Oh yeah – there are more than one!
Use Jump Plot to discover execution trends, benchmark metrics, and highlight bottlenecks.
Check out some variations that you can use to tweak your plots.
Jump Plot – The Original
Compare repeated executions of event sequences against one another, looking for outliers.
Checkpoints are visualized with circles on the x-axis. Checkpoints may be organized with:
- Static Placement – Checkpoints are placed equidistant and spacing of checkpoints has no significance
- Variable Placement – Checkpoint spacing is based on a measure value (e.g. average time between checkpoints).
Hop height is visualized with Bezier curves connecting checkpoints across the sequence. It is not required to have a hop across all checkpoints. Height of hops are usually based on a measure of time (e.g. milliseconds, seconds, minutes, etc) compared to zero (x-axis). Hops can also be based on volume instead of a measure of time. The standard Jump Plot will identify out of sequence hops below the x-axis.
Threshold-Based Jump Plot
Benchmark event executions against expected values on a per-hop basis to find outliers.
In place of zero, a threshold metric is used as the x-axis measure. Thresholds can vary by hop, series and/or view as desired. When thresholds are used on the x-axis, the reader should know (or be shown) the threshold measure values. Hop height – Similar to Jump Plot specifications, however hops are compared to associated threshold metric values instead of zero. Hops will be shown both above (larger than threshold) and below (smaller than threshold) the x-axis. Hops on the x-axis are equal to the threshold value.
Sift through high-volume event sequences to find macro trends.
Hops within in a jump plot can be aggregated based on quantity or volume measures. In this variation of the Jump Plot, measures are aggregated (e.g., summarized, averaged, etc.) for each instance of duration and hop. The aggregated amounts can be used to vary size and/or color of the hops within the Jump Plot to assist the viewer. Aggregation can be applied with all variations of Jump Plot. Aggregation is most beneficial when plotting high volume sequences.
Compare event sequences which have varying durations or volumes between checkpoints.
Percentage transformation can be applied to both Jump Plot and Threshold Based Jump Plot. In this variation, hop heights are transformed from actual values to a percentage of corresponding values. In Jump Plot, the values are usually shown as a percentage of total of the maximum hop value (although this may have to be adjusted if skewed).
In Threshold Based Jump Plot, the values are usually shown as a percent change against the threshold value.
Slice and dice your events into panels for more granular analysis.
Each of the above variations can be displayed in a trellis panel view. As with all panel views, it is recommended to have a lower volume of panels (e.g., less than 50) within a Jump Panel. If possible, panels should also be sorted by an aggregated metric (e.g., average duration per series).
Evaluate macro and in-line trends through a high data density, simplified version of Jump Plot. Refer to this write up from Edward Tufte on the value achieved through in-line visualizations (e.g., spark lines and whisker plots). Jump Line is to Jump Plot as spark line is to line graph.
This simplified version of Jump Plot will incorporate three or less series (usually just one) to evaluate macro and/or in-line processing trends. When working with in-line views it is also beneficial to add indicators to Jump Line in the form of colored hops or an additional indicator mark. Jump Line can leverage basic, threshold based and percentage transformation Jump Plot variations.
Data from ESPN Stats & Info Group.