In SAS^® Real-Time Decision Manager, a decision campaign might return an unexpected result in the following circumstances:

1. The decision campaign includes at least one Score node, and the underlying scoring model is based on a PMML file that has been imported into SAS^® Model Manager.
2. During execution of the campaign flow, one or more of the Score node's input variables contain missing values.
3. Although this campaign execution does return the correct result, if one or more of the Score node's input variables contain nonmissing values during a subsequent execution of the campaign flow, the Score node output variables might not return the expected value(s). Therefore, the Reply node's reply variables also might not return the expected value(s).

A similar problem might occur during the execution of a test case in SAS^® Customer Intelligence Studio. Consider this example. You have two test cases. In the first test case (Test Case A), the Score node's input variables contain missing values. But in the second test case (Test Case B), the Score node's input variables all contain nonmissing values. If you run the test cases independently, then both test cases return the expected results. If you run the test cases together and you run Test Case B first, followed by Test Case A, then both test cases return the expected results. However, if you run Test Case A first and then Test Case B, Test Case B returns unexpected results.

Workaround:

If you are importing a new model from a PMML file, perform the following steps after you import the model in SAS^® Model Manager:

1. On the Model Properties page, select Score Code to edit the score code.
2. Select the Edit check box.
3. Add the following code line just after the header:
   
   PSCR_WARN=0;
4. Save the new version of the score code.

Then, perform the following steps if you have already created a Model node definition in SAS^® Customer Intelligence Studio and the underlying scoring model is based on a PMML file that has been imported into SAS> Model Manager:

1. Edit the score code for the model(s), as described above.
2. Create a new Model node definition. (Do not overwrite an existing definition. The best practice is to add the version number in the definition name.)
3. Amend any decision campaigns that already use the Model node definition by changing Model Nodes to use the new definition.
4. Mark the campaign for deployment.
5. Deploy the campaign.
6. Activate the campaign.

Click the Hot Fix tab in this note to access the hot fix for this issue.

Operating System and Release Information

A fix for this issue for SAS Decision Services 6.4 is available at: