Question
Newfold
US
Last activity: 22 Feb 2024 5:08 EST
Random Control Group: Q103a-d. Its Propensity Reporting?
Q103a. Where and how can one see the Propensity scores in the Pega Ai Control Groups?
Since the Control group comes from a random uniform(1,0) distribution, I want to see and verify that the control groups have a volume around 2% and an average propensity score around 0.50 (a 50/50 coin toss).
Q103b. Does Pega have an out of the box report which shows the average propensity scores for its Test group vs. its Control group? If yes, what is that report called and where can I find its two respective average scores?
I also want access to (want to see) each records individual propensity scores and its Test vs. Control flag (its categorization or property).
Q103c. Where in the Pega CDH internal database can I get access to these individual records and their control group propensity scores? From the version 8.6 IH data model, Pega Customer Decision Hub Interaction History data model | Pega , it appears that appropriate dimension/class/property is the “modelcontrolgroup” property in the pr_data_ih_dim_customer table. Please confirm.
Q103a. Where and how can one see the Propensity scores in the Pega Ai Control Groups?
Since the Control group comes from a random uniform(1,0) distribution, I want to see and verify that the control groups have a volume around 2% and an average propensity score around 0.50 (a 50/50 coin toss).
Q103b. Does Pega have an out of the box report which shows the average propensity scores for its Test group vs. its Control group? If yes, what is that report called and where can I find its two respective average scores?
I also want access to (want to see) each records individual propensity scores and its Test vs. Control flag (its categorization or property).
Q103c. Where in the Pega CDH internal database can I get access to these individual records and their control group propensity scores? From the version 8.6 IH data model, Pega Customer Decision Hub Interaction History data model | Pega , it appears that appropriate dimension/class/property is the “modelcontrolgroup” property in the pr_data_ih_dim_customer table. Please confirm.
Q103d. If yes, does that modelcontrolgroup property need any other properties, fields, or filters for it to report out correctly? For example, does the modelcontrolgroup property need to be set or filtered by a date, a validation step, or an outcome type (such as “controlgroupvalidtystart”, “cgvalidityend”, “pyoutcome”, and/or “pyresponse”)???
Accepted Solution
Updated: 22 Feb 2024 5:08 EST
Pegasystems Inc.
NL
Hi Peter,
Please find the answers below,
Kind regards,
Ivar
Q103a. Where and how can one see the Propensity scores in the Pega Ai Control Groups?
Since the Control group comes from a random uniform(1,0) distribution, I want to see and verify that the control groups have a volume around 2% and an average propensity score around 0.50 (a 50/50 coin toss).
- The propensity scores are stored in the Interact History. The modelcontrolgroup='Control' indicates that the customer was in this control group. Please note that this will not be distributed uniformly as the prioritization will bias higher propensities
Q103b. Does Pega have an out of the box report which shows the average propensity scores for its Test group vs. its Control group? If yes, what is that report called and where can I find its two respective average scores?
I also want access to (want to see) each records individual propensity scores and its Test vs. Control flag (its categorization or property).
Hi Peter,
Please find the answers below,
Kind regards,
Ivar
Q103a. Where and how can one see the Propensity scores in the Pega Ai Control Groups?
Since the Control group comes from a random uniform(1,0) distribution, I want to see and verify that the control groups have a volume around 2% and an average propensity score around 0.50 (a 50/50 coin toss).
- The propensity scores are stored in the Interact History. The modelcontrolgroup='Control' indicates that the customer was in this control group. Please note that this will not be distributed uniformly as the prioritization will bias higher propensities
Q103b. Does Pega have an out of the box report which shows the average propensity scores for its Test group vs. its Control group? If yes, what is that report called and where can I find its two respective average scores?
I also want access to (want to see) each records individual propensity scores and its Test vs. Control flag (its categorization or property).
- In the monitoring of a Prediction the different success rates for these two groups are shown, and the difference expressed as Lift
- Additional detailed reporting can be done offline on the Interaction History data
Q103c. Where in the Pega CDH internal database can I get access to these individual records and their control group propensity scores? From the version 8.6 IH data model, Pega Customer Decision Hub Interaction History data model | Pega , it appears that appropriate dimension/class/property is the “modelcontrolgroup” property in the pr_data_ih_dim_customer table. Please confirm.
- Correct, modelcontrolgroup='Control' indicates the customer was in the control group where a random propensity is generated
Q103d. If yes, does that modelcontrolgroup property need any other properties, fields, or filters for it to report out correctly? For example, does the modelcontrolgroup property need to be set or filtered by a date, a validation step, or an outcome type (such as “controlgroupvalidtystart”, “cgvalidityend”, “pyoutcome”, and/or “pyresponse”)???
- This is not required