Renewal Optimization in SAS Dynamic Actuarial Modeling
Introduction
The SAS Dynamic Actuarial Modeling Premier solution offers a comprehensive renewal optimization workflow specifically designed to streamline and enhance the process of optimizing price increases for premium renewals. This workflow is tailored to support actuarial teams in analyzing, modeling, and executing pricing strategies that balance profitability with customer retention. By leveraging advanced data analytics capabilities, the workflow provides actionable insights to determine optimal price adjustments during the renewal process, ensuring alignment with both organizational goals and market dynamics.
This article focuses on setting up the optimization model and it assumes that you have a basic level familiarity with terminologies involved in optimization, insurance pricing, SAS Model Studio and SAS Dynamic Actuarial Modeling solution.
|
You can obtain more information on SAS Model Studio and SAS Dynamic Actuarial Modeling using the following links.
Machine Learning Using SAS Viya
SAS Dynamic Actuarial Modeling Site Page
SAS Dynamic Actuarial Modeling Support
|
The Workflow
SAS Dynamic Actuarial Modeling Premier offers a Renewal Optimization Workflow (RENEWAL_OPTIMIZATION_WORKFLOW_<version number>) designed specifically for optimizing price increases during premium renewals.
Following represents the steps involved in the workflow.
Illustration of steps in the Renewal Optimization Workflow.
Select any image to see a larger version.
Mobile users: To view the images, select the "Full" version at the bottom of the page.
Below is a concise overview of the main steps and key objectives involved in the workflow.
| Task | Objectives | Supporting Tool |
| Import Data |
This task loads data into the CAS server, including a Customer Table and a Rate Increase Table:
|
SAS Data Explorer |
| Prepare Data | This task enables you to validate the input data and map or append any required columns needed for optimization. | SAS Scripts |
| Analyze Data | This task enables you to analyze information assets, such as data tables, utilized throughout the cycle. | SAS Information Catalog |
| Explore Solution | This task enables you to explore your optimization results. | SAS Visual Analytics |
| Review Results | This task enables a senior analyst or actuary to review results | SAS Visual Analytics |
| Validates Results | This task enables you to filter optimization results based on a retention level or the primary constraint defining the scenario. | SAS Scripts |
Open Optimization Model Task
The Open Optimization Model task is executed using SAS Model Studio, a powerful platform for advanced analytics and modeling. This task provides the tools and framework necessary to construct a robust optimization model. By leveraging SAS Model Studio's capabilities, users are enabled to define objectives, incorporate constraints, and integrate various data inputs to develop a model that optimally aligns with the organization's goals. This step is essential in ensuring that the optimization process is tailored to address specific business needs, such as balancing profitability and retention.
It is important to note that for an insurance company, increasing premiums too much may result in customers leaving, and too low premiums might result in a loss in net customer value or overall profits. The plan is to use analytics to optimize the premium hike. Optimization helps to deliver a value of premium increase that has low or negligible impact on customer turnover rate.
In the solution, there are four prerequisites for carrying out renewal optimization model related activities. These are as follows:
- First, it is necessary to have a churn or renewal model that calculates the probability of churn for a specific customer profile and price.
- Second, it is necessary to train the model using data from the same rise or decrease range for the price. For instance, if a model is based on data with increases up to 10%, the result for a 12% increase might not reflect actual customer behavior. It might be safer to stay in the known rise or decrease range of data in certain circumstances, or a pilot is required to retrain the model.
- Third, it is necessary to optimize a curved function with a mathematically possible optimal solution. You should use the function that is most appropriate for your organization. There is no optimum for a linear function. In general, the function is the profit margin function multiplied by the renewal probability. The profit margin function is an estimate of expenses and claims for the next year per customer profile.
- Fourth, it is necessary that the constraints should avoid overpricing the vulnerable customers and should provide enough increase to the high-cost claimants. You should create constraints that protect vulnerable customers. For example, limit the percentage increase for economically weak customers with specific features. This enables you to avoid overpricing them.
Pipeline Nodes
The task involves configuring the data node by assigning roles, such as the target variable, and setting up the optimization node by defining the objective function, calculated measures, and constraints using the node's property panel.
Illustration of the nodes in optimization pipeline.
The Data node prepares the input data set for analysis by subsequent nodes. In the Data tab, you must review and, if necessary, modify each variable's role, level, and transformation values. For instance, in the sample data, the variable NEW_PREMIUM_AMT is automatically assigned the role of the target variable, but you should verify and adjust roles for your data needs.
Illustration of Data tab for setting up variable roles.
The Pricing Optimization node attempts to find the most advantageous pricing strategy for insurance renewals or other similarly defined optimization problems. The node can be used to optimize multiple metrics. Some examples of metrics that you might want to optimize are the number of customers that are retained, the total amount of retained premium and the measure of profitability for an entire portfolio. The pricing optimization node enables you to set up objectives and constraints for your optimization model.
An Example of Objective Function and Constraints
Here’s an example of how to configure an objective function in the solution. The objective function is represented by the measure Total_NetCustomerValue, which needs to be maximized. To set this up, enter the objective function and associated variables using the Objective Functions Editor, found in the objective function options in the property panel for Pricing Optimization node.
Illustration of objective function and calculated measures.
In this example, an insurance company aims to maximize the profit from its existing portfolio of policies. The company defines the new premium (new_premium_amt) as the previous premium plus an increase. The rate increase table contains the base variable for this increase. The New_Profit is calculated as the new premium minus costs like claims, expenses, and commissions.
To define the objective function, a Net_Customer_Value formula is used, which multiplies the new profit by the probability of renewal (representing expected profit). This process is repeated for all customers, with two key measures: the Net_Customer_Value for each individual customer and the Total_NetCustomerValue, which is the sum of individual values across the portfolio. Note that you must run Data node before proceeding for the configuration of objective function and constraints. You must use the property panel for the Pricing Optimization node to configure the settings. You must provide the appropriate rate increase table in the property panel. The following figure highlights the important parts of the property panel for the node.
Illustration of the property panel for pricing optimization node.
You must click Open Objective Function to open a new window that enables you to define calculated measures and the objective function. Following is a sample of values you enter.
Next, click the Add Calculated Measure (+) icon. For the Variable Name, enter new_premium. Leave the Description blank, as it is optional. In the Code field, input the formula: &previous_premium_amt.*(1 + &increase_rt.).
Illustration of Calculated Measures and Objective Function window with sample values.
You can add constraints as well. As an example, following two constraints to avoid or limit high increase for low premium paying customers e:
- For customers with premium amount less than 1000, the increase rate should not be more than 4 percent. This constraint protects low premium customers from a high increase rate.
- For customer with premium amount greater than 3000, the increase rate should be up to 6 percent. This constraint ensures that higher premium customers have a relatively higher increase rate.
To add a constraint, click Open Constraints in the property panel, which opens the Add Constraint window. Provide a Constraint Name, an optional Description, and specify the Customer Filter (for example, PREVIOUS_PREMIUM_AMT < 1000) and the Rate Limit (for example, Rate Increase < 0.04). Use the dropdown menus to select the appropriate variables for the filters. Once completed, press OK to confirm and Save to store each entered constraint.
Illustration of windows used for adding constraints.
Results
The execution of Pricing Optimization node produces a variety of results including tables and charts. An Efficient Frontier Chart visually represents the trade-offs between retention probability (x-axis) and Total_NetCustomer_Value, the objective function value (y-axis).
Illustration of an Efficient Frontier chart.
The chart illustrates that as retention probability increases, the objective function value decreases. This inverse relationship occurs because higher retention probabilities result from lower rate increases, which reduce premium collection and subsequently lower the objective function value. The chart provides decision-makers with a clear set of options to evaluate and select optimal retention rates while considering their impact on the total net customer value.
Note that the optimization workflow using SAS Model Studio along with SAS Visual Analytics provides a variety of tasks and options to review and validate the results by exploring optimization result tables.
Illustration of a tabular output for optimization solution.
Conclusion
The article introduces the renewal optimization workflow and provides examples of the setting up objectives and constraints using SAS Model Studio in SAS Dynamic Actuarial Modelling for your optimization needs. This article shall motivate actuaries and modelers to try out and test several configuration options based on your requirements. Such steps might be beneficial in the development and implementation of renewal optimization models at your site. For more details refer to the documentation of the solution available at Documentation for SAS Dynamic Actuarial Modeling
Find more articles from SAS Global Enablement and Learning here.
