Design Considerations for Trellis Views and Microcharts
The following are some ideas to consider when designing content displayed in trellis views:
For comparisons, choose the Simple Trellis subtype.
For trend analysis, choose the Advanced Trellis subtype.
The inner graphs that make up a trellis should be readable and not too dense, so a trellis view is not especially useful for displaying multiple series or multiple groups. If you cannot easily target a data point with your mouse (to be shown a tooltip), then it is likely that the inner graph is too dense to be readable.
- When using the Simple Trellis subtype, note the following:
Designing a simple trellis is like designing a pivot table, except that the total number of cells that can be rendered is much less for a trellis.
The main difference between designing a simple trellis and designing a pivot table is that for a trellis, one or two of the dimensions can be associated with the visualization; so, that many less dimensions must be added to the outer edge.
It is best to design the trellis with a small number of outer-edge dimensions. The entire graph series should be visible at once (for easy comparison of like to like) with no need to scroll. If you must show additional dimensionality, consider adding the dimensions to the graph prompt.
When determining which data to show in column headers and which to show in row headers, the column headers should show one or two dimensions (each dimension with a small number of members). Most often, the dimension shown in column headers is time. Place the remaining dimensions in the row headers or in graph prompts.
- When using the Advanced Trellis subtype, note the following:
The key use case for an advanced trellis is to show trend graphs alongside numeric values, in a compressed form. So a typical advanced trellis contains a combination of spark graphs alongside number representations of the same measure.
Ideally, place no dimensions in the column headers, just place the measures here.
The dimensionality typically associated with a spark graph is time. As there are no visible labels in a spark graph, it is important that the data visualized is intrinsically ordered. For example, a spark graph visualizing regions would be meaningless, because the ordering of the regions (which would be the specific bars, in a Spark Bar graph) is unintuitive.
Just as when designing pivot tables, you generally display time on the horizontal axis, with the other dimensions displayed on the vertical axis. The eye then scans from left to right to see how the dimensionality changes over time.
When Might a Trellis Not Be the Best Visualization?
Hierarchical columns do not work well with the Simple Trellis subtype, because when a hierarchical column is displayed on the outer edge, parents and children (such as Year and Quarter) will by default be shown using a common axis scale. However, because Year and Quarter have different magnitudes, the markers in child graphs may be extremely small and hard to read against the parent scale. (Hierarchical columns do work well with the Advanced Trellis subtype, however, because each data cell is a different scale.)