New Mobility

Automobile data – fuel for the insurance company? 

The transformation of the car into a networked data centre with more and more technology is creating alternative possibilities for risk evaluation in motor insurance. However, the use of the data is limited, warns ERGO Board Member Karsten Crede: “Current telematics solutions are reaching their limits, as insurers do not have direct access to the vehicle data.” A promising approach for the optimisation of existing risk models therefore lies in the combination of driving behaviour data and usage data of the assistance systems.

For hardly any other industry and its business model data is as elementary as for the insurance industry. We have learned to adequately assess and insure almost all conceivable risks on the basis of data. With regard to motor insurance, a stable but static model has become established, too.

Actuaries determine the expected loss requirement and thus also the premiums on the basis of variables such as vehicle type, number of claims, place of residence, age and many other factors. Unfortunately, the underlying process landscape is very complex, and in an effort to achieve competitive differentiation and underwriting excellence, more and more parameters have been introduced, which have led to a lot of of questions and a high administrative burden for the customer. Even digital tools such as comparison platforms are based on complex analogue processes and only compare prices based on largely identical performance of the products.

 

However, with the transformation of the car into a networked data centre with more and more technology, the starting position is changing. Powerful assistance systems and sensors, modern software and immense computing power on board the vehicles produce vast amounts of data – in the terrabyte range per day. Ten years ago, data was already being collected on GPS position, the frequency of electronic seat belt tightening or the types of roads travelled – in the meantime, the options are many times higher.

For insurers, the question now is which risk-related derivations the data make possible and whether it can be used to fundamentally digitalise the business model of car insurance. First, let's take a look at what is technically possible.


Surroundings data recorded by sensors and cameras enable the reconstruction of accident sequences, the clarification of guilt or also the identification of fraud. Collisions are recorded and automatic emergency calls are sent. It is possible to fully trace which assistance systems were active at which time and – with regard to autonomous driving functions – who is responsible in the event of an accident, the driver or the vehicle.

Vehicle data can also be used to provide the user with further insurance offers and services that are relevant to him or her. A simple example is travel insurance that is offered on the basis of GPS data or any destinations in the navigation system. However, intelligent warranty products based on real condition data of the individual components are also conceivable. The list of digital and data-based application possibilities is getting longer and longer.

Better risk models for motor insurance

The decisive question, however, is whether it is possible to develop more accurate and better risk models for motor insurance by using vehicle data, in contrast to the current approach. Today, almost every insurer already has motor insurance offers in its portfolio that take behaviour-based driving data into account – so-called telematics tariffs. Put simply, the existing risk model is supplemented by a “driving style factor”. This factor is determined by acceleration values, cornering speeds or the number of abrupt braking manoeuvres.

In my view, however, the current approaches have two decisive weaknesses: First, there are no real consequences, i.e. higher premiums, for risky driving behaviour. The “experienced” discounts are “financed” exclusively by a reduced need for claims, which is supposed to occur automatically in the users of these offers through a more careful driving style. Without this premise, telematics products would abruptly lose acceptance. Particularly in an age that is “full on flat rate”, for example when it comes to streaming films and music or even smartphone use, offers with sudden additional costs would not be compatible with key customer requirements.

On the other hand, there is no direct access to the vehicle data. The data required for the telematics tariff must be generated via external devices such as dongels or smartphones. This limits the quality and completeness of the models. HUK-Coburg board member Jörg Rheinländer recently illustrated this problem with a very good example: a certain braking manoeuvre could be evaluated much better if data from the rain sensor were also available and one knew whether the road was wet or dry.

Combining driving behaviour data and data of assistance systems

Furthermore, the question arises as to whether data on pure driving behaviour, i.e. braking, acceleration, speed, among other things, actually improve the quality of the risk models sufficiently. My thesis is that the highest informative value lies in the combination of driving behaviour data with usage data of the assistance systems. It would therefore be important to know how often an emergency brake assistant or a front collision warning system intervened, to what extent the parking pilot was used or how often lane departure and lane change assistants had to intervene.

A first, albeit still expandable, model is Tesla's Safety Score. Tesla provides the mathematical calculation of this score value on its homepage. A total of eight factors are included in this value. Seven of these factors are behaviour-based, i.e. they relate to driving behaviour in a narrower and broader sense. One factor relates to an assistance system – specifically, the number of front collision warnings.

A closer look reveals that even if all seven “behavioural factors” represent the worst value and only the “assistance system factor” is optimal, the final score is still 94.6 (100 is the maximum value). One possible interpretation would be that even a single assistance system provides more information on the need for damage than seven behaviour-based factors. A remote diagnosis as to whether the explanatory power of the behaviour-based factors is really so low or whether the score value is more of a marketing instrument than an actuarial tool is not to be made here.

A comprehensive approach in which several driver assistance systems are included in the risk assessment is currently being developed by ERGO's Mobility Technology Centre and the development and software specialist in-tech. The goal is a fully integrated, digital and data-based insurance solution based on the data of the installed assistance systems. Unlike telematics approaches, data collection and risk assessment are to be carried out directly by the vehicle.

Manufacturer-independent and transferable approaches are being tested in which data from up to twelve assistance systems, for example Adaptive Cruise Control, Park Pilot and Lane Change Assist, are collected via original interfaces such as ODB-II or CAN bus in order to quantify their efficiency and the corresponding influence on the need for damage – all of this, of course, always under the premise of the highest level of data protection conformity.


Enormous potential in the data of modern, connected vehicles

In addition to motor vehicle insurance, there are also other interesting application possibilities for data-driven solutions, for example with a view to high-voltage batteries and a valuable remarketing of electric vehicles on the used car market. For corresponding used car programmes, battery certificates or even interactive tips on battery-saving driving, stable forecasting models are needed that can only be developed with the help of real vehicle data.

For the time being, access to the data remains critical to success. Currently, the car manufacturers have the power of disposal, which is viewed very critically by insurers, workshops and automobile clubs. The GDV called for an end of the manufacturers' data monopoly only at the beginning of the year. The ADAC has also taken a clear position and demands that the data should belong to those who “produced” it – and that is the drivers. 

The EU Data Act, which is to come into force in 2025, is a first step in the right direction, but it is very general. For example, it also applies to networked refrigerators and televisions. It is therefore doubtful whether it will do justice to the complex overall situation surrounding vehicle data. Decided regulation that takes into account the special requirements of manufacturers, insurers, users and all other stakeholders would be more effective.

All in all, I am convinced that there is enormous potential in the data of modern, connected vehicles. The goal must be to establish digital insurance solutions for all sales channels based on simple processes. The customer must be completely freed from any avoidable administration. I see the quality impression of Bosch products in cars as an example for the insurance industry – you don't see them, but you have a good feeling when they are on board.

Ähnliche Beiträge