Creating a Predict Inclusion Report

Create a Predict Inclusion Report to calculate the likelihood of a person being included in a particular group based on what is known about them.

Procedure

1. Drag the tool to the workspace or select it from Create > Engineering > Predict Inclusion.
2. Specify a Target Group: This is the segment of records that you wish to base your model on. It is the people that already exist in a particular group that you wish to use to predict what other people are likely to join that group in the future. In our example, it is the people that have already bought a particular product. Drag in a segment of records or use the Create new drop-down to engineer a new segment.

3. Specify a Universe: You can add an optional universe filter to limit the scope of your calculation. In our example we could potentially limit our model to a limited time period, or a particular geographic region. Use standard drag and drop functionality to add a filter.
4. Optional settings: if you want to make the column permanent so that it is saved in the database, and can be viewed and selected in the Data Explorer tab, click the Optional expand icon to display the following optional settings:
   • In the Table Column Name field, enter the name you want to use to display the column in the Data Explorer tab.
   • If you want the new column to be indexed automatically, select the Index Column check box.
5. Columns: these are the attributes used to profile the records in a segment to then calculate the prediction model. It is the information you know about the people in your database and could include their Age, Income, Occupation, MOSAIC Group or any previous transaction history you have. Drag in columns from the Data Explorer that you wish to include in the model. You can use columns from the same table as the target or from a superior table. For example, if the Target is at Customer level, columns from Customer and Household will be valid, but Order columns will be ignored. The Tool can process three types of column:
   • Categorical: these are converted into flag columns (1,0); only columns with a maximum of 500 discrete values can be used.
   • Numeric: for example, Age or Income.
   • Date: are converted into expression columns, for example, Year and Month.

   Invisible Columns

   During the model building process, a number of invisible columns will be automatically created as the model processes the selected columns to establish statistical trends. For example, if you drag in the column 'Age', a number of banded columns may be created to analyze if particular age ranges are statistically important. Invisible columns contribute towards the specified 'Max Model Column' and will remain in the database until the Engine Cleanup tool is used to remove them.

   Max Model Columns

   If you have selected a categorical variable like 'Occupation' as a column, this will be turned into a series of flag columns. You can add a number to this field and the tool will use the most relevant columns up to the specified number. So if you add a number of 50, then the 51st most relevant column will be excluded from the calculation. Note that this figure includes any invisible columns that the model creates automatically.

   Ideal Sample Size

   Ideal Sample Size limits the sample size that is used by the model and applies to both the target group and the prospect pool. If a sample size is 5,000, the model takes that number of people from our target group, and the same number of people from our prospect pool.

6. Optional Advanced Settings:
   • Minimum Significance for Model Inclusion: This is the minimum entropy score that is required for a column to be included in the model. Lowering this value may allow a model to be created but is likely to increase execution time and may produce less robust results.

   • Model Diagnostics Enabled: select this check box if you want to generate a diagnostics file on the server that can be used to investigate reported issues.
7. Click the Process buttonto run the model.

• Predict Inclusion Tool
  • Creating a Predict Inclusion Report
  • Analyzing the Results
  • Further Results Analysis