Learn how to use Narrative's Data Discovery Assistant to quickly uncover new sources of predictive data
The Data Discovery Assistant leverages seed data in the form of either User-ID List (MAIDs only) or App List (URIs / bundles) to help uncover high-value apps. Simply upload a list and Narrative will analyze & produce a report to help discover attributes with the highest correlation to the seed data. Data Discovery Assistant currently supports analyzing MAIDs & Apps only.
To start, create a New Report and choose either a MAID or App list for Narrative to analyze as your seed data. Select an existing MAID list in your account or manually input a list of App URIs.
New Report - App
New Report - User ID List
- The first view provided is the Scatterplot
- The X axis (ID Count) represents how many MAIDs are associated with each data point
- The Y axis (NPMI) is the Normalized Pointwise Mutual Information
Pointwise Mutual Information is a way of measuring the relation between two
events. When normalized (as in NPMI), this relation is represented by a
number between -1 and 1. When knowing that one event occurs gives you more
information about the likelihood of a second event occurring, the score
between the two events moves closer to 1. A score of 0 indicates that
knowing about one event tells you nothing about the likelihood of the second
- The Data Points on the graph represent an Application that falls between the values of the X & Y axis.
- Click a specific app to get more information associated with that data point, including ID Count & NPMI.
- Adjust Values of the X & Y axis to change the view of the data. Use either slider to focus on data points that fall between a specific NPMI or ID Count range.
- Table View provides the same data but in list-view with both NPMI Score & ID visible, can be sorted by Name, Score, or ID Count
- Click an App to see the App Store or Google Play page
- If an app does not display an icon or link, the app is no longer available for download
What does a negative NPMI score mean?
A negative score assigned to an app on the scatter/table view means that the seed data being analyzed (MAIDs or Apps) is less-likely to use that app.
Can I change what type of data I see in the report?
Currently, the Data Discovery Assistant uses Apps as the main data point of comparison to the seed data.