Semantic Model Practice

HI powerbiexpert22​ 
if you want to learn from the starting about Semantic model. It is better to starts with it.
Link : https://learn.microsoft.com/en-us/power-bi/connect-data/service-datasets-understand

Essentially, those are somewhat related (good data engineering breeds simple semantic models, and bad business process creates "complex" semantic relationships)

Government datasets should have some excellent source options for you that would have interesting time-series analyses to practice on. Look for something that requires multiple time-series components and that will help the most with DAX and modeling.

Hi powerbiexpert22​,

If you want to practice semantic modeling rather than DAX or Power Query, I would deliberately choose scenarios where the relationships and grain are the hard part.

A few good practice cases would be:

1. Sales and Budget as two fact tables sharing Date, Product, Customer, and Geography dimensions.
2. Customers belonging to multiple Accounts or Students belonging to multiple Courses, so you have to design a proper bridge table for a many-to-many relationship.
3. A fact table with Order Date, Ship Date, and Delivery Date using the same Date dimension as different roles.
4. Two fact tables at different grains, for example daily Sales versus monthly Targets.
5. A slowly changing Customer or Product dimension where attributes change over time.
6. A factless fact table, for example employee attendance, promotions eligibility, or course enrolment.
7. A model where bi-directional filtering looks tempting, and then redesign it to avoid ambiguous filter paths.
8. A composite model that mixes Import with DirectQuery or Direct Lake tables and forces you to think about cross-source relationships and performance.

Microsoft’s star schema guidance is still the best foundation for these scenarios, and the many-to-many relationship guidance is particularly useful once you start introducing bridge tables.

I would also spend time on the bi-directional relationship guidance, because one of the most useful modeling skills is knowing when not to use Both-direction filtering.

A good challenge is to take one business scenario and model it three ways, then compare:

• simple star schema
• bridge-table design
• composite model

That teaches you much more than isolated exercises because you start seeing the trade-offs in filter propagation, ambiguity, performance, and maintainability.

AI-assisted drafting: AI was used to help structure and phrase this response. I reviewed and validated the technical content before posting.

Hi powerbiexpert22​ 

Have you had a chance to look through the responses shared earlier? If anything is still unclear, we’ll be happy to provide additional support.