UPCOMING WEBINAR: Cluster analysis: Classic, powerful, and still worth knowing Displayr automatically groups variables into variable sets, which speeds up and simplifies the process of creating tables and visualizations. However, variable sets can contain many variables that are not shown in the table or visualization (e.g., the corresponding categories may be hidden). Greater speed can be obtained by creating a new variable set just for the table or visualization, or, by removing irrelevant variables from the variable set.
Worked example
Consider the funnel chart shown below. It shows the data for Pizza Heaven. If we only want to see this visualization and don't need to see the table, then creating a visualization in this way will have a performance cost.
The table above was created from 60 variables. However, only four of them are required to perform the visualization, so every time the data is updated (e.g., filtered):
- 56 unnecessary variables are retrieved.
- These variables are added to a cache so they can be more quickly retrieved in the future. This can cause other data to be pushed out of the cache, causing other analyses to be slower.
- The table is calculated using all 60 variables, rather than the 4 that are required, making the calculation slower.
- The whole table of data needs to be transferred and shown on the users' screen, which will take longer than a smaller table.
A much more computationally-efficient approach is to create a new variable set using just the variables required for the visualization, create a table from them, and then hook the table up to the visualization:
Note that this is an example of the more general strategy of Minimize the Size and Distance of Data Being Moved.