Overview
Distribution heatmaps provide benefits above and beyond histograms.
Histogram charts are excellent for understanding the distribution of a given metric on a single dimension or window of time. However, understanding how the distribution changes across dimensions require analysis of multiple histograms at once.
Distribution heatmaps solve this issue by merging and flattening several histograms into a single, easy-to-understand visualization. The heatmap is colored by the density of test runs with similar performance, building a straightforward way to understand if the distribution of test runs varies across dimensions.
How to Use
Use this chart type to spot patterns and abnormalities based on time, location, ISP, day of the week, and more. If the density of test runs shifts to a particular cell, it is an immediate indication something has changed. Distribution Heatmaps are available in the Analysis section as well as dasboard widgets.
Once in the Explorer module, select hte Heatmap button within the Visualization section.
By default, the # Buckets will be 10 and Type will be "Linear". Yu can change these settings to alter then number of value-ranges that your data is broken into, and how those ranges are distributed.
The distribution heatmap can be generated using any one of the standard metrics at a time. Below is an example of a heatmap chart broken down first by Test over the last 30 days. It is apparent a heatmap makes it easier to analyze lots of data points over a longer period of time with the ability to notice any trends.
In the example below, the arrow is pointing to a darker shaded cell.
Hovering over that cell will display the Upper Bound, Lower Bound, # Runs and % Runs in the tooltip. In the example below, for the bucket with an Upper Bound of 1,120 ms and a Lower Bound of 960 ms, there were 10 runs or 83.33% of the runs
Use Cases
Use Case 1
The line graph above paints a picture where there is consistent performance below one second, however there are customers sending complaints describing intermittent issues. If the data points above are re-drawn into a Distribution Heatmap the user is then able to see that there are hidden outliers, which explain the customer complaints. This Distribution Heatmap can be seen below.
Use Case 2
If there is a test that runs very frequently, it can be difficult to find the right breakdown in order to view all of the individual data points on a scatter plot. This is where the Distribution Heatmap can be useful. For example, the chart below has more than 500 individual data points per hour. The Distribution Heatmap displays trending data in a clearer format when compared to a scatter plot.
