Understanding the Recommendations

When the Recommendation Engine reports its findings, notice the features of the interface and the actions you can now perform:

initial recommendation screen
  1. To choose another app, click Back To App Selection.
  2. To start implementing the advice of the Recommendation Manager, click Create Recommended Analytical Views.
  3. To ensure that the analytical views are ready to use, click Refresh New/Stale Analytical Views.
  4. Status of Selected Visuals table shows the ID and Title of each visual and filter in the dashboard, the Status of the recommendation (green for 'existing', blue for 'recommended', red for 'not provided'), and the clickable link for SQL.
  5. To see the SQL definition of each artifact on the dashboard, click the corresponding SQL option in the Status of Selected Visuals list. Here, we are showing the SQL definition of the date filter p_dt_date.

  6. Recommended Analytical Views table shows both existing and recommended analytical views, and has the following information columns:

    • State (Up to date, Stale, Needs initial refresh, Invalid, or Recommended)
    • Analytical View Strength, a scale of 0 to 5 stars, from week to strong
    • Actions available for analytical views
    • Details

    Note that when dashboards use multiple base tables, the recommendations group accordingly.

  7. You may rename the recommended analytical views by double-clicking on the name. This is not possible for existing analytical views.

  8. The available Actions for each analytical view appear on each status line. They are:

    • Create view
    • Refresh view
    • Delete view
    • Schedule view refresh

  9. You can see analytical view definitions in the Recommended table by clicking Details. It has several tabs with information about the recommended view, the base storage structures, and the requesting query (either the visual or the filter):

    • Score Detail
      Aview Advantages
      These considerations determine the rating of the analytical view.
      File format conversion
      Analytical views are always in parquet format. If original data is not in parquet format, then converting to an analytical view applies parquet's default compression techniques.
      Join materialization
      Join computations are cached, eliminating the need to perform join computations after query routing.
      Data reduction
      Aggregation functions can result in data reduction. Analytical views cache the results after aggregation, eliminating to perform these computations after query routing.
      Incrementally refreshable
      If analytical views are partitioned, they can be incrementally refreshed. Only the partitions containing stale data will be refreshed, leaving the valid partitions untouched.
      Score breakdown
      The following elements indicate how the score is calculated.
      Max possible score
      This number is 5.
      Deductions / No Deductions
      This number is subtracted from the maximum score.
      Total Score
      This number represents the final rating after applying deductions.
      Recommendation Details: Score Detail
    • Base Table Info
      Table Name
      The name of the table appears in the form database_name.table_name
      File Format
      Format of the table, such as PARQUET
      Row Count
      Number of rows in the table
      Row Size
      Typical size of each row, in Bytes
      In Memory Size
      The size of the table when stored in memory, in Bytes
      On Disk Size
      The size of the table when stored on disk, in Bytes
      Partitions
      The names of the columns that partition the table; can be none or several
      Recommendation Details: Base Table Info
    • AView Info
      Aview Name
      The name of the analytical view appears in the form database_name.aview_name
      File Format
      Format of the analytical view, such as PARQUET
      Row Count
      Number of rows in the analytical view; on recommended analytical views, this number is an estimation
      Row Size
      Typical size of each row, in Bytes
      In Memory Size
      The size of the analytical view when stored in memory, in Bytes; on recommended analytical views, this number is an estimation
      On Disk Size
      The size of the analytical view when stored on disk, in Bytes; on recommended analytical views, this number is not available, so we report is as To Be Determined (TBD)
      Partitions
      The names of the columns that partition the analytical view; can be none or several
      Recommendation Details: Aview Info
    • AView SQL
      Aview Name
      The name of the analytical view appears in the form database_name.aview_name
      Measures
      Measures used in analytical view; these are typically aggregations, such as sum(1)
      Dimension
      Dimensions used in analytical view; dimensions that match a partitioning column have a green triangle flag
      SQL
      The SQL SELECT statement that defines the analytical view
      Recommendation Details: AView SQL
    • Visual SQL
      Name of the Visual and its ID
      The name of the visual that originates the query request, and its ID, in the form visual_name (ID: visual_id)
      SQL
      The SQL SELECT statement that defines the visual
      Recommendation Details: Visual SQL

Proceed with Create Analytical View.