Tableau user here. It took me a while to get my head around the concept of'data blending' as the documentation a while back was aimed at non-technical users (the beauty of Tableau is that it's so easy to pick up and start making data visualisations with). I'm not sure about Alteryx, but in Tableau, a blend is fundamentally just a left outer join (so everything in the left table is always returned by the query). The clever bit is that Tableau will aggregate the data in the left table based on the fields (and therefore level of granularity) you choose from the right table(s) before the join is done. Hence, data blending is very useful when you want to join datasets that are at different levels of granularity. One thing to keep in mind though is, as an outer join is done between the left tables and right table(s), performance can quickly reduce as the sizes of the table(s) in your query get bigger.
Just my 2 cents of managing a an analytics team that works daily with Tableau and Alteryx: 1) Alteryx is super expensive and can be replaced with Jupyter Lab and Python, those that do Alteryx well can also do basic data prep in python; 2) forget Tableau for anything other than fancy, online, dashboards (so forget pivot tables, email notifications and pdfs that look good)
There is no good reason to ever use Alteryx over R/Python. But their company is really good at selling to non-technical management.
what do you mean by "forget pivot tables, email notifications and pdfs that look good".
they don't work? are expensive? hard to use? limited?
Tableau straight out refuses to make tables that look good because they want users to do charts. So if your users want cross tab looking reports like Pivots, Tables, etc., you'll be left with a very limited solution. Tableau does also very limited notifications and those are normally "subscribed" by users, not centrally managed. Finally, no pdf or print out or screenshot ever looks as good as those from Excel, I don't know why, but they know it is a problem and they still think it's ok. In summary Tableau is about the "Tableau way" which is great if you have the right type of user, but completely sucks if you don't. For instance, if you are trying to move people away from Excel, Tableau is too much of a leap and will hurt your adoption of an online solution. Things like "SeekTable" are better if all you need is moving away people from crazy Excel files.
In my experience doing data transformations in Tableau is almost always a mistake. It's opaque what's actually happening and you can't version control the actual operations. Instead, preprocess data into a view in your database or with a script and read that into Tableau.
Not sure why they've grouped together Alteryx and Tableau - they're completely different pieces of software used for completely different purposes. By the time your dataset gets to Tableau it better be ready for visualisation with minimal changes.
I'm a QlikSense developer, so I've always wondered about the other technologies in the BI and Data Engineering Realm. It's hard to do what you can do with programming languages in trying them out, because they are all so expensive.
Alteryx / Tableau is indeed a powerful combo. But it is also bloody expensive.
Correct, I am definitely going to replace Alteryx with Jupyter / Python and maybe with vega we could replace Tableau at one point too.
Just learned about vega from your comment. It seems very cool (coming from a person who loves D3). But replacing Tableau with a higher level D3... seems like you'd lose in man hours what you make up in subscription cost savings...
Vega isn't really a dashboarding solution. Look into Dash or Bokeh instead.