Seeing the Business (Ch 13)

Opens Part 3. Reis’s central argument: technical people build schemas without understanding the business processes those systems serve. The “Hellta” airline case study shows a two-year platform migration that launched successfully as infrastructure but failed catastrophically because nobody modeled the business.

Three Audiences for Every Model

People (visual maps and diagrams), applications/data systems (precise schemas for integrity), and AI agents (context via metadata, semantic layers, ontologies). A change in business process must be reflected everywhere humans and machines consume data.

Process Discovery Method

Identify trigger event, business object in motion, actors (including AI agents), and sequence of steps. Map process flow, then translate to a data model: entities from business objects, events from workflow actions, state changes from steps, relationships from actor handoffs.

Domain-Driven Design for Data

Bounded contexts are linguistic boundaries where terms have unambiguous meaning. Every time a process arrow crosses a swimlane, you are likely crossing a bounded context — a data quality risk zone. Ubiquitous language insists the business’s vocabulary match the data model’s column names. EventStorming (orange = domain events, blue = commands, yellow = aggregates) is a rapid alternative to formal BPMN.

Context Collapse

When domain distinctions, temporal dimensions, and actor identities are flattened, data becomes ambiguous. Symptoms: dashboard contradictions, zombie data, semantic slush. Metadata is the structural glue that preserves reality.

RDCO Relevance

Process-first modeling aligns with our dbt consulting approach: start with business questions, map domains, then build. The Hellta case study is potent client-facing content. Cross-ref SDG pipeline articles on stakeholder alignment.