Good Strategy for Marketing Technology

Every growth-stage company I talk to has the same quiet problem. They have spent real money on marketing technology and AI. The dashboards are green. The roadmap is full. And no one can say, with a straight face, what any of it did for the business.

They are not alone, and the research says so plainly. In MIT’s 2025 study of enterprise AI, about 95% of generative-AI pilots delivered no measurable profit-and-loss impact, while only about 5% drove rapid revenue. A separate S&P Global survey found AI-initiative abandonment climbed to 42% in 2025, up from 17% a year earlier. Most of the money going into AI marketing buys activity. Little of it buys performance.

The instinct is to blame the tools, and the instinct is wrong. When an AI initiative underperforms, teams reach for a better model, a newer platform, another vendor. I have watched that reflex burn quarters. The tools are rarely the binding constraint. The work has no strategy underneath it.

I mean strategy in Richard Rumelt’s sense, not the marketing-plan sense. In Good Strategy / Bad Strategy, Rumelt shows that most of what gets called strategy is fluff: a vision, a list of goals, a roadmap. Real strategy is three things working together: an honest diagnosis of the core problem, a guiding policy for dealing with it, and coherent actions that carry the policy out. Bad strategy skips the diagnosis and jumps straight to goals. A wall of green metrics and a roadmap full of features can feel like strategy. Underneath, it is a wish list with instrumentation.

Good strategy starts by naming the crux, and in marketing technology the crux is almost always the same. The organization has fragmented, untrusted capability that every team is quietly rebuilding. Marketing has its data. Product has its own. Analytics keeps a third copy. Each team ships its own AI feature on its own slice, and no one trusts the numbers enough to act without double-checking them first. The expensive truth sits one level down: the dashboards are green because they measure the wrong thing. They confirm that a model ran, a campaign sent, a feature shipped. They stay silent on whether the answer was right, the message reached a real person, or the feature moved the metric it promised to move.

I have learned to distrust a green dashboard on an AI product. I once paused the launch of an AI product because my own end-to-end testing found it returning confident answers that were quietly wrong, while every engineering metric read green. Those metrics were tracking whether the system responded, not whether it was correct. Green dashboards lie by omission, and in AI they lie hardest where the cost of being wrong is highest.

The guiding policy that follows from that diagnosis is counterintuitive: build the one asset that is both your long-term advantage and this quarter’s win. Most teams treat those as a trade-off. You can ship something fast for the business, or you can build a durable platform for the future, and the budget rarely stretches to both. Good strategy declines the trade. The move is to find the single intelligence layer that, built once, removes an acute daily pain and compounds into an advantage a competitor cannot easily copy. One team I worked with was drowning in customer feedback spread across a dozen channels, with no time to analyze it and no confidence it was complete. They built a single trustworthy feedback layer that standardized the signal, routed each issue to the team that owned it, and grew more valuable every month as the company scaled. It solved Monday’s problem and became the moat at the same time.

A guiding policy is worthless without coherent action, and this is where most martech strategy quietly dies. Three actions carry the policy, and teams tend to skip all three. First, earn trust in the asset by gating it on outcomes rather than activity: define what “good enough to ship and perform” means before you build, treat it as a hard gate, and hold the line when the work does not clear it. Second, align the organization around the shared asset. The hardest sentence in this work is “why are you building your own when we already have a head start,” and someone has to say it, more than once, or every team drifts back to its own copy. Third, prove it: drive adoption, show the business outcome, and let the advantage compound. This is the unglamorous part. It is coordination, standard-setting, and saying no, and it is the work that separates a strategy on a slide from one that performs.

So if you take one thing from this, start with the diagnosis, not the tool. Before you approve the next platform or the next pilot, sit with the harder question: what is the actual crux, and what single asset would be both the fix and the advantage. That question is worth more than most roadmaps.

The discipline itself is simple to state, even if it is hard to hold: find the crux, make the pivotal bet, earn trust, align the organization, and prove it. The tools will keep getting better every year. The teams that win will be the ones with the strategy to point them at the right thing and the discipline to prove it worked.