This implementation is DEPRECATED, please refer now to Negotiation Guidance Accelerator.
Running a Segmentation-Optimization model is not a one-time thing. The prices keep evolving over time and it is necessary to periodically refresh the model with latest information. There are three levels of model refresh that need to be done with different frequencies:
Minor Refresh
A small refresh that only includes re-running the Scoring and Pricing steps (possibly only the Pricing step) using the latest set of transactions in the Datamart. This can be done regularly, even on weekly basis.
Workflow
-
Run the DataPrep step (if necessary).
-
If the DataPrep step has run in the past, the transaction data is stored in the model and this step needs to run again to load the latest data.
-
-
Run the Scoring step (optional).
-
The Scoring step can be run in a usual way.
-
-
Run the Pricing step.
-
When re-running the Pricing step, it is recommended to set a unique Run ID (available in the logic since 03/2020). This will ensure that a new set of Policy Records is created and the original Policy Records will remain intact. In case of any issue with calculation, it is possible to run the Pricing step again with the same Run ID.
-
-
Put the new Policy Records to use.
-
To use the newly created Policy Records, the Pricing Run ID model parameter needs to be set to the same Run ID that was used in Pricing step calculation.
-
Standard Refresh
A full model update that includes re-running Segmentation, Scoring and Pricing steps using the latest set of transactions in the Datamart. A Standard Refresh is typically done to reflect new data that significantly changes product or customer attributes. In some cases, Standard Refresh can significantly affect the price recommendations. Therefore, to maintain some level of consistency, it is not recommended to run a Standard Refresh too often. A reasonable frequency is 2-4 times per year or after significant data changes.
Workflow
A recommended workflow for a Standard Refresh is using a new version of the Optimization model.
-
Create new model.
-
Copy the model and change its name (for example, add a postfix containing the current date).
-
-
Set model parameters.
-
Model parameters are not copied over to the new model so it needs to be done manually.
-
Tip: The table with parameters on the Parameters tab only shows up after running a calculation step. There are two options to change the parameters before running any of the steps:
-
Start any calculation and cancel it immediately which will cause the table to show up.
-
Change the parameter values directly in the Company Parameter table Model Params <ModelName>.
-
-
-
Run the model.
-
Run all the steps of the model that are needed (the required steps are Segmentation, Scoring and Pricing) and validate the results.
-
-
Ensure the new model is used in Quoting or any other logics.
-
The model name is usually set in a library code or in a special Company Parameter table.
-
Major Refresh
A Major Refresh is the same as Standard Refresh, but it also includes:
-
Deployment of the latest Optimization logic
-
Comprehensive validation of the model calculations and results
-
Diagnostics
-
Clean-up of Datamarts, Rollups, Charts, Dashboards, Optimization Models, Optimization Logics, Company Parameter tables...
-
Improvement suggestions and best-practices recommendations
-
Update of the filter that excludes older transactions
A recommended frequency of a Major Refresh is once per year.