Significance analysis compares metrics for a selected audience with the entire population of a specified user tag dataset. The following figure shows an example.
The audience for the analysis must be created using Tag Filtering, or using cross-filtering or audience intersections and unions that include tag filtering. Otherwise, the analysis results will not contain any data.
Unlike audience profile metric analysis, significance analysis shows only the top 10 tags that have significant differences between the audience and the entire dataset population. For example, consider two tags. For the "Gender" tag, the male-to-female ratio is 1:1 in the entire dataset population, but 9:1 in the audience. For the "Region" tag, the Beijing-to-Hangzhou ratio is 1:1 in the entire dataset population, but 4:6 in the audience. When you compare these two tags, "Gender" is the one with the significant difference.
Procedure
Select an audience to analyze. You can do this in two ways:
Go to Workspace > User Insights > Audience Management > Audience Analysis. On the Audience Analysis page, select an audience from the Audience drop-down list, as shown in the preceding figure.
Go to Workspace > User Insights > Audience Management > Audience Management. In the audience list, click the
icon next to the audience to go to the Audience Analysis page.
On the Significance Analysis tab, select a user tag dataset (including custom tag datasets) from the panel on the right. The dataset must use the same data source. If the dataset contains tags associated with the audience, these tags become available as analysis metrics.
NoteNot all tags in a dataset can be used for significance analysis. The available tag types are enumeration and multi-value. Multi-value tags can be used only when the number of tag values after splitting and deduplication does not exceed 50. When you create a dataset, all available tags are enabled for profiling by default. You can also disable specific tags for profiling when you edit the dataset. For more information, see Manage user tag datasets and Manage custom tag datasets.
Only the top 10 tags with significant differences between the audience and the entire dataset population are displayed.
From the Search Tags drop-down list, select the tags to analyze. The corresponding analysis charts appear below.
NoteYou can also select tags from the tag list below. The tag list can be sorted by total tag popularity, profiling frequency, filtering frequency, or push frequency. When you hover over a tag, the corresponding data is displayed.
The system automatically selects a suitable chart based on the number of tag values. The rules are as follows:
Number of tag values
Select a chart
<=10
Column chart
>10
Condensed column chart
You can click the icons in the upper-right corner of a chart to perform operations. These operations include filtering out null values, switching the chart type (column chart, pie chart, or treemap chart), sorting (default, descending, ascending, or custom), viewing data, exporting, zooming in, and closing the chart.
Filtering out null values excludes records with unknown (null) tag values from the analysis, as shown in the following figure.
Only column charts support custom sorting. After you select Sort > Custom, you can drag the items in the dialog box to adjust their display order in the column chart, as shown in the following figure.
NoteCustom sorting rules are specific to each user and cannot be shared.
Hover over a shape in the analysis chart to view the number of users and the percentage for that segment, as shown in the following figure.
