Challenge #DIC014-ENE015: Vehicle Insurance – UServ Product Derby

Description

For the Challenge de Diciembre 2014 – Enero 2015 , Decision Management Community web proposes participate in the “UServ Product Derby”. Over 10 years ago BR vendors had a chance to demonstrate their product capabilities at Business Rules Forums, using the same use case known as “UServ Product Derby”. The majority of BR vendors openly competed during these major BR events and potential users had an opportunity to compare different solutions.

The business case deal with vehicle insurance problems including eligibility, pricing and cancellation policies for a hypothetical insurance company. In this link we have a document with a more detailed description of the business case with a model of business entities and the test scenarios which we have to solve.

To deal with this solution we will use the BRMS IBM Operational Decision Manager 8.7.0. This new version of the product, let us simulate the impact of changing decisions for the client portfolio.

Below we detail our solution to the business case.

Rule Project

Our solution receives a proposal for Policy trading. From these data will go running rules, calculating premiums for both cars as drivers containing the policy and in the final step estimate the annual policy premium.

The flow of execution of rules or workflow is as follows:

The first package of rules decides the level of risk depending on the characteristics of cars and drivers.In the package “Pricing”, from previously calculated risk level, assigns a score to each participants in the policy (cars and drivers) and depending on the total score determines whether the client is eligible to contract a policy.If the client is eligible, we calculate the value of the annual premium. If the client is not eligible, the workflow execution finishes.

Business Console

Business Console Decision Center is the platform used by business users and / or analysts to manage the rules repository.

As shown below rules that form the Service Decision are written in natural language, enabling users of the platform to keep these rules without depending on IT department. This provides greater flexibility in making decisions because we separate application logic from business logic.

The solution we have defined for the business case consists of 65 rules. Let’s see in detail some of these rules:

Eligible Driver senior driver and training certification

In this rule, if the driver is Senior and has training certificate have it marked as eligible for the policy.

In this example we see a rule for the package “Pricing”. Where we will increasing or decreasing the premium based on conditions. On the right side, we can see the data and some properties of the rule. User who edited the rule, the state of the rule and  date properties that we can use for the rule to be valid for a certain period of time.

Besides business rules have also decision tables. In this picture we can see a table we use for scoring. In the table we can define some preconditions that decide if the table is executed or not, then we need as many columns as columns conditions and action.

At the head of the column we define the condition or action and rows value.

From the scenarios presented to us in the business case, complete the information in excel. These excel with a simulation run them within the Enterprise Console and we graphically show the results.

Solution

The business case proposes a series of scenarios to validate. When you run these scenarios, we have to provide some results to ensure that the solution was correct..

Some rules use the current year, for the implementation of the scenarios we will use 2015.

At the end of the article we show the implementation that we have order to reach solutions shown below.

Preferred Client Business Rule V1

For the implementation of the scenarios the tool provides excel sheets to enter data and the results we expect. We fill the information in excel with data from this scenario.

On the first tab we define the policy with drivers and cars.

On the next tab we define the policy drivers.

And finally we define the policy drivers.

Result the execution of a simulation scenario. The client type is  “Preferred Client”.

Preferred Client Business Rule V2

We generate excel with new data from this scenario.

Result the execution of a simulation scenarios. Although the data are almost identical to the previous client, the client type is not “Preferred Client”, price increases..

In this scenario only change data of the car, and tell us that vehicles have one more year. To do this change the year of the vehicle.

Result the execution of a simulation scenarios. In this scenario re-rated the first customer and still “Preferred Client”, the price varies because the vehicle is older one year more and is not within the range “between 5 and 10 years”.

Eligibility Within and Outside an Elite Client Relationship

Result the execution of a simulation scenarios. Shane’s client Type is “Elite Client” because she is within father’s portfolio. Therefore his “Eligibility Score” is 0

Result the execution of a simulation scenarios. Shane is out of the portfolio of his father and therefore his client type not is “EliteClient”. With his data his “Eligibility Score” is 100 and therefore the eligibility result is:

“the client’s application/policy renewal must be reviewed by underwriting manager who will determine whether the client is eligible for auto insurance“

We calculate the value of his annua lpremium considering that after review it has been considered eligible.

Simulations

We take the Challenge to show simulations in the Enterprise Console, simulations are becoming available from version 8.7. With simulations, besides show the results we wish, from change of a rule we can  observe the impact it has in our client portofolio. We will show the workflow for creating a simulation and a brief description one of each elements needed for simulation.

Metrics The metrics define the elements used in the ICR. You must create metrics to make ICR

KPIs o ICR The ICR show the results of running a simulation.

Data Simulations using business information which is supplied as input. To run a simulation, you must use an Excel scenario file or a file from a data provider.

Report formats When a simulation runs, it shows its results into a report. You can define multiple formats so that a simulation generates various reports.

Simulations You can set up a simulation to collect ICR, input data, a server and a report format. When a configuration simulation runs, it uses rules associated service decisions and puts the results in a report.

In this example, we show a graph where the variation of the annual premium is shown in the first 3 scenarios previously executed.

Conclusion

In this new Challenge we have seen as a BRMS can help insurance companies, convinced that this technology should be part of its Strategic Plan Systems because:

• Improve your “Time To Market” accelerating decisions in production environments.
• Critical services allows the sector to be legible and traceable improving the attention of customer complaints.
• Reduce dependence on IT department thus reducing maintenance costs.
• Centralized in a unique system Decision Services allowing these are centralized and reusable for different applications and technologies (On-Line, Batch).
• Simulates the impact of changing decisions on the portfolio.

Implementation

We will detail the development, to show how we determine if the client is eligible for the policy and how we calculate the annual premium.

By default the algorithm we use is RetePlus, exit criteria none and order by priority. The default priority is 0. When we want rules are executed before others, we do it giving a higher priority.

Eligibility The first packet execute in the flow is “Eligibility”.

The first of the subpackage is “AutomobileEligibility” in wich we’ll determine the category of risks based on vehicle characteristics. The order would be : 1-PotentialTheftCategory: We calculate the vehicle category based on the risk of theft. Although several rules can be executed in the same package by the characteristics of the vehicle, it is always assigned the most restrictive. We achieve this by placing a higher priority to the least restrictive categories. In this rule for example is determined that the category is high if the vehicle is greater than 45000. All that the category is “high” we put a low priority (-100) for run the latest. In this other if the price of the car is between 20000-45000 and is not in the list of the high theft probability auto , the category is “low”. We will put the highest priority to run the first. 2-PotentialOccupantInjuryCategory: We calculate the vehicle category in terms of risk of injury to the occupants. In the same way you always assign the most restrictive and we got priority. In this rule if haven no airbag assign the category “extremely high”. The airbag attribute is a list of String. We assign lower priority for the last run. In this one if you have different types of airbag, the category is “low”. We will put the highest priority to run the first. 3-AutoEligibility: Define auto eligibility based on the category that we have assigned. If the category is “extremely high” assign “not eligible”. If the category is “high” assign “provisional”. 4-Default AutoEligibility: Last case you have not defined any auto eligibility above, mark it as eligible. We only have one rule, if it has not been assigned any eligibility, mark it as eligible.

The next of the subpackages is “DriverEligibility” drivers categorized according to their characteristics. These rules do not need priority because you can not execute multiple rules in a package, because are exclusionary. We divide it into 2 packages: 1-DriverAgeCategory: In this package we categorize the driver for his age and license. From this categorization evaluate whether it is “eligible”. ExperienceDriver:Depending on the age and sex, we assign a driver experience. TrainingCertification: We mark a true has training certificate. Eligible: Depending on the experience and if you have certificate assigned if eligible, assign eligible. 2-DrivingRecordCategory: Depending on characteristics of the driver will mark it as “high risk driver”

The next subpackage is “ScoringEligibility“. No need priority, are exclusionary rules: 1-Scoring: assign a score based on the categorizations we have given to cars and drivers. 2-Scoring result: Depending on the score of the previous package, we decide whether the Client is Eligible.

Pricing In this package we calculate premiums for cars and drivers. No need priority.

1- AutoPremiums In this package we will calculate the premium for each cars involved in the policy. In the premium car we have a method “increase the premium” which increase to the premium amount. BasePremiums Depending on the model of car we assign a premium base. This premium base will be the initial value of the premium car. NewCar One rule that evaluates if a car is new and based on that increases the value of the premium. In the BOM have a “is new” method that evaluates whether the year of the car is the same as the current. The rule have a Boolean condition and the b2x where contain all operations IncreasePremiums: Depending on vehicle characteristics increase the value of the premium. 2-AutoDiscounts: Apply discounts premium car. The percentages accumulate in the variable “discount” of each car. 3-TypicalDriver: Assign experience “Typical Driver”. 4-DriverPremiums: Calculate premium each of the drivers involved in the policy. Also in the premium driver have a method “increase the premium” which adds to the premium amount. 5-MarketSegment: Depending client type will discount amount to the annual premium.

AnnualPremium In this package is where we calculate the annual premium. Here need priority because we need to execute rules before others.

In this order will go executing 4 last packet rules: First we calculate the premium car using the discount. In this rule we add the premiums of all cars, we will save the value in a variable “Sum the premiums of the cars”. Later we use In this rule we add the premiums of all cars, we will save the value in a variable “Sum the premium of the drivers”. Later we use. Finally, we calculate the Annual Premium