Analyze the distribution of numeric properties that describe people, their behaviors, and how they traverse through your experience using Interana distribution view.
Why use distributions? Sometimes it’s not enough to know what the average user is like in your experience. For example, if you have different user segments who exhibit different behaviors, you might find that there are different modes according to your usage metrics. Or you might want to look for power users who act like extreme outliers. Distributions can help illuminate these cases.
Use distribution view
To get started:
In the Interana UI, navigate to Distribution.
In the Event property field, pick an actor property, event property, or flow property.
In the Starting and Ending fields, select a time range you want to examine that property over.
Hit GO. Interana suggests a default binning and displays a histogram.
For example, say your data set has an actor property BodyHeight. In this example, select the BodyHeight actor property. When you press GO, you’ll see a histogram of your user height.
Adjust the range
To look at one part of the distribution, you can apply a filter to your property. For example, if BodyHeight is in units of inches and you want to look at people greater than 6 feet tall, apply “Filtered to BodyHeight is greater than 72” to the original binning property.
Or, to zoom in on any particular bin, click the bin and then click Zoom in. Interana applies the relevant filters in your query.
Either way you adjust the range, to zoom out, delete the value filters for your binning property.
Note that ranges (and percentiles) identify inclusive and exclusive elements. A bracket indicates an inclusive range, and a parenthesis indicates an exclusive range. For example, integers described with this notation are as follows:
- (0, 3) = 0, 1, 2, 3.
- [0, 3) = 1, 2, 3.
- [0, 3] = 1, 2.
To adjust the binning while you are zooming or adjusting the range, click Modify Bin+Measure.
You can adjust the bin size or count. When you pick one, Interana automatically recalculates the other based on the range of values in your bin property and applied filters.
When investigating event properties, decide whether you want to distribute all instances of an event or unique attributes.
For example, consider an event property called SongLength.
If you are interested in the distribution of songs played over the past couple weeks, irrespective of the unique songs, then distribute across SongLength, using the default query.
However, if you are more interested in seeing the distribution of unique songs, then click Modify Bin+Measure. The default measure specified is count of events. To count the unique songs played, in the Measure field, select Song. As a result, you will get a distribution of SongLength for unique songs.
Calculate binning properties
You can go beyond counting unique things: you can sum or average different properties over your binning property.
Let’s keep our original example binning property: SongLength. To learn how SongLength relates to total play for all song plays, adjust your measure to sum of SongLength. Your resulting distribution might not look like a Gaussian curve anymore, but you might be able to discover whether your users spend more time playing shorter or longer songs.