This article focuses on how to consolidate output results across model points using Product Output Groupings. Output Groupings are used for data consolidation and reporting purposes in projections.
Product Output Groupings allow you to create subsets of aggregation totals for product variable results within a given product, which are dynamically grouped by another variable value on that same product. For example, if you wanted to see reserve totals grouped by valuation basis or pricing metrics grouped by age, gender, and risk class within a product, then Product Output Groupings will allow you to create these subtotals during the projection runtime in an efficient manner. The results of these aggregations are also made available in the reporting module for you to create custom reports against.
Creating Output Groupings
Product Output Groupings are defined at the model level for all Products on the Product List screen. To add a new Product Output Grouping, click the "Add Grouping" button at the top of the Groupings section on the Products page.
A window will open allowing you to select up to 4 Product Variables to include in an Output Grouping. The Name field for the Output Grouping is optional. If it is omitted, the Product Output Grouping will be automatically named for you based on the names of the variables selected in the grouping.
The variables list in the dropdown menu will display a comprehensive list of all variables that are contained within any of the Products with the current model. Output Groupings can only be based upon product variables, not model point fields. When a model point field is desired to consolidate model point results, simply assign the model point field to a new product variable and include the new product variable within the Output Grouping definition.
Click the Save button when completed. This Output Grouping will now be available for use in Projections.
How They Work
Product Output Groupings create subsets of the total Product results grouped by the distinct values from the variables defined in the Output Grouping configuration. For example, looking at the screenshot above, a set of results will be created for each unique combination of Gender and Risk Class that exists in the output from model point calculations.
The groupings are created dynamically each time a projection is run, and are only created for model points that are actually processed in that run. So, if model points are excluded due to Model Point Filters being applied, or the run terminates early due to error limits being reached, the Product Output Groupings may not contain every possible value that could exist had the full set of model points been processed.
Product Output Groupings work in conjunction with the Aggregation Variable Property on product variables. Every product variable which has the Aggregation property set (i.e. it is not 'None') will be included in results for the Product Output Grouping and an aggregate result will be calculated within the subset grouping. The calculation performed will be the same as is done for the full product aggregation, but only including the model points that fall into the specific subset.
Because of this behavior, Product Output Groupings are best used with Group By variables that have a low cardinality in your data. Creating Product Output Groupings using variables with a high cardinality can result in negative performance impacts during projection execution.
Using Product Output Groupings on Projections
Once a Product Output Grouping has been created and configured, it will be available for use on any future projection for this model. The groupings can be selected when creating or editing a product on any Projection Template or Projection.
You may select 0 or more Output Groupings to process for each product configured. It is also possible to set different Output Groupings for different products within a single run if desired. The Output Groupings will be processed independently for each product.
Reporting with Product Output Groupings
Projection results for Product Output Groupings are saved in the Product Results (Grouped) table in the results database. Output Groupings are created for reporting purposes for each unique combination of values across all product variables that comprise a given Product Output Grouping. For example, assume a Gender variable exists with two possible values (Male or Female) and a RiskClass variable exists with three possible values (Preferred, Standard, Substandard). When processed, the Output Grouping defined as Gender-RiskClass will split all model points into six (= 2 genders X 3 risk classes) distinct groups.
If a third product variable is added to this named Age with attained ages ranging from 20 to 69, then the Output Grouping defined as Age Gender RiskClass will result in 300 (=2 x 3 x 50) groups.
Those variables will become Group By options in the reporting module for all aggregate variables after the projection is run. This allows users to examine cross-sections of model point data within a given product without having to aggregate results manually from reporting module output which will allow for more efficient loading of report results. Output Groupings allow for consolidation of model points across up to four (4) variables. Note that this maximum limit of four variables can easily represent hundreds of model point groupings.