Note: This tech note is not intended to act as a recommended method for data modelling. It is simply intended to answer the above question by describing a basic example.
A common Strategy question is when to create a single attribute composed of multiple attribute forms or whether to create each attribute form as the ID form for it's own individual attribute.
Consider a situation where a particular data model includes Airport and Location data. Of course, these can be associated since each Airport is built in a particular Location. So what is the difference between setting up two separate attributes or just creating one Airport attribute with a Location attribute form?
That depends on whether or not the relationship between Location and Airport is a 1-1 relationship where no aggregation is needed, or whether it is a Many-1 relationship in which case Airport data may need to be aggregated up to the Location level.
A 1-1 relationship can be modelled using one attribute with two attribute forms. This is because each Location corresponds to one and only one Airport. One possible schema set up for this would be to have the Airport column on the fact table with fact data, and a lookup table that has both the Airport and Location columns. No additional information is required to properly aggregate and display a report including Location, Airport, and fact data.
However, now consider a data model in which many Airports can exist in a given Location, and it is desired that reports display aggregated data. Strategy needs to be explicitly told that there is a Many-1 relationship between Location and Airport, which can only be done by creating them as two separate attributes. In this case, Strategy would know to aggregate Airport level fact data up to the Location level in when for example, only Location and Revenue need to be displayed on a report.
