✍ The Tableau 2020.2 release introduced a feature for data modeling called Relationships as seen below. Relationships are another approach to combining data from multiple tables for analysis on matching fields. Relationships have semantic behavior with Smart Aggregations, and Contextual Joins. Smart Aggregation allows a measure to aggregate to it's level of detail of their pre-join source table. Contextual Joins allow unmatched values to be handled individually per viz. Tableau provides the below description of a Relationship.
☕ Buy me a coffee! 🙏 Glad my article helped you!✍ In Tableau, Blending is a method that helps combine data from multiple data sources. The secondary data source brings additional information and displays it directly along with data from the primary data source. Blends never truly combine the data but queries each data source independently. The final results are then aggregated to an appropriate level even though the sources might be at different levels of detail. The primary data source is indicated by blue, and the secondary by red as seen below.
✍ It is advisable to always avoid Blending when possible and opt for Relationships. With Relationships, you can have unmatched values from two sides, or you can have inequality joins, or do distinct counts, and relate as many tables as you like. With Blending, you will encounter limitations soon and it is similar to using a VLOOKUP. Relationships, on the other hand, are contextual database joins. These are some of the key differences between Relationship and Blending.
✍ With a Relationship instead of a Join when using two tables, Tableau first performs the aggregation of the measures separately and then joins them in the views or calculations based on the Relationships we specify. All metrics are calculated at the level of detail of their source table. Performance-wise also it is optimal since Tableau will only perform the join if you are building a view or calculation that needs both the tables. Whereas Blends can be termed as a limited version of Relationships, and they can be useful for quick cases and scenarios. But again, we cannot Relate or join published data sources, so our only option for such cases is Blend.