Large Loss

Overview

• By combining Mind Foundry's world-class AI and Machine Learning capabilities with Aioi Nissay Dowa Europe’s (AND-E) rich driving dataset and expertise in auto insurance, we rapidly created a highly effective model that predicts the likelihood of major accidents occurring. This allows preventative action to be taken that can significantly reduce claims costs whilst potentially preventing severe injuries and even saving lives.
   

Detail

Background

• Identifying and predicting large losses within the auto insurance market is a formidable challenge, primarily due to the inherent complexities. These complexities stem from the diverse and dynamic characteristics of these risks combined with the infrequent yet severe nature of large losses. Additionally, data quality issues and fragmented information can further compound the problem.

​

Hypothesis

• Innovative AI and Machine Learning solutions can be applied to address these challenges and help predict the occurrence and nature of large losses. The findings can then be used to inform programmes of activity to reduce the human and economic consequences of these terrible events.

​

R&D

• AND-E’s knowledge of driving behaviours captured in its massive bank of UK telematics data together with range of other data sources, such as geospatial and environmental, were combined into a single AI and Machine Learning solution.

• This solution looked at the likelihood that a driver may suffer a severe accident resulting in a large loss for the insurer, as implied by their driving behaviour.

• Machine learning models were used to analyse continuous behavioural and historical driving data, looking specifically at a customer's driving patterns, like speeding and breaking, as well as road familiarity because the data showed that drivers are more likely in certain circumstances to have an accident on roads that they are more familiar with.

• This is then paired with road risk data, such as how many accidents there are on a certain motorway each year.

• The models are then aggregated and enriched with human insight from AND-E’s team of auto insurance experts to provide an all inclusive risk score that can be used in a range of ways to reduce large losses such as providing enhanced driving advice to insurance customers.

 

Next step

• Since the approach utilises the principles of Kaizen by using on-going experiential data to continuously improve the insights that are generated, the benefits of the initiative will continue to be realised.

• The positive impacts were apparent very soon after being implemented in the UK in 2023 with a reduction in the large loss ratio of 3% percentage points after only a few months for the risks to which the model was applied.

• Plans are in place to expand this know-how beyond the UK to reduce the number of tragic accidents in other countries.