← Back to the blog

Blog

Managing Invisible IT Infrastructure for Value

Selling invisible IT infrastructure is hard, steering it even harder. A recent GPM article on linked budget data in Berlin shows how standards, data models and a shared terminology often decide the later value of digital applications.

In April 2025, the initiative published Berlin’s budget data as Linked Open Data. Berlin and Schleswig-Holstein jointly developed a semantic data model across state lines; Berlin also documents open standards and the basis for a knowledge graph spanning public authorities. What the project visibly delivered was a reliable prerequisite for future analyses and services, rather than a new citizen-facing application.

Foundation work needs a different business case

Infrastructure projects tend to fall into a governance trap. The costs arise today; the benefits only appear in subsequent projects. Steering committees then report technical completion, while the question of who will actually use the new capability remains unanswered.

Four questions help make the benefit tangible:

  • First user: Which process or follow-up project will put the result to use first?
  • Useful increment: What can go into productive use before the target architecture is complete?
  • Adoption: Which teams, suppliers or organizations need to apply the standard?
  • Benefit chain: Which later decision, automation or quality improvement does this enable?

The Berlin case shows a pragmatic order of steps: vocabulary and data model first, publication on an available platform next, long-term infrastructure later. That is a deliberate, manageable increment rather than an improvised workaround.

The pattern in IT projects

I see the same pattern in migrations and integration programmes. A canonical data model, an interface convention or reconciliation logic initially looks like technical preparation. If it is only accepted as a document, its value stays hypothetical.

That is why I would require a first usage path in the steering committee: one real dataset, one defined consumer, measurable quality rules and a decision log for unresolved mappings. In migration reconciliation, “schema complete” is not enough. What counts is whether a business process can compare target and actual data transparently and approve deviations with clear accountability.

What project owners should do now

Every infrastructure roadmap should include a benefit hypothesis, a named first user and at least one production-ready increment. Reporting needs to cover adoption, reuse and unresolved governance decisions alongside the technical milestones.

Whether the foundation is complete matters less than which next project it demonstrably makes faster, safer or possible in the first place.

← Back to the blog