Tableau has a few different options when it comes to choosing your data model. This can be quite confusing because there’s usually not a “right” option and it ultimately comes down to your own judgement. In this blog I’m going to discuss the different data model options within Tableau and give you a few pointers as to when these options are most appropriate.
The Physical Layer:
Consider joins when…
- …data relates one to one (i.e. 1 row in Table A matches 1 row in Table B)
- …you have data keys or predefined joins within a schema.
- …the data doesn’t require level of detail calculations to prevent duplication of rows.
- …using a main table and lookup table(s) to bring in additional data.
Consider unions when…
- …data tables stack on top of each other, make sure that the number of columns and the columns themselves are identical (note that you can always add columns with null rows if you need to union mismatching tables).
The Logical Layer:
Consider relationships when…
- …data doesn’t relate one to one (i.e. 1 row in Table A matches to N row(s) in Table B).
- …you require a full outer join between data at different levels of aggregation.
- …you are using a relational database with predefined relations.
- …using a main table and lookup table(s) to bring in additional data, but you don’t necessarily need to have all the data at once (improving performance for large datasets).
Consider blends when…
- …you have tables at different levels of aggregation (e.g. daily and monthly data) and want to have separate datasource extracts in your workbook.
- …you’re not using Tableau 2020.2 (when relationships were added) and need to relate data which exists at different levels of aggregation.
- …you want different relations based on the chart you’re building (relationships apply to the whole workbook whereas blends only apply to a sheet, making them more flexible).
If you’re ever in doubt the safest bet is to use relationships to avoid exploding your data, I hope you find this blog useful!