The backbone of every good and fast DAX calculation is a well-defined data model.
What is a Data Model?
In simple terms, the data model consists of all the tables in your Power BI/Excel/analysis services database that are connected through relationships. The definition also includes other, non-related tables, as well as other objects in the model like Measures, calculation groups etc., but for now, we will stick to the simple definition.
Why Data Model?
If you come from an Excel environment, you might think why would you need a data model, when you can have it all in one single, big fact table? There are several reasons that go in favor of preparing a proper data model prior to any calculation or data analysis. We will mention a few of them:
- Easier understanding of the data
- If you have multiple columns that have the same name (e.g. Address which can be linked both to customer, buyer, employee, or other parts of the data model), it’s much easier to spot the correct one if it is a part of the dimension (e.g. Customer[Address]), rather then denormalized in a single table with prefix.
- Handling data with different granularities
- Imagine having actual data on an SKU level with daily sales. You need to compare it to the forecast values that are on the monthly/brand level. It would be a daunting task to incorporate both tables into a single fact table, due to their obvious differences in granularities. In the case of a data model, you can prepare the dimension that can simultaneously filter both actuals and forecast data.
- Simpler and faster DAX code
- You can forget about using time-intelligence functions without a date dimension linked with a fact table
- Also, most complex DAX code can be simplified with the help of a well-defined data model
- More flexibility
- With additional tables, you can introduce parameters, calculation groups, field parameters, and other objects that cannot be a part of a single table.
- Better security
- Row-level-security is best defined on dimension levels that filter necessary table/s.
There are many more advantages to having a proper data model than a single table that we haven’t thought to mention. In the following articles we will explain optimal data model structures, often referred to as a star scheme.
