By Eleanor Brodie, Data Science Manager, Insurance, LexisNexis Risk Solutions and Carla McDonald, Senior Vertical Market Manager, Insurance, LexisNexis Risk Solutions
With retrospective value analysis (retros) for LexisNexis® Vehicle Build, and the ADAS risk classification now coming to completion with major insurers in the UK and in the EU, we wanted to give an update on the product and the people involved in the process.
When we conduct a retro, we play back actual data from a client’s book of insurance business (a data append) from the recent past and we’re able to join that with the Vehicle Build score classification to show real-world benefits for an insurer. We’ll come to some examples of the numbers and the benefits in terms of the underwriting uplift a little later.
But a few observations first about how we got to the current point: why is it that ADAS vehicle data holds such promise for the future of motor underwriting? What are the ADAS combinations and clusters that are really predictive of insurance loss? What type of path is it creating for motor insurance rating in the future?
The LexisNexis ADAS classification system standardises vehicle safety features for motor insurance providers. It is no industry secret that there are multiple sets of proprietary ADAS features and their diversity is further complicated by automotive manufacturer naming conventions.
The logical sequencing and classification of hundreds of these variations into a common taxonomy enables insurers to more easily ascertain how these features affect a vehicle’s risk profile. From these many variations, the classification derives approximately 60 unique features, all of which are delivered through LexisNexis® Vehicle Build.
Why is ADAS important?
At LexisNexis Risk Solutions we have normalized a huge amount of data that’s been sourced from car manufacturers so that we can actually return this to the insurance provider as actionable insights in form of attributes. This identifies whether the ADAS features were fitted as standard, or if they are optional extras, because LexisNexis® Vehicle Build is a VIN level solution. But we don’t just identify if the feature is present or not, we are also able to tell the insurance provider, how does the feature behave? Is it there to warn the driver? Is it there to assist the driver? Or, as is increasingly the case, when a feature is there to actually take over operational control of the vehicle (such as automatic emergency braking).
With LexisNexis® Vehicle Build we are incorporating a lot of the infrastructure of the future that’s needed to connect insurance to the growing roster of ADAS features and vehicle autonomy. We wanted to be able to tell the insurance provider, so, how does a specific feature perform, what’s its purpose? Is that feature there to avoid a collision? Is it there to reduce the severity of a collision that should happen? Or in some circumstances, is it there to maintain safe driving conditions, on the road, or even post-collision?
For example, child-proof door locks and automated locking is going a step further with safe exit assist and its several variations: central locking may be released in the event of a collision so passengers can exit the vehicle. Door opening can be controlled depending on traffic approaching from the rear. Post-collision ADAS can open the bonnet hood so emergency services can access the engine bay in the case of a fire.
Within LexisNexis Vehicle Build we’ve also created core feature groups. Classification of each core feature group were created from research on how ADAS moved the needle for reduction of claims, plus an understanding of the market.
These core feature groups and the way they interact are what we’ve found will deliver a reduction in claims frequency. There are 12 in total, and they can fitted at the front of a vehicle, for example, such as Forward Collision Mitigation, including several ADAS groups. The graphic below represents the cluster of features that combine together to make up one feature group, Forward Collision Mitigation, mapped to show how the elements interact.
ADAS features can be fitted at the side of the vehicle, such as blind spot warning. Or they could be at the rear of the vehicle, such as Rear Collision Warning. At a very top level, these core feature groups essentially place a higher level of protection around the vehicle.
Forward Collision Mitigation: ADAS feature group example
The insights shown below highlight the impact and claim relativity when a vehicle is equipped with at least one core ADAS feature. 37% of the VINs that we studied were equipped with at least one core feature, and for these vehicles we found a 27% reduction in bodily injury claim frequency relativity, and a 19% reduction of property damage frequency relativity. The really interesting thing that we found, that while there was a decrease in claim frequency, the impact on claim severity was negligible.
Examples: ADAS feature combinations and impact on claims frequency
- Overall industry claim relativities when one or more core ADAS features is present: bodily injury -27%claim frequency, property damage -19% claim frequency.
We have also carried out a multivariate analysis to see the impact of combinations of ADAS vehicle data on claim frequency. We used a decision tree machine learning algorithm, and this considers interactions between variables.
We found that when vehicles are equipped with Forward Collision Mitigation, Adaptive Cruise Control, Blind Spot Mitigation and Lane Departure Mitigation we see a reduction of 25% in Property Damage Claim frequency relativity. When vehicles are equipped with Blind Spot Mitigation, Lane Departure Mitigation and Adaptive Cruise Control we see a 13% reduction in Property Damage Claim Frequency relativity.
However, some of these features are optional extras, so just having model type level information isn’t sufficient. It’s crucial that we have VIN level information so that the insurance provider can know exactly what’s on the vehicle. Therefore, the Vehicle Build solution is at VIN level.
From our US studies, we can see that these features work in reducing claim frequency. We’ve also started testing in Europe and so far, we have tested nearly three million VINs with a 97% match rate. We’ve found the UK vehicles are equipped with more ADAS features than their Italian and Spanish counterparts. The figure for ADAS features on UK vehicles has been surprising to some. We’ve discovered that there are in fact as many as eight features on average per UK vehicle, and furthermore at least 64% of vehicles in the UK are equipped with at least one core feature grouping.
ADAS rating indicator
So, we’ve looked at the core features, their prevalence and impact in claim frequency and severity. Next, we consider the rating indicator.
What we’ve done is to really focus in on the ADAS features and ADAS feature groups, but we have also looked at the vehicle as a whole, to allow the insurance provider to really be able to compare vehicles with one another. To do this we’ve created what we call a rating indicator. This is a value between zero and five that allows the insurance provider to look at all the features that are fitted to a vehicle and how they perform in relation to reducing claims frequency.
The rating indicator was developed using insights from our LexisNexis Risk Solutions study in the US of 11 million vehicles with make year from 2014 to 2020. We looked at the bodily injury and property damage claim type. These claim types are where the vehicle causes an accident and is at fault and someone is injured or there’s damage to a vehicle. The rating indicator is a value between one to five that relates to each VIN. This value can be used as a base in combination with other ADAS features to determine the level of claim frequency per VIN.
The higher the number, the better the performance in reducing claim frequency. For example, a value of one shows that the vehicle does have some level of protection. This isn’t necessarily at the rear; it could be at the side or the front. A value of two or three shows that the level of protection around the vehicle is increasing and a value of five indicates all of the vehicle has protection. This additional intelligence delivers to the insurance provider a great starting point, as we found with clients that have tested the product, when this is used in conjunction with other ADAS feature data, it allows for greater risk segmentation.
Risk insights from US, UK and EU vehicles
So, in summary, what we found from our analysis in the US was that the core features have the biggest impact on claim frequency with negligible impact in severity. In the UK, 57% of vehicles have an ADAS rating indicator of one or higher. As technology develops, we would expect this percentage to increase. The crucial thing though, when we analysed a sample of European policies, was a 14% reduction in pure premium loss ratio when the vehicle had an ADAS rating indication of one or more. So, our US analysis shows a reduction in claim frequency, and our European analysis shows that our ADAS features show additional segmentation beyond insurers’ current predictions.
Ultimately LexisNexis® Vehicle Build is seeking to deliver insight into the insurance market, on these safety features that are fitted to any individual vehicle that insurers are seeking to place on cover today. Taking into account the current traditional demographic score model that’s in place today, we’re able to say with certainty that Vehicle Build contributes to delivering a more accurate insurance quote, based on the specific claim relativities for any individual vehicle (bodily injury, property damage and collision costs) that we’ve identified in the market today.
It is an ongoing exploration with data that is continuing to deliver better and better underwriting results as we continue to grow our data sample. It is already starting to really help insurers with their pricing and underwriting decisions.
In terms of the ongoing rise of vehicle technology and assisted driving systems, LexisNexis® Vehicle Build is the start of that journey and it plays into all the live, dynamic data possibilities that are gradually coming into view with the truly connected car.
So, when we talk about Vehicle Build, this is informing our clients about data in terms of a vehicle as it has rolled off the assembly line. These are the safety features that are fitted to the vehicle. Over time the modern car with embedded connectivity will enable us to show more about how these features are being used and triggered in real time. We are already seeing the start of this evolution with LexisNexis® connected car solutions and the Vehicle History and Vehicle Mileage products.
At LexisNexis Risk Solutions we believe the way we normalize this vehicle data and prepare it for insurance consumption is a key differentiator for us in the market. We don’t just tell the market what features are on a vehicle. We go beyond this and define the features and feature clusters in terms of their predictive value of a collision.
Since vehicle technology classification is a new rating factor, another important part of our work in this area is coaching, educating, testing in partnership with clients, and evolving this innovation together.
The data shows us that each area of the ADAS classification can impact on different driver profiles and different parts of any insurance book of business. Some of the testing, and building the necessary data infrastructure takes time, as this is a new type of risk attribute. Opportunities are different for each insurance provider, and we work with each one to optimise the solution. One challenge in the UK for example was to build a bridge for the VRN (vehicle registration number) to VIN conversion and then to be able to deliver that at point-of-quote.
We have built flexibility into the LexisNexis® Vehicle Build solution. For example, at the most granular level you have ‘individual ADAS features’ and they can be rolled up into ‘ADAS feature groups’ and even ‘core ADAS feature groups’ at a less granular level if required. Should insurers not want such a deep level of granularity and prefer a more ‘off the shelf’ solution, they can incorporate the ADAS rating indicator into their existing pricing models. This should aid all types of insurance providers on any challenges with ingesting the data.
Follow the link to the LexisNexis Risk Solutions website to find out more about how we support insurance providers.