Anja Pilz (WDR) and Matthias Thar (BR)
The ARD’s Digitale Erneuerung (Digital Renewal) initiative is reshaping how data is managed, shared, and leveraged across a complex media ecosystem. At its core lies a product-centric approach to data: treating data as a product rather than a byproduct.
The concept of data as a product has steadily evolved over the past decade. In 2012, DJ Patil described a data product as “a product that facilitates an end goal through the use of data.” In 2020, Zhamak Dehghani established data as a product as a core principle of Data Mesh, embedding it within a broader socio-technical architecture. More recently, Jean-Georges Perrin emphasized reusable, active, and standardized data assets designed to deliver measurable value through product-thinking principles. Mario Meir-Huber distilled it even further: a data product delivers value to its consumers.
All definitions converge on a single principle: value delivery. And once data is intended for consumption rather than mere storage, certain characteristics become essential. Reusability, standardization, quality guarantees, and active lifecycle management are necessary for creating reliable, long-term value.
To translate this principle into practice within ARD and to cultivate a living data-product ecosystem, we have identified three key pillars. In a large, federated organization, purely greenfield methods rarely scale; instead, an approach is needed that provides structure without sacrificing agility.
With this in mind, we introduce a scalable and reusable model for data product development that places product thinking at its core. Drawing on proven experience from other industries, we combine visual collaboration aids such as data-product canvases, formal usage agreements like data contracts, and a central data marketplace where data products can be published, shared, and discovered.
1. Starting with the data product canvas
When setting up a new data product, we start by collaboratively filling out a data product canvas. This tool clarifies ownership, domain context, purpose, and concrete use cases. Quality requirements, access patterns, dependencies, and lineage considerations become explicit early on. By aligning all stakeholders before implementation begins, the canvas establishes shared understanding of what the data product is, who it serves, and why it exists.
We find that this framework can be applied consistently across teams and domains, even in large, federated organizations, helping embed product thinking and co-creation.
2. Formalizing expectations with data contracts
A central outcome of the canvas process is the data contract. Data contracts function much like APIs for data: they define schemas, semantics, quality expectations, SLAs, and usage terms in a standardized, machine-readable format. They act as a formal usage agreement between producer and consumer and foster trust by ensuring predictable, stable behaviour over time. Much like labels on food products, they help consumers understand precisely what they are receiving. Machine-readable data contracts are also essential enablers for future agentic automation.
3. Publishing and discovering data products via a data marketplace
To maximize their impact, we need to make data products FAIR: “findable, accessible, interoperable and reusable”. To this end, data products are published on a central data marketplace. This platform is the entry point for discovery and access for every individual in the organization. Users find data products and request access by agreeing to the associated data contract, ensuring consistent expectations around quality, stability, and permitted usage.
Once accessed, data products can be consumed as-is or used as inputs for transformations and aggregations. The resulting outputs may then be published again as new data products through the same marketplace.
Building on this model, data products are developed in an agile, outcome-oriented manner. Teams hold end-to-end ownership of their data products, supported by shared platform capabilities that ensure autonomy and consistency. This balance enables ARD to scale data usage, build trust, and increase impact over time – even amid legacy systems and cultural transformation.
This article first appeared in the March 2026 issue of tech-i magazine.