Power BI semantic models that survive reorganisation

A semantic model that breaks whenever the organisation restructures was built against the organisation chart instead of the business process. Reporting lines change every year or two; the underlying processes of selling, servicing, delivering and invoicing change far less.

Design choices that hold

• Model the process, and treat organisational structure as a dimension with history rather than as model structure.
• Keep hierarchies in slowly changing dimensions so historical reporting remains accurate after a restructure.
• Define measures once in the shared model. Report-level measures are where definitions quietly diverge.
• Apply row-level security by role, sourced from a maintained mapping rather than hard-coded identities.

Governance that keeps it trustworthy

1. A named business owner for each core measure, recorded in the model description and the data catalogue.
2. A written definition in business language for every published measure.
3. A change process: definition changes are proposed, reviewed and communicated, with dependent reports identified before the change lands.
4. Certification for models intended as an authoritative source, so users can tell them apart from personal work.

Retire deliberately

Usage metrics show which reports are genuinely used. Retiring the rest is not housekeeping. It reduces the surface that must be revalidated every time a definition changes, and it makes the remaining reports easier to trust.