Understand Visualizations Generated by Contextual Insights

The Contextual Insights algorithm generates three types of visualizations by leveraging the related columns.

To generate these insights, the algorithm selects a subset of related columns that show the highest contrast between your selection and the rest of the data. The order in which the insights are displayed depends on the data. In each insight, your selection appears in orange while rest of data appears in blue.

  • Breakdown by attribute - These dual-axis visualizations provide a breakdown of your metric across various members of a selected attribute in your data. The bars represent the breakdown of the selected data and the line represents the breakdown of the remaining data. The overall shape of the distribution helps you quickly see which members in your selection contrast significantly with the rest of the data. The members are sorted from highest to lowest based on the rest of the data.

    For example, this insight breaks down sales by product container. The description provided indicates that in the case of the Small Box product container, the selected data has significantly lower sales when compared to the rest of the data. It also points out that the Jumbo Drum and Jumbo Box product containers for the selected data have much higher sales than the rest of the data.
    Description of ci_breakdown_example.png follows
    Description of the illustration ci_breakdown_example.png

  • Trending - Trending visualizations compare the relative growth of a metric over time for your selection and the rest of the data using lines to show the evolution of each. The algorithm uses the first date column that appears in the Related Columns section of the Grammar panel.

    Each line starts with a base index value of 1.00 set at the initial time period. The evolution of the metric over time shows the relative value in the following periods compared to the index value of 1.00 in the initial period. When looking at absolute values for the metric, discrepancies in values make it difficult to properly compare any growth or decline over time.

    For example, this insight shows the trend in sales by ship date. The description provides additional insights about the overall performance of sales over time. The description also highlights the intervals in the data where there's a significant difference in the trend between your selection and rest of the data, in this case, 2014 to 2015.
    Description of ci_trending_example.png follows
    Description of the illustration ci_trending_example.png

  • 80/20 - This type of visualization shows what proportion of your metric value consists of the top 20% of the records and what proportion consists of the bottom 80% of the records when data is ordered by your metric. The visualization also shows the same for the rest of the data. This is computed at the most granular level of data in your source visualization.

    For example, this insight shows the 80/20 proportions, sorted by sales, using two bars: the first for the rest of the data and the second for your selection. The description highlights the fact that the proportion is noticeably different between the two.
    Description of ci_80_20_example.png follows
    Description of the illustration ci_80_20_example.png