Explain
Explain is an experimental feature that is disabled by default. You can enable explain from the Advanced tab in the Settings.
Explain allows you to select a section of data and receive automated, machine driven, explanations of what is contributing to the data that was selected.
On a line chart you might see a spike, a dip, a rise, or a fall. Using explain you can figure out which values in which dimensions are responsible for the observed trend.
Clicking on the star button in the highlight menu activates the Explain feature.
After selecting a feature of interest (a spike in the above example), a series of explanations are presented with possible actions.
In this case we can exclude the user Bearcat
and immediately we see that the selected spike disappears.
Explain is not limited to the line chart. You can select any slice of data of interest and receive a list of driving factors.
Explain also has a specific mode to work when time comparison is enabled - Explain will analyze the difference between the current values and the comparison values.
Similarly in a dashboard, selecting a section of data and clicking on the star button will activate the Explain feature
Explain is domain agnostic, it builds a model of your data on the fly when activated and can therefore be used within any filter.
There are three algorithms within Explain:
- Change compare - suitable when a comparison is chosen. It compares and finds the contributing factors in the current value vs the comparison. Change compare is automatically chosen when compare is selected.
- Trend analysis - suitable when selecting a trend such as a spike, dip, rise, or fall. It will automatically detect the trend and then find the contributing factors.
- Breakdown - suitable when selecting a singe slice of data (without specific reference). It will find the contributing factors to the selected slice.
The appropriate algorithm is automatically chosen given the selected data and other factors such as the filters and whether a comparison is used.