What is Power BI semantic model?
A Power BI semantic model is the structured data layer at the heart of every Power BI report — defining tables, relationships, calculations, and measures so business users can answer questions consistently without rebuilding queries each time.
The semantic model is the difference between a Power BI report that produces the same answer for the same question every time and one that produces five different answers depending on who is asking. It is where business definitions live: revenue, gross margin, customer count, retention rate. Built well, a semantic model becomes the trusted analytical layer for the entire organisation. Built badly, it becomes a maintenance burden of one-off measures and conflicting calculations. Codnity Data builds star-schema semantic models with documented business rules, governed measures, and incremental refresh from accounting and operational source systems.
What it includes
- Star schema (fact and dimension tables)
- DAX measures with documented business logic
- Time intelligence (YoY, MoM, rolling averages)
- Row-level security and tenant isolation
- Incremental refresh and partition strategy
- Governed dataset for reuse across reports
How it works
Model the business
Identify facts (sales, transactions, events) and dimensions (customer, product, time, geography). The shape mirrors how the business reasons about itself.
Build the model
Star schema with cleanly defined relationships, calculated columns where the source cannot deliver them, measures for everything else.
Govern the measures
Single source of truth for every business definition. Documented in the model and referenced consistently in reports.
Optimise refresh
Incremental refresh on large fact tables, partitions for performance, scheduled refresh aligned to business cadence.
Frequently asked
Star schema or snowflake?
Star schema for almost every Power BI use case. Simpler for users, faster to query, easier to maintain. Snowflake only when the cost of denormalisation is genuinely prohibitive — rare in practice.
Calculated column or measure?
Measure when the calculation depends on filter context (most cases). Calculated column when it is row-level and reused across many measures, where computing once at refresh saves cost.
How big can a Power BI model get?
Power BI Pro: ~1 GB per dataset. Power BI Premium: 10–400 GB depending on capacity. Beyond that, partition strategy and DirectQuery patterns become essential.