM, transactional platforms, and modern digital behaviour trackers with identifiable signals.

What it should not be doing is playing cleanup for five-year-old legacy platforms spitting out unstructured noise. If a system can’t validate its own data quality or doesn’t know who a customer is, it’s not your CDP’s job to guess.

Remember: garbage in = creepy personalisation out.

Build for feedback loops, not one-way pipelines

A healthy CDP architecture is built on feedback. This means validating that data sent from Point A landed accurately at Point B. It means enriching profiles and getting signals back from the activation layer to improve targeting next time.

Too many stacks have a one-way “fire-and-forget” mentality. That’s how you end up with personalisation strategies based on data that’s six months old and wrong.

Make it a loop. Automate the sanity checks. Measure the value of every integration. Otherwise, you’re just wiring pipes in the dark.

Batch vs. real-time: Choose the hill you want to die on

Real-time is the sexy thing to put on a slide. “We’re real-time,” says every vendor ever. But is it always worth it?

Not necessarily. For campaign prep, onboarding flows, or audience planning, batch is often faster, cheaper, and more reliable. Save real-time for use cases that need it: triggered messages, fraud detection, cart recovery.

Real-time is not a religion. It’s a tactic. Use it wisely.

AI: Finally earning its keep?

AI’s been loitering around the CDP party for years, sipping punch and offering vague promises. But it's starting to justify its invite.

Where AI adds real value now:

• When to communicate: Optimal timing based on intent signals and behavioral patterns

• Where to communicate: Channel selection driven by engagement likelihood

• What to communicate: Personalized messaging, subject lines, offers—based on profile data and predictive models

But here’s the kicker: AI is only as good as the data feeding it. A poorly integrated, overfed CDP will just serve up smarter junk. Garbage in, algorithmic garbage out.

Ease of use is the new Enterprise currency

Here’s something enterprise stacks often miss: marketing teams are not data engineers.

If your CDP integration requires a PhD and three Slack channels to operate, it’s not serving its purpose. The real value comes when a marketer can confidently pull the right audience, activate the right segment, and trust the data is both fresh and accurate.

This means clarity in data flows, intuitive interfaces, and ruthless prioritisation of integrations that actually drive campaigns forward.

So, what should a CDP talk to?

Only the tools that:

• Contain person-level data you can act on

• Are masters or sources of truth, not echoes of confusion

• Can give or receive data in a way that improves outcomes

• Support a measurable feedback loop

• Help AI do its job, not muddy the waters

• Make marketing faster, smarter, and more accountable

Everything else? Politely tell it to get in line.

Final thought: Build for purpose, not proof-of-concept

Too many CDP rollouts start with “let’s integrate everything,” instead of “let’s integrate what matters.” The result? Frankenstein stacks, overwhelmed marketers, and AI models trying to read tea leaves made of glitter.

The next wave of customer data strategies won’t be defined by how many tools you connect. It’ll be defined by how smartly, selectively, and securely you do it.

Choose quality over quantity. Choose clarity over complexity. And if you're not sure whether your CDP should be talking to a system?

Ask it this: Does this help us know the customer better or serve them smarter?

If the answer’s “meh,” unplug it.