Smart Discovery Results
Smart Discovery generates results that provide you with an overview of your data.
Smart Discovery results are displayed in the following pages:
-
Overview
-
Key Influencers
-
Unexpected Values
-
Simulation
Smart Discovery Overview
The Overview page provides visualizations to summarize your Smart Discovery results.
The Overview page is split into two halves. The first half provides a summary of your data. The visualizations in the second half of the page change according to the type of data you've selected.
Even if the Smart Discovery's insight quality is considered to be poor or low, this page is still created, as it's just an overview of your data.
Smart Discovery Key Influencers
About This Smart Discovery Section
The About this Smart Discovery section displays an overview of the analysis. Included are the specific date of the snapshot of the customer data used to run the Smart Discovery, and the columns that have been excluded from the analysis. Removing the highly correlated columns improves the overall quality of the Smart Discovery results. The overview also indicates the quality of the Smart Discovery results which is assessed by the underlying machine learning model.
All the relevant measures and dimensions are aggregated to the level of the entity. For dimensions that have multiple values for a single entity, the number of dimension members of each entity is counted.
How to Interpret the Charts Section
The How to interpret the charts section explains the four displayed charts based on your live data. Use the first two charts to select a key influencer and analyze how it impacts on the entity's Target. In the second set of charts, you can select an additional key influencer to better understand the relationship between the influencers, and their respective impact on the Target.
Key Influencers Section
The Key Influencers page lists (ranked from highest to lowest) up to 10 dimensions and measures or accounts that significantly impact your target in relation to your entity.
The key influencers are displayed in a table and sorted by their degree of influence from highest to lowest:
-
Strong
-
Moderate
-
Weak
The second column indicates the name of the dimension or measure determined to be a key influencer, and the last column shows the dimensions or measures correlated with each influencer.
We aggregate all the relevant measures and dimensions to the level of the entity. If one of your dimensions has several values for one entity, we count the number of dimension members of each entity. We rename these types of dimensions as Number of <Dimension>, where <Dimension> is the original name in your dataset.
Smart Discovery Unexpected Values
The Unexpected Values page includes a table listing existing values, and the predicted values are displayed along with the other corresponding dimensions.
To add or remove columns from the displayed table, select Edit Columns.
This page isn't displayed if the insight quality is considered to be too low or there are no unexpected values. The underlying machine learning model assesses the quality of the analysis. Each chart has an insight quality checkbox indicating when the quality of the Smart Discovery results is good. If the underlying machine learning model analysis indicates low confidence in the analysis quality, you're also alerted in the chart.
Two additional interactive charts (scatterplot and bar/column) are provided to help visualize the difference between the predicted and existing values listed in the table. Use either the Search or Edit Columns options to focus or filter out columns from the table. When you sort the table on a particular column, the columns in the bar or column charts reflects this sort order. Any changes in the table are automatically reflected in the charts.
Smart Discovery Simulation
The simulation view helps you understand how changing the values of your key influencers could have an impact on the value of your target measure in relation to your entity. For more information on the simulation view, see Smart Discovery Results: Simulation.