Insurance Score for Personal Lines
Objective
To educate readers on how Zendrive’s Personal Lines (PL) Insurance Score can be used to measure risk for better risk segmentation and pricing. This document also conveys the minimum data specifications necessary to achieve a reliable and stable score, and the API to retrieve the data.
What is the PL Insurance Model?
The PL Insurance Model is Zendrive’s scoring model for Personal Lines that predicts collision frequency risk based on driving behavior events and factors such as nighttime driving.
This model uses verified detected collisions as an objective measure of risk, which makes it suitable for PL insurance underwriting, rating, and discounting. Zendrive PL Score was developed in accordance with Actuarial Standards of Practice (ASOP) 56 and endorsed by Milliman, a major insurance provider.
What Variables Go Into the Model?
The following table provides a list of PL Insurance Scores, with Hard Braking and Distracted Driving being the strongest indicators of risk:
Hard braking
Count of hard braking events.
Hard acceleration
Count of hard acceleration events.
Distracted driving
The total amount of time for which the driver's phone was in use during the trip. This value includes physical handling of the phone, handheld calls and texting.
Nighttime driving
The total amount of time spent driving in the night, between midnight and 4 A.M. local time.
High speed
The total amount of time during which the vehicle's speed exceeded 75 mph or kmph.
The PL Insurance Score model generates an Insurance Score for each driver, which assigns a risk value based on the predicted number of collisions per 1 million miles.
How Well Does the Insurance Score Model Perform?
The following lift chart illustrates the model’s predictive power, and its ability to segment between the best and worst risks. To create this chart, Zendrive’s data was sorted from the lowest to highest predicted frequency, then grouped into deciles of equal exposure, as follows:
Decile 1: The group of drivers with the lowest predicted frequency (safest drivers).
Decile 10: Behind the group with the highest predicted frequency (riskiest drivers).
This chart shows the relative frequency of collisions per 1 million miles for each decile of the Zendrive testing dataset, based on equal driver-months:
What Is Insurance Score?
Insurance Score is the risk value generated for a driver, based on Zendrive’s predictive model. This score predicts the expected number of collisions per 1 million miles of driving. This score can be used to determine a driver’s insurance premium, discounts, and/or underwriting qualification as determined by the carrier using the collision relativity of each score group. The score itself does not automatically dictate what the premium or discount of a particular driver should be.
How Is The Insurance Score Generated?
A driver’s risk score is calculated based on all the qualifying trips taken within a selected time frame as requested through the userStats API. This API currently allows the carrier to select a window of 7, 14, 21, or 30 days, based on what makes the most sense for their program.
In addition to the insurance score, the API contains the following metadata of the driver’s performance during the requested time period:
Number of miles/kilometers captured
Number of trips taken
Event scores
Total driving hours
Nighttime driving hours, and more.
How Can The Insurance Score Be Used For Risk Segmentation and Discounting?
The PL Insurance Model measures the relative risk of the driver population used in the model. These relative risks can be applied to a driver’s rating, to better match the price to the risk.
However, the exact values and application of discounts or pricing factors to be used as part of UBI \ BBI offerings will be up to the carrier and their actuarial/pricing teams, as this will vary greatly across customers based on their business operations, goals of their telematics programs, amount of discounting they’re able to take.
1
0.06
0.305
2
0.10
0.41
3
0.13
0.541
4
0.16
0.652
5
0.20
0.794
6
0.24
1.047
7
0.29
1.18
8
0.37
1.457
9
0.48
2.032
10
0.71
3.043
What Is The Minimum Driving Data Required?
The Insurance Score model does not require any minimum data threshold to generate an insurance score via the driverInsuranceScore API , as long as at least 1 driving trip was taken within the selected insurance score window. However, there are minimum data parameters that must be met to achieve a fully stable score.
What Is The Minimum Driving Data Required For Optimal Score Stability?
The PL Insurance Score requires the following minimum thresholds for optimal stability and predictability:
400 miles or kilometers of driving
20 trips
These thresholds help ensure the score truly represents the driver’s general driving risk and isn’t skewed by a few long trips. If a score is generated on less than 400 miles of driving, the generated score will show greater variability.
The following table helps illustrate the decrease in accuracy of discount calculation with lower mileage. It is the carrier’s responsibility to build business rules to accommodate their use case. For example, the carrier could reduce price sensitivity based on driver's driving behavior at lower mileage levels.
Miles
0-50
51-100
101-200
201-300
301-399
400+
Discount Factor
0
0.125
0.25
0.5
0.75
1
Refer to the Zendrive insuranceScore API documentation to understand how the insurance score is retrieved for users of the Zendrive App.
The insuranceUserStats API Endpoint v2 remains available and supported for existing users. For more details, refer to Zendrive insuranceScore API V2.
Was this helpful?