Benchmarks display qualitative differences in data by comparing the distribution of a metric based on all of the data with the distribution of a metric based on a selection in the data.
They provide a quick reference of normal behavior by illustrating what a visualization would look like if no selections were made. The differences in the data are based on relative values, not absolute values.
Benchmarks can represent differences in data for the dataset or for any data subsets that you create. For more information about subsets, see Working with Subsets.
The following graphs show the sessions on a website by day of the week and hour of the day. The visualization on the left provides a good representation of the pattern that the site traffic takes over time. To compare the overall site traffic pattern to the traffic pattern on weekends, you would select Saturday and Sunday in the Sessions by Day of Week graph as shown in the visualization on the right.
After a selection is made, benchmarks (the thin white vertical lines highlighted above) appear in the Sessions by Hour of Day graph on the right. From the benchmarks, we can see that the traffic pattern on weekends when compared to the overall traffic pattern generally includes fewer sessions earlier in the day (08:00 to 11:00).
The Radar Visualization is also useful for analyzing benchmarks. See Radar Visualization.