Snowflake’s Dennis Buchheim reimagines data measurement for the AI era

Snowflake is best known as a data cloud — a platform that allows organizations to unify, analyze and securely share data across teams, partners and applications. But as the worlds of media, marketing and technology converge, the company has emerged as a central player in reshaping performance measurement.

Leading that charge is Dennis Buchheim, global head of media, marketing and entertainment technology at Snowflake. Buchheim joined the company earlier this year after senior roles at Yahoo, Meta and IAB Tech Lab, where he focused on how data is used for targeting and measurement, with privacy and interoperability at the core.

At Snowflake, he brings those threads together to help media and entertainment companies break down silos and build what he calls an “open, composable ecosystem” for data and AI. The thinking goes: When data is unified and discoverable — and customers can embed AI directly into their workflows — marketers can innovate faster and do so securely with privacy governance.

Here Buchheim discusses why the industry must move beyond counting impressions to measuring impact and how AI is rewriting the future of marketing capabilities.

This conversation has been edited for clarity and length.

How do you see the role of Snowflake’s “data cloud” evolving in the ad tech ecosystem, especially with the rise of AI?

AI is changing how data is used, but it hasn’t changed the fundamentals: Data quality and access still come first. Companies can securely bring models, data and applications together in one place. It’s no longer about exporting data to external systems for analysis. The fundamental shift is bringing AI to the data so marketers can act in real time with governance and privacy built in.

You have spoken about the importance of having a data-first strategy. What are the most common misconceptions or mistakes you see media companies make when approaching their data strategy?

The biggest misconception is that data strategy starts with tools. It doesn’t. It starts with purpose. Too often, companies jump straight to technology decisions before they’ve clarified what problems they’re solving. A use-case definition is step one, solutions and technologies come late in the process.

Another common mistake is over-collecting data. You don’t need every data point — you need the right ones, and you need to understand how they connect. A smart data strategy focuses on clean, governed, AI-ready data that can serve multiple use cases. And privacy is part of the value exchange with consumers, not just compliance with current regulations. The more transparent and responsible companies are, the stronger their first-party data becomes.

How do you balance openness and interoperable data collaboration with privacy, security and proprietary interests of publishers, advertisers and platforms?

That balance is really the art of it. Our philosophy is simple: collaboration without compromise. You can have data that’s interoperable and still private.

Media and entertainment companies are sitting on enormous silos of fragmented data, from viewership logs to ad performance metrics, but the real value comes from connecting those pieces across teams and partners. It fosters an ecosystem where publishers can contribute audience insights, marketers can measure real performance and both sides can innovate faster while maintaining control and trust.

What trends in formats, audience behavior or measurement are you watching most closely? Any surprises or disruptors?

Two stand out. First is the rise of agentic AI systems that connect data, applications and workflows autonomously. That’s going to reshape how marketing and media operations run and drive a new level of automation and intelligence. Custom creative generation will transform how campaigns are built and optimized.

Fanatics, for example, is using [AI] to ask complex questions about customer behavior and inventory without a data scientist in the middle. What used to take weeks of back-and-forth with data teams, are now moments of insight and innovation.

The second is how publishers are rediscovering the value of their own content through data. Publishers can control and license their archives to AI models securely. That could open up entirely new business models and give more power back to content creators.

Speaking about publishers, how do you plan to convince or onboard legacy media companies, many of which might be struggling with their businesses, to invest in more data-driven, collaborative models?

It starts with meeting them where they are. These companies already have valuable data. They just haven’t had the infrastructure or incentives to fully activate it. Our role is to show how a secure data infrastructure built for collaboration and AI readiness can strengthen their position, not threaten it. When a media company controls its data in the cloud, it can define the terms of engagement, package audience insights more effectively, and even monetize content or archives in new ways. It’s about participating more actively in the ecosystem, not giving data away. 

And what about CTV, which is often credited as a frontier for ad innovation. You’ve said “household targeting” is a paradigm shift. How is this area changing?

CTV is fascinating because it’s highly measurable but highly fragmented. AI will play a major role in solving this. Smarter planning models can help optimize reach and frequency across channels, while secure collaboration with media providers will make measurement more transparent. The technology is ready but what’s needed now is alignment on standards, currencies and trust across the ecosystem.

What are the risks and guardrails you feel are essential when it comes to AI?

The risks are real. Data privacy, bias and misuse of generated content — all require serious oversight. The solution is embedding governance and transparency at every step, knowing what data trains models, keeping that data secure and ensuring human judgment remains in the loop.

Ad tech has long wrestled with attribution, incrementality, “last click” biases, viewability, etc. What is your view on the state of measurement today? What’s broken, what’s improving and what still needs radical rethinking?

Measurement is improving, but it’s still running behind the way audiences actually move across channels. The main challenge now is fragmentation. We’re seeing real progress through clean room-based measurement and multi-touch attribution models that connect owned, paid and earned data in a single view.

Ultimately, the industry has to move from measuring impressions to measuring impact. That’s how we’ll understand what truly drives outcomes.

Looking ahead then, how will clean rooms, IDs and measurement solutions evolve in the coming three to five years?

We’re moving from simple two-party setups to broader multiparty collaboration where several brands, publishers and agencies can analyze shared data securely in real time. The next breakthrough will come when measurement becomes continuous when AI can analyze performance mid-flight, recommend optimizations and close the loop automatically, even as modeling fills the gaps left by less observable data. The key is to maintain rigorous privacy and transparency as these systems become more intelligent.