Using percentile charts for analyzing numeric data distributions

Keshif features a unique percentile chart option, as a rich exploratory refinement of box-whisker plots. The percentile charts present information in a compact chart area, and can effectively visualize numeric data trends across multiple data groups.

A percentile chart reveals the median (50-percentile) and other percentile characteristics (10-20)...(80-90) of a numeric data attribute.  The median line (50-percentile) shows the number for which half the records have larger value, and the other half have smaller value. Each percentile range, colored from dark to light from the median line, includes roughly one-tenth (10%) of the data. The visualization quickly reveals the skewness of the data, that is, whether larger or small values are more common.


Viewing or hiding percentile chart

Learn how to view or hide the percentile chart.


Using a percentile chart for group comparison

To compare trends across groups, simply select the data groups for comparison. Learn more.

In the chart capture below, percentile charts show that the entire apartments have higher prices than private rooms, which have higher prices than shared rooms. The bottom row shows the distribution of the whole dataset. The histogram chart above, in contrast, is less intuitive to make sure observations with confidence.


Special Features

  • The percentile plot is not linked to analytics settings, such as measurement mode or breakdown mode. For example, if average rating is the measurement for aggregated bins in the histogram, the distribution of the percentiles will continue to be based on the count-based simple distribution of the numeric attribute.

  • The percentiles share the same value axis with the histogram above it, including binning scale type (linear or log).