Quick answer: which one should you use?
- Relationship: the default. It links tables on matching fields but keeps them separate, so Tableau joins at the level of detail of each view. No duplication. Use it first.
- Join: merges columns from two tables into one table before analysis, using inner/left/right/full outer. Rows can duplicate when the match is one-to-many.
- Union: stacks rows of tables with the same structure, for example Sales Q1 + Sales Q2.
- Blend: combines two separate data sources per sheet (primary and secondary), behaving like a left join on linking fields. It’s legacy; use it only when you can’t relate the data.
Since Tableau 2020.2, dragging a second table onto the canvas creates a relationship (the ‘noodle’) by default. Joins and unions still exist: double-click a logical table to open its physical layer, and build them there. Here is what you see on the data source view:
Double-click a logical table to get to the physical layer, where you can create joins and unions:
Tableau says that relationships have mostly superseded blends. Relationships allow you to combine also avoid data duplication but are more flexible.
| Combines columns or rows? | Match on key field duplicates data? | Limitations | Tableau Documentation | |
|---|---|---|---|---|
| Relationship | Combines columns | No | Relationships cannot be formed between tables from data sources published to Tableau Server or Tableau Cloud. | Relationships |
| Join | Combines columns | Yes | Cross database joins will not work with all data sources | Joins |
| Blend | Combines columns | No | Fields from the secondary data source cannot be brought in to the view unless there is a match on the key field. This makes a blend somewhat comparable to a left join, since data from the primary data source is always brought into the view even if there is no match to the secondary source.
Blends are only able to combine two tables, a primary and secondary data source. More information on limitations of blending here here |
Blends |
| Union | Combines rows | No, since unions do not use key fields. | Data sources used in the union must come from the same connection | Unions |
Join Vs Union
Joins combine columns from one table with columns from another table. Here’s an example of how a join looks in Tableau:
And here’s what the data looks like after the two tables have been combined using the join. Notice that the first three columns come from the Book table, while the next three columns come from the Award table.
Unions combine rows from one table with rows from another table. Here’s an example of how a union looks in Tableau:
And here is what the data looks like when you combine Sales Q1 and Sales Q2 with a union. Notice the three of the rows come from Sales Q1 and three of the rows come from Sales Q2
Join Vs Relationship
When tables are joined, the result table has a row each time the join criteria is met. For example, if we have two tables, one with transaction data showing purchases:
and a second table with Monthly Sales targets, with one row per month:
Let’s see how these are combined in Tableau, first using a relationship. I use datetrunc so that 1/14/2022 becomes 1/1/2022 so that it matches to the value in the Monthly Sales target table.
Here’s my view in Tableau:
Now try the same but with a join. Notice that I have the option of Inner, Left, Right, and Full Outer, which I didn’t have when I did the relationship.
Also notice the data preview area. The values of Sales Amt Target are replicated. The values from the sales targets are replicated for each row of the purchase data.
If I don’t do anything to handle this duplication, and simply apply a sum to the Sales Amt Target field, I end up with inflated values for Sales Target:
So this illustrates a couple of the key differences between joins and relationships:
- Joins give the option of different join types – inner, outer, left, and right – which handle nonmatches differently
- Joins produce a row each time the join criteria is met, which can cause duplication. Relationships do not produce duplication.
You can download the Join vs Relationship workbook here: https://public.tableau.com/app/profile/lukas.halim/viz/JoinVsRelationship/Join
Types of joins in Tableau
| Join | Keeps |
|---|---|
| Inner | only rows that match in both tables |
| Left | all rows from the left table + matches from the right |
| Right | all rows from the right table + matches from the left |
| Full outer | all rows from both, with nulls where there’s no match |
How to create a join: drag the first table to the canvas → double-click it → drag the second table next to it → click the Venn icon and choose the join type and the join clause fields.
Why joins inflate your totals (and relationships don’t)
Join one-row-per-month targets to many-rows-per-month transactions and every target value repeats on each transaction row, so SUM([Target]) comes out far too high. Fixes: use a relationship, or with a join use {FIXED [Month] : MIN([Target])} or switch to MIN/AVG.
Blend vs relationship
Blends are set per worksheet, need a linking field, and only bring in secondary fields that match (shown with an orange link icon). Relationships are defined once in the data source and work in every sheet. Use a blend when the sources can’t sit in one data source, for example two published data sources.
Practice what you learned
- practice questions on joins and blends
- relationships quiz
- Data Analyst practice: multiple data sources
- Desktop Specialist study guide (domain 1)
FAQ
What’s the difference between a join and a relationship in Tableau?
A join physically merges tables into one and can duplicate rows. A relationship keeps tables separate and queries them at the right level of detail for each view, so it avoids duplication.
What’s the difference between a join and a union?
Joins add columns side by side, matched on a key. Unions add rows, stacking tables with the same columns.
Are data blends still used?
Yes, but mostly for combining separate published data sources. Relationships replaced most blend use cases.
Which is on the Tableau certification exams?
Both exams test it: Desktop Foundations objective 1.2.3 (“when to use a join versus a relationship”) and Data Analyst 1.2.4 / 1.3.2–1.3.3 (relationships, joins, unions, and Prep joins/unions).