Go ahead and open the Chapter 02 Starter.twbx workbook located in the Learning TableauChapter 02 directory and navigate to the Tableau Paradigm sheet. Take a look at the following screenshot, which was created by dropping the Region dimension on Columns and the Sales measure on Rows:

The Region field is used as a discrete (blue) field in the view, and so defines column headers. As a dimension, it defines the level of detail in the view and slices the measure such that you get one bar per region. The Sales field is a measure aggregated by summing each sale within each region. As a continuous (green) field, Sales defines an axis.

For the purpose of this example (although the principal is applicable to any data source), let's say you were connected live to a SQL Server database with the Superstore data stored in a table. When you first create the preceding screenshot, Tableau generates a VizQL script, which is translated into SQL script and sent to the SQL Server. The SQL Server database engine evaluates the query and returns aggregated results to Tableau, which are then rendered visually. The entire process would look something like the following diagram in Tableau's paradigm:

There may have been hundreds, thousands, or even millions of rows of sales data in SQL Server. However, when SQL Server processes the query it returns aggregate results. In this case, SQL Server returns only four aggregate rows of data to Tableau—one row for each region.

To see the aggregate data that Tableau used to draw the view, press Ctrl + A to select all the bars, and then right-click one of them and select View Data.

You can actually see the aggregate data that Tableau used to render the bars and the underlying records by following these steps:

1. Navigate to the Tableau Paradigm sheet in the Chapter 02 Starter workbook.
2. Press Ctrl + A to select all four bars.
3. Right-click one of the bars and select View Data... from the context menu as follows:

The View Data screen allows you to observe the data in the view. The Summary tab displays the aggregate-level data that was used to render the view. The Sales values here are the sum of sales for each region. When you click the Underlying tab, Tableau will query the data source to retrieve all the records that make up the aggregate records. In this case, there are 9,426 underlying records, as indicated on the status bar in the lower-right corner of the following screenshot:

Tableau did not need 9,426 records to draw the view, and did not request them from the data source until the Underlying data tab was clicked.

Database engines are optimized to perform aggregations on data. Typically, these database engines are also located on powerful servers. Tableau leverages the optimization and power of the underlying data source. In this way, Tableau can visualize massive datasets with relatively little local processing of the data.

Additionally, Tableau will only query the data source when you make changes requiring a new query or a view refresh. Otherwise, it will use the aggregate results stored in a local cache, as illustrated here:

In the preceding example, the query based on the fields in the view (that is, region as a dimension and the sum of sales as a measure) will only be issued once to the data source. When the four rows of aggregate results are returned, they are stored in the cache. Then, if you were to move Region to another visual encoding shelf, such as color, or Sales to a different visual encoding shelf, such as size, then Tableau will retrieve the aggregate rows from the cache and simply re-render the view.

Of course, if you were to introduce new fields into the view that did not have cached results, then Tableau would send a new query to the data source, retrieve the aggregate results, and add those results to the cache.