The last mile: why great first-party data still doesn't make great marketing

The last mile: why great first-party data still doesn't make great marketing | Databricks Blog Skip to main content

Summary

*The martech stack is broken by design. Decades of layered, siloed tools have left marketing teams unable to activate the customer data sitting in their modern data platforms — creating a costly gap between data infrastructure investment and actual campaign outcomes.

*The composable canvas closes the gap, but only with the right approach. Databricks provides the unified data foundation and the Agentic CDP that eliminate integration bottlenecks; the last mile requires building the bridge between that foundation and real-time, AI-powered marketing execution.

*Brands building on this architecture today are already winning. From automating manual data workflows that consume entire weekends, to deploying AI agents that trigger personalized campaigns from live data signals, the composable marketing future is being built now — not in three to five years.

Last month, a data team and a marketing team sat in the same room and talked past each other for 45 minutes. The engineers spoke in Delta tables and medallion architecture. The marketers spoke in journeys, segments, and send time optimization. Same customer. Same data. Two completely different languages. Nobody was wrong. They just couldn't hear each other. That dead space — between where data lives and where it activates — is where millions of dollars in customer value go unrecovered every week, at companies with world-class data infrastructure and world-class marketing ambitions. They have the rocket ship. They forgot the launchpad. Scott Brinker's research report,  The New Martech "Stack" for the AI Age , published in partnership with Databricks, puts the right name on what needs to replace the architecture that created this problem. He calls it the composable canvas. The diagnosis is exactly right, and the direction it points is exactly where this partnership is built to go. What the composable canvas actually means for marketers If you've ever exported a CSV from your data warehouse, emailed it to the campaign team, and waited three days for it to get uploaded into your marketing platform — that's the old stack model failing you. For two decades, martech was built in rigid vertical layers: data at the bottom, systems of engagement in the middle, campaigns at the top. Each layer was its own box. Getting those boxes to talk to each other required pipelines, connectors, sync jobs, and teams of people whose entire job was moving data from one system to another. As Brinker's report notes, integration remained a top-three challenge for the majority of marketing organizations surveyed — not in 2015, but in late 2025. After thirty years of vendors promising to solve it. The composable canvas is the architectural alternative: a unified data foundation where every tool, from your customer engagement platforms, to your AI agents, to your analytics, operates on the same shared substrate, without data ever having to move. No middleware. No lag. No Sunday night spreadsheet ritual. Bryce Peake, former VP of Marketing Decision Sciences at Domino's, puts it plainly: We've been trying to modify the same martech stack we've had since the internet started interneting. Folks, we're going to have to build a new one. Modern data platforms like Databricks provide an open, shared foundation that marketers, agents, and apps can all operate on together. And the business case is concrete: "Speed: From months to minutes. Today, adding a new marketing capability means integration projects, data pipelines, and IT tickets. In the composable canvas, new tools and AI agents plug into a shared data foundation instantly." The architecture: five rings, one center of gravity Brinker's framework organizes the composable canvas into five concentric rings, each with a distinct role.

Data Core:  the unified foundation of customer, company, content, code, and control data; the gravitational center of everything Semantic Layer: shared definitions that make data consistent and meaningful across every system that touches it CaaS (Context-as-a-Service): platforms like CDPs that package relevant data and context for the apps and agents that need it Decisioning: where AI engines optimize next-best actions and resolve contention when multiple agents want to reach the same customer simultaneously Apps & Agents: the outermost ring, where customer experiences actually get delivered

The payoff of this structure is a dramatic reduction in integration complexity. In the old point-to-point model, ten systems could require up to 45 integrations. In the composable model, each new capability joins a coherent shared ecosystem rather than a web of brittle pipelines. As Rick Schultz, CMO of Databricks, puts it: All CMOs are trying to solve the same three things: effectiveness, efficiency, and self-service. The composable canvas is how all three become achievable simultaneously. Elizabeth Dobbs, AVP of Marketing Technology at Databricks, describes the impact from firsthand experience: Once all of our marketing data was centralized in the lakehouse, we revisited models that had never fully leveraged that richness... The rebuilt model converted at 4X the previous rate. From there, we quickly replicated the same approach to launch a highly predictive account scoring model in just four weeks. The last mile: the gap between your first-party data platform and a live campaign Brinker is explicit that this report is a North Star, not a step-by-step implementation guide,  and that's exactly what makes it valuable. What it intentionally opens up is the practitioner question: how does a unified data foundation actually become a triggered campaign? How does a decisioning layer get built for a marketing team that doesn't speak data engineering? How do the five rings move from an architecture diagram to a customer experience that fires at the right moment? These questions are where the real work of implementation lives — what Stitch calls the last mile. Here's what the gap between the starting point and the finish line of the last mile actually looks like in practice: The autonomous agents with nowhere to go. A marketing leader at a major brand recently described a maddening situation: her engineering team had just deployed an autonomous agent system on Databricks that could reason across their entire...