Connection types
The connect pane splits into To a File — Excel, text, JSON, PDF, spatial, statistical — and To a Server, covering the databases and cloud services.
Each database has a native connector, which is faster and more capable than a generic ODBC connection. Use the native one where it exists.
You can also connect to a published data source already on Server, which is usually the right answer in an established team — someone has already modelled and governed it.
The data source page
Everything you do here applies wherever the data source is used, which makes it the cheapest place to fix things:
- Rename fields to what a business user would call them
- Hide fields nobody will need
- Set data types and geographic roles
- Create calculated fields shared by every workbook
- Set default formatting — currency, decimal places
This is the cheapest performance improvement available in Tableau, and the data source page is where you do it.
Custom SQL
The connection pane allows custom SQL. It works, and it is usually the wrong tool.
Tableau wraps your query as a subquery, so it cannot optimise through it — filters and aggregations that would normally be pushed to the database are applied afterwards. On a large table that difference is dramatic.
Where logic is genuinely needed, a view in the database performs far better and can be reused.
Where preparation belongs
| Work | Belongs |
|---|---|
| Joins, renames, hidden fields | Data source page |
| Repeatable multi-step cleaning | Tableau Prep or the warehouse |
| Business logic used by several teams | The warehouse |
| View-specific calculations | The workbook |
The general rule: the further upstream a transformation happens, the more places benefit from it.