Have you examined this article in the Knowledge Base about defining custom joins with the custom SQL option
http://kb.tableausoftware.com/articles/knowledgebase/custom-sql-join? It may point you in the right direction with your specific data. Additionally, there is another forum topic that has been answered about joining multiple SQL Server tables, which may help you: http://community.tableau.com/message/122497#122497.
I had read both of those pages, in fact I'm following their instructions for joining tables. I feel like I'm missing a key piece of Tableau. If I had to create duplicated data sources every time I need to join tables, using this on a larger database, with several dozen tables would be complicated.
I read this article:
It directs you to have rows with only piece of information instead of multiples. In my current workflow following the join instructions, I am left with underlying data that is similar to this.
Thanks for the information!
I also attempted a blended view (based off of one of the above links recommending using multiple identical data sources) of my data but it wouldn't allow me to continue because it knew it was from the same data source, so that alternative doesn't seem to work.
1 of 1 people found this helpful
I think the key piece you think you were missing was, in fact, data blending. When data blending occurs, Tableau issues a warning saying that a join is also possible if it is from the same data source. This is simply to offer an alternative, not to say that blending is necessarily an incorrect approach in a particular case.
When data is joined, not on unique identifiers, data gets brought in multiple times, thus the inaccurate aggregates (provided the field isn't dimensional and the multiple data situation isn't desired). To handle this, Tableau has data blending, where the goal is to associate the aggregates together on common dimensions rather than to join them at the database level.
This comes down to a row level vs. aggregate join scenario, where in a data join, this is done at each individual row, where blending deals with an aggregate first, then a join when using a common dimension. This blend should provide for correct (unique) aggregations. The query you proposed accomplishes the same thing as data blending.
A good KB article on joins vs. blends can be found at http://kb.tableausoftware.com/articles/knowledgebase/join-vs-relationship-60. Please let me know how this works out for you!
can you provide sample data or workbook ? And what type of join you are using in your case, with some more knowledge of data , i think we will find a way out easily.