Power BI semantic model: the layer every report relies on
Power BI semantic model is the current name for what Microsoft used to call a dataset. It is the layer between raw data and the charts a reader sees. The model holds tables, relationships, measures and security rules. Because every visual reads from it, the quality of that layer decides whether reports are fast, consistent and trusted.
- Formerly called a dataset
- What lives inside
- Why one shared model helps
Last updated
What is a Power BI semantic model?
Raw data arrives as rows and columns with technical names. A business reader, however, thinks in terms such as revenue, customers and last quarter. The semantic model translates between the two.
It stores the tables, either as an imported copy or as a live link to the source. It also records how those tables relate. Then measures written in DAX define each business figure once.
So the word semantic points to meaning. The model tells the platform what the data means, and a report simply displays it. When you publish from the desktop application, the service creates both items side by side.
What does a Power BI model contain?
Six elements make up the layer, and each deserves attention.
Tables
Facts hold events and numbers, while dimensions hold descriptions. Together they form the base of everything else.
Relationships
Links between tables let a filter on one reach another, so a chosen region narrows the sales.
Measures
DAX formulas calculate figures on demand. Because they live in one place, every report agrees.
Hierarchies
Levels such as year, quarter and month give readers a natural path for drilling down.
Security roles
Row-level rules limit which rows each person sees, and they apply in every report on the model.
Storage mode
Each table can import data, query the source directly or, in Fabric, read straight from the lake.
Why should many reports share one Power BI model?
In many tenants, each report carries its own private copy of the data. At first that feels quick. Later it means ten refreshes of the same tables and ten slightly different versions of margin.
A shared model fixes both problems. Authors connect new reports to it through a live connection and add only visuals. Since the logic sits in one place, a corrected measure updates everywhere at once.
It also supports self-service safely. Business users can build their own reports, or analyze the model from Excel, without touching the underlying logic. Therefore central teams keep control while others keep freedom.
Signs of a healthy Power BI semantic model
Use this list as a quick check on any model you own.
- A star schema with clear fact and dimension tables
- Friendly names, with technical keys hidden from readers
- Measures grouped in folders and described in plain words
- No unused columns or tables adding size
- A named owner and a documented refresh schedule
- Security roles tested with real user accounts
How to build a Power BI semantic model that lasts
These four steps apply whether the model is new or due for a rebuild.
Agree the business terms
First settle what each key figure means with the people who use it. After that, the formulas are easy.
Shape the tables
Then prepare facts and dimensions, as far upstream as your platform allows.
Write and test measures
Build each measure once and compare results with a trusted source before anyone relies on it.
Publish and endorse
Finally, publish to a managed workspace. Mark the model as promoted, so authors know which one to use.
Questions about Power BI models and datasets
Is a Power BI semantic model the same as a dataset?
Yes. Microsoft renamed datasets to semantic models, and the item itself did not change. Older articles, scripts and some menus may still use the earlier word. So when a colleague mentions a dataset, the two of you are talking about the same thing under a different label.
What is the difference between a model and a report?
The model holds data and logic, while the report holds pages and visuals. One model can feed many reports. When you publish a file from the desktop application, the service separates it into both items. Because of that split, teams can manage data and presentation independently.
Can Excel connect to a published Power BI model?
Yes. Analyze in Excel opens a pivot table that reads from a published model. The data stays in the service, and row-level security still applies. For finance teams that love spreadsheets, this route gives familiar tools on top of governed numbers, without exporting copies that go stale.
How large can a model in Power BI get?
The limit depends on the license or capacity behind the workspace, and Microsoft documents the current figures. In practice, design matters more than the ceiling. Removing unused columns, reducing detail and using incremental refresh keep most models well within bounds. We check those options before suggesting a higher tier.
Can you take over a model someone else built?
Yes. We start by reading the model, its measures and its refresh history, and then document what we find. A free 30-minute assessment comes first. Power BI Agency is a division of Sataware Technologies, works remotely inside your own tenant and is independent of Microsoft.
Let's build your first dashboard together.
Tell us what you report on today. A consultant replies within 2 hours with next steps.
Trusted partners
Microsoft Solutions Partner · AWS Partner
Cloud & AI Platforms
Azure SQL, Microsoft Fabric and the data pipelines behind your reports.
AI Business Solutions
Power BI reporting on your Dynamics 365, Business Central and Power Platform data.
Security
Row-level security and Entra ID sign-in on every report and workspace.
Amazon Web Services
Power BI on your AWS data, with connectors for Redshift, Athena and S3.