Dashboards people trust start with the semantic model

Every organisation has a graveyard of dashboards nobody opens. The common cause is not visual design — it is that each dashboard was built straight on top of raw tables, so two reports of the same metric disagree, and trust evaporates the first time someone notices.

The semantic model is the product

A semantic model is the shared definition of your business: the measures, the relationships, the hierarchies, expressed once and reused by every report. Build it as a star schema — fact tables surrounded by conformed dimensions — and define every metric as a measure, not a column.

Net Revenue =
CALCULATE(
    SUMX(
        fact_sales,
        fact_sales[quantity] * fact_sales[net_price]
    ),
    fact_sales[is_returned] = FALSE
)

Defining Net Revenue once, in the model, means every report that references it agrees by construction. A number computed in the report layer is a number that will eventually disagree with another report.

Design for the question, not the export

• Start from the decisions the report is meant to support.
• Show the metric, the trend, and the one comparison that gives it meaning.
• Resist the request to "add every column" — that is a request for a spreadsheet, not analytics.

A report built on a solid semantic model can be simple, because the model carries the complexity. That is the report people keep open.