Land O’Lakes rewrites the rules of product-based data alignment

The benefits of a product operating model are now well established. Organizations that shift from project- to product-based IT achieve tighter collaboration between business and technology, orienting autonomous, empowered teams around business capabilities to deliver direct business impact. In pursuit of those outcomes, many organizations embed a business lead within each product team, someone responsible for shaping roadmaps, evaluating throughput, and dialing up resources when supply fails to meet demand.

It’s the right idea. But it doesn’t always work.

“Despite the best intentions, the matrixed nature of this relationship doesn’t always deliver the intended outcome,” says Jeremy Lembeck, director of data and analytics at Land O’Lakes, the Minnesota-based agricultural cooperative where he’s spent the past decade helping modernize data capabilities across the business. “The spirit of the model is sound, but in practice, the influence business sponsors have is often diluted.”

So the Land O’Lakes team tried something different. Rather than align business stakeholders to virtually lead data teams that still report into technology, they flipped the model. They positioned data product teams to reporting directly into the business.

“AI is something everybody wants more influence over,” says Lembeck. “We wondered: Could we achieve better outcomes by giving business units more ownership of their data product teams — data engineering, data science, and product leadership?”

It’s not a silver bullet, he admits, but there’s something powerful about putting business units in the driver’s seat. Ownership creates accountability. When business leaders own the backlog and the reporting relationship, it becomes more difficult to outsource responsibility — or to underestimate what it takes to build scalable solutions.

“Although the cost centers, and the talent within them, still sit in tech, the dynamic has flipped,” Lembeck explains. “Instead of IT endlessly petitioning for more resources to hit the scope, the business now champions the investment. And that change in directionality shows up as more speed to market.”

A BU-centric structure

To make the model work, Land O’Lakes embedded a data product owner into three of its business units: WinField United, Truterra, and Animal Nutrition. These product owners report directly into the business unit and work alongside business leadership to define value creation goals that support BU-specific objectives.

Each product owner is supported by a multidisciplinary team: data engineers who build the pipelines that transform raw data into usable formats; data scientists who match the right algorithms to the right use case; and BI analysts who visualize the results — descriptive or predictive — through dashboards and tools that support decision-making. The teams use agile ways of working, but the emphasis isn’t on the ceremony. It’s on outcomes.

Supporting each of these BU-specific product teams is a centralized platform engineering team that reports into Lembeck. “This team really serves as the connective tissue,” he says. “They connect platforms, build accelerators, and ensure the factory is efficient.”

Breaking from the matrix

Most organizations still favor a matrixed model, where business stakeholders lead data work in spirit, but where resources ultimately report into technology. Land O’Lakes did the opposite. The majority of data roles now report the business.

Critics of that approach often raise two concerns: How do you prevent shadow IT, and how do you ensure that specialists like data engineers and data scientists are still learning their craft?

Lembeck’s response: “There’s always the possibility of ‘shadow IT.’ But if we’re doing our jobs building trust and keeping the conversation going, we can still shape outcomes even in a decentralized model,” he says. “And because architecture and technology decisions are federated to the teams, the people doing the work own the how instead of waiting for a central group to dictate it. That ownership is where the magic happens.”

He credits his boss, Land O’Lakes CTO Teddy Bekele, with shaping this approach: “Teddy’s taught me so much about the power of working horizontally. By talking to enough people across the business, you get a pulse on what’s working, what’s broken, and where to focus your energy.”

As for capability-building, Land O’Lakes has taken a page from Spotify. Lembeck established “guilds” for data engineers and data scientists, functional communities that share best practices across business units and help raise the bar on delivery quality. “They don’t just share what worked,” he says. “They share the scars too.”

Making the mindset stick

To ensure the product mindset takes root, Lembeck and his team avoid getting hung up on semantics. “We try not to spend time debating whether someone is a product owner or a product manager,” he says. “The language matters less than the behaviors.”

Instead, they focus on the fundamentals. Are we proactively defining what’s next? Are we reviewing product with the business before we ship? Are we speaking their language? Are we measuring outcomes, not just outputs?

“The ways of working matter,” he says, “but the product mindset isn’t about the lingo — it’s about the mindset and the outcomes.”