We're Building Too Much Marketing Tech for No Clear Reason

The infrastructure boom in martech mirrors the data center problem—and could end the same way

Marketing technology spending hit $23 billion in 2023, yet marketers use just 42% of their stack capabilities. Not the advanced features. The basic ones. Publishing content, tracking performance, connecting systems—tasks that should take minutes eat up hours.

Sound familiar? It should. The same pattern is playing out in AI infrastructure, where companies are pouring over $5 trillion into data centers, chips, and computing power while actual revenue-generating use cases remain sparse. McKinsey found that fewer than 15% of AI pilot projects in surveyed firms succeed. The tech gets built. The applications don’t materialize. The investment doesn’t pay off.

Martech is following the same script.

The accumulation problem

Most marketing stacks weren’t designed. They were accumulated. Someone needed faster launches, so they added a tool. Someone needed better tracking, so they plugged in a platform. The growth team needed their own analytics, so they bought their own license. Over time, all those point solutions add up to a sprawling mess that nobody fully understands.

The average B2B organization now manages between 162 and 269 marketing applications. That’s not vendor counts—that’s distinct tools. Each one solving a problem that seemed urgent at the time. Most sitting largely unused today.

Here’s the tell: when you ask marketing teams what percentage of their tools they actively use, the answer clusters around 30 to 40%. That means 60 to 70% of the stack is waste. Not underutilized advanced features. Entire platforms that teams forgot they pay for.

Gartner research found that organizations with $250 million in revenue waste nearly $4 million annually on marketing technology that doesn’t deliver value. Scale that up to enterprise budgets and you’re looking at real money disappearing into software subscriptions that solve problems nobody remembers having.

The AI parallel

The data center boom offers a preview of where this goes. Companies are spending massive sums to build AI infrastructure without clear visibility into what it’ll be used for. The logic is: we need the capacity to run AI models, train algorithms, and process data at scale. Build it now, figure out the applications later.

Some of this will work out. AI will find productive uses, businesses will figure out how to monetize capabilities, and the infrastructure will get utilized. But history suggests a lot of it won’t. Look at the late-1990s telecom bubble, when companies laid fiber-optic cable and built network capacity far beyond what demand required. Billions in investment got written off when the bust came.

Martech is showing the same warning signs:

Redundant capabilities everywhere. How many tools in your stack can send an email? How many can track a conversion? How many store customer data? Most organizations have three to five platforms that do roughly the same thing, purchased by different teams at different times for slightly different reasons.

Integration nightmares. Connecting tools costs more than buying them. When one system updates, integrations break. Teams spend weeks getting data to flow between platforms that are supposed to work together but don’t quite. The promise of a unified ecosystem becomes a patchwork of workarounds.

Low adoption rates. New platforms get purchased with grand ambitions, adopted by one team, and then left to stagnate. Licenses renew because nobody wants to be the one to decommission something that might still be useful. The result is shelfware at scale.

Why this keeps happening

The explanation isn’t incompetence. It’s incentives. Here’s how the cycle works:

A new platform launches with compelling capabilities. Early adopters see real results. Case studies circulate. The platform raises funding, builds more features, expands its reach. Marketing leaders see peers using it and worry about falling behind. Procurement processes favor established vendors with lots of features, even when simpler tools would work better.

So the stack grows. New tools get layered on top of old ones. Nobody wants to remove anything because what if we need that later? Tech debt accumulates. Complexity increases. Team productivity decreases.

Then someone realizes the stack is a mess and launches a “consolidation initiative.” This usually means buying an enterprise platform that promises to replace five point solutions. Except the enterprise platform is harder to use, takes longer to implement, and doesn’t quite do what the point solutions did. So the point solutions stick around. Now you have the enterprise platform and the old tools. The stack just got bigger.

Sound like infrastructure over-building? That’s because it is.

The utilization myth

Here’s a dirty secret about martech utilization metrics: they’re misleading. When reports say teams use only 30% of their stack, the implication is we should be using more of it. What if the real problem is we bought too much?

Software doesn’t work like physical goods. You don’t waste money by not using every feature. You waste money by buying features you never needed in the first place. The instinct to “maximize utilization” leads to Feature Creep Syndrome: teams force-fitting workflows into complex tools because those tools are there and expensive.

Better question: What capabilities actually drive business outcomes? Everything else is optional.

Recent data shows 68.6% of organizations now use generative AI tools, making them the 6th most popular martech category. That adoption happened in under two years. But when asked what they’re using AI for, the answers cluster around content generation—the easiest, lowest-value application. The hard problems (customer lifetime value prediction, attribution modeling, personalization at scale) remain mostly unsolved.

That’s the utilization trap. Organizations adopt technology because it exists, not because they have a clear problem it solves. Then they scramble to find uses for it.

What actually matters

The marketing organizations getting this right aren’t the ones with the biggest stacks. They’re the ones with clear principles:

Build around data, not tools. Most stacks fail because they’re organized around software vendors instead of information flows. Start with what data you need, where it comes from, and where it needs to go. Then pick tools that fit that architecture. Not the other way around.

Fewer tools, better integration. Six well-integrated platforms beat twenty disconnected ones. Every time. The cost isn’t in the software licenses. It’s in the integration work, training, maintenance, and ongoing confusion. Consolidation isn’t about following vendor roadmaps. It’s about ruthlessly cutting anything that doesn’t directly support priority use cases.

Measure outcomes, not outputs. Marketing technology should make revenue more predictable, not just easier to report on. If a platform’s main benefit is dashboards, you don’t need it. Ask what decisions the platform enables that you couldn’t make before. If the answer is vague, that’s your signal.

Resist shiny object syndrome. Every year brings new categories: marketing automation, account-based marketing, customer data platforms, AI-powered personalization engines. Most are solving problems that only exist because your stack is too complicated. Before adding another layer, ask whether simplifying what you have would work better.

The forecast nobody wants

Here’s where this likely goes: martech spending keeps growing for another few years as AI capabilities get embedded in every platform. Organizations keep buying because competitors are buying and nobody wants to miss out. Utilization stays low because the underlying problems—too much complexity, unclear strategy, weak data infrastructure—don’t get fixed.

Then something breaks. Maybe it’s a recession that forces CFOs to scrutinize software spend. Maybe it’s a major platform failure that exposes how brittle these systems are. Maybe organizations just hit a breaking point where adding more technology makes everything worse instead of better.

When that happens, the correction will be sharp. Marketing technology that can’t demonstrate clear ROI will get cut. Vendors that built businesses on selling features instead of solving problems will struggle. And a lot of money currently being invested in building martech capacity will turn out to have been wasted.

The smart move is to get ahead of that. Build the minimum viable stack that can grow with your business. Focus on integration and data quality over feature count. And stop treating technology as the solution to problems that are really about strategy and organizational design.

Infrastructure matters. But only if you build it for problems you actually have.