Data platforms are often funded as if plumbing were the outcome. The outcome is a decision that can be made on time, with a number that can be explained. Plumbing without an operating model recreates the warehouse: central, expensive, and still not quite believed.
Products, not pipelines
A data product has a consumer, a contract, a steward, and a service level. It has a way to change without surprising finance or operations. Pipelines can implement that. They cannot invent it.
- Who is the steward, and what are they allowed to change?
- What is the contract—grain, latency, allowed nulls, lineage?
- What is the service level when the source system is late?
- Who is the consumer, and how do they raise a defect?
Those questions belong in discovery, not in a hypercare war room. They are also why Data & AI work has to sit with technology and finance rather than as a lab beside the estate.
Build the operating model for the product. Then the platform has something to serve.