For years, we’ve been told that data is the moat.
The theory was straightforward: collect more customer data, unify it, build a better customer profile and resolve identities across channels. Put everything into a CDP, and the company with the richest customer data would have the advantage.
I’m increasingly convinced that’s no longer true.
That doesn’t mean data has become less important. Quite the opposite. Customer data is becoming so fundamental to the modern marketing stack that I think we need to stop treating it as a source of differentiation. Data is becoming infrastructure. And infrastructure, by itself, isn’t a moat.
The interesting question now isn’t who has the most data or even who has the best customer profile. It’s who makes the best decisions with that data, and who can execute those decisions most effectively. That’s where I think the next competitive moat is being built.
We spent a decade assembling the customer
The rise of the CDP made perfect sense. Customer data was scattered everywhere, with the ESP knowing one version of the customer, ecommerce another, loyalty another and mobile yet another. The CRM had still another.
Compounding the problem was the fact that legacy ESPs were built on relational databases and were unable to store unstructured data like events. So where to put that data?
Hence the CDP.
In its original form, the value proposition was compelling: ingest customer data from multiple sources, resolve identity, create a persistent customer profile, build audiences and make those audiences available to downstream systems. That fixed the problem of unifying structured and unstructured data, but it also created yet another platform that invariably added latency and friction to messaging.
So where are we now?
Cloud data platforms like Snowflake and Databricks increasingly sit at the center of enterprise customer-data architectures, while identity capabilities are appearing in more places. Reverse ETL and warehouse activation platforms can make data available to marketing applications without necessarily copying it into another persistent database.
At the same time, modern ESPs have expanded in the opposite direction. Braze, Zeta, Iterable, Bloomreach, Cordial and others are taking on more responsibility for data, intelligence, orchestration and decisioning.
The clean little boxes we used to draw around CDP, ESP, personalization, analytics and data warehouse are getting awfully fuzzy. That isn’t an accident. It’s what happens when a technology category matures.
Identity is infrastructure
I’ve written and talked quite a bit about identity because I believe it is one of the most important capabilities in the entire customer stack. If you don’t know who the customer is, everything downstream from personalization and measurement to journey orchestration and AI, suffers.
Identity is the gatekeeper for almost everything we want to do next. But here’s the distinction I think matters: being essential does not necessarily make something differentiating. Electricity, networking and cloud computing are all essential. We don’t normally describe any of them as the source of a company’s competitive advantage. I think customer identity is moving in the same direction.
Enterprises absolutely need a reliable customer profile and a coherent identity strategy, including governance, consent and clear rules around which systems own which data and how that data moves. But once those capabilities become broadly available across data platforms, CDPs and engagement platforms, simply having a unified profile stops being particularly special.
It becomes table stakes. And that changes the conversation.
AI accelerates the shift
AI makes this even more interesting. There is a growing argument that AI needs direct access to all of the enterprise’s customer context and therefore the data platform naturally becomes the center of marketing decisioning.
I’m not convinced.
AI certainly needs access to context. But access to context and responsibility for making the decision are two different things.
An airline may have thousands of attributes about me sitting in Databricks. Great. But when my flight is cancelled at 4:17 PM, something still has to decide what happens next.
Should I get an email, a push notification or an SMS? Should the promotional campaign scheduled for 4:30 be suppressed? My status may determine whether I get a hotel offer or some other accommodation, while the fact that an agent is already helping me might mean the best action is no action at all. And if I’ve already rebooked by the time the decision is made, the message needs to change again.
Those aren’t data questions. They’re decisioning questions. And once the decision has been made, something still has to execute it. That distinction is going to become increasingly important.
The new stack has a division of labor
This is why I think one of the most important things companies can do before selecting marketing technology is define the future-state division of labor.
Not the feature list. The division of labor.
Before you start comparing vendors, decide what Databricks or Snowflake should own and what the ESP should own. Determine where identity resolution and audience creation should happen, where real-time context should live, and which platform is responsible for decisioning and orchestration. Consent and suppression need an owner, too, as does execution. And somebody needs to decide what happens when two systems disagree.
Those questions are much more important than whether Vendor A has 412 features and Vendor B has 397. Increasingly, multiple platforms can perform many of the same functions. The real architectural question is which one should. That is a very different way of evaluating technology.
Decisioning is where things get hard
Sending an email is not particularly difficult anymore. Neither is storing an attribute, and even building an audience is becoming increasingly commoditized. Making the right decision at the right moment across millions of customers is much harder.
Every decision has to account for constraints and competing priorities: eligibility rules, frequency caps, channel preferences, promotional campaigns competing with service messages, business rules, regulatory requirements and even inventory considerations. Now add AI-generated recommendations, along with the governance, explainability and human override they require.
That’s a considerably harder problem than storing the customer data that feeds the decision. Which is why I think this is where the real battle is moving.
The question is no longer simply, “Who has the customer data?” Everybody is going to have access to the customer data. The question is who can turn that data into the best next decision—and then execute that decision reliably, immediately and across the right channel.
That is much harder to commoditize.
Advantage ESPs?
This is where my ESP roots probably show.
I think modern ESPs may be better positioned for this transition than many people expect. For years, ESPs were viewed primarily as execution engines: build the campaign, select the audience and send the email.
But the best modern platforms have been moving steadily upstream. They now ingest behavioral events and maintain customer state while using that information to determine eligibility, personalize content, orchestrate journeys and coordinate activity across channels. Increasingly, they’re also making those decisions in real time.
And, importantly, they already sit very close to the moment of execution. That matters.
There is a big difference between knowing something about a customer and being operationally responsible for deciding what happens to that customer next. Could a warehouse-native architecture ultimately handle all of that? Of course.
But I don’t think the answer is automatically, “Put everything in the warehouse and let AI figure it out.” The operational side still has to deal with latency, queues, retries, suppressions, frequency management, deliverability, channel coordination and what happens when something fails.
Execution is messy. And execution is where the advantage lives.
So what happens to the CDP?
I don’t think the CDP disappears. I think the category changes.
Some CDP capabilities will migrate downward into the data infrastructure while others migrate upward into engagement platforms. There will also continue to be enterprises where a standalone CDP makes perfect sense because their architecture, organization or use cases justify it.
But “we need a CDP” is becoming a less useful starting point. The better question is: what capabilities do we need, and where should each of those capabilities live? The answer might be a CDP. It might be a warehouse-native architecture, a modern ESP with substantial CDP functionality, or some combination of all three.
Architecture should determine the answer, not the category name.
The moat moved
For a long time, the industry focused on assembling the customer. That work mattered, and it still does.
But we are reaching the point where assembling the customer is no longer enough. The next generation of marketing technology will compete on what happens after the customer has been assembled.
The winning platforms will need to interpret the signals and understand the context well enough to decide what should happen next. They’ll have to coordinate competing priorities, recognize when doing nothing is actually the best decision, and then execute the decision reliably at scale.
That is where I believe the next moat is being built.
The data moat isn’t really a moat anymore. It’s the foundation. The competitive advantage is what you build on top of it.
Decisioning. Orchestration. Execution.
That’s where the fight moves next.
And if I were placing a bet today?
Advantage ESPs.
