Before the computer was invented, back-office management purposes in order to record, process, and restore large volumes of information were processed in paper format. The period 1960-1970 was characterized by the adoption of information technology which was applied by a very limited number of users, but it already enabled a considerable reduction in processing times and substantial savings. Since 1970, the development of machines, programming languages, and the Internet allowed industries to use information technology for accounting and management control, execution of payments and cash management, operational and financial planning, payroll management, and more generally human-resource-related issues, monitoring, and measure of business activity. Moreover, the development of Internet use through smartphones and the development of social networks have accelerated the use of big data in insurance. It started bolstering the development of a solution-oriented ecosystem in the insurance industry.
An Industry Developed Through The Use Of Data
The explosion of the Internet bubble helped industries, companies, and individuals toward expanding commercial relationships. The use of big data in insurance industry was one of the first to adopt information technology — an ideal tool for carrying out numerous, successive and segmented tasks. This adoption has profoundly altered the relationship between insurance companies and their clients, and optimization of management has gradually given way to a real commercial struggle, gradually making insurance a consumer product. From the advent of information technology within the industry, data represent a fundamental asset for insurance companies, the latter being necessary for product pricing, reserving, and claims management.
In insurance companies, data are at the heart of the relationship between policyholders and insurers that help in establishing the conditions for their mutual commitment. Insurers conduct their businesses and manage the risks to which they are subject by anticipating in particular the probability of occurrence of future claims. As regards prevention and claims management, access to an increasing volume of information and the ability to process these data in real-time or near real-time contribute to better support of the policyholder and a reduction in claims cost.
Link Between Data And Insurable Assets
To be insurable, a risk must be random, futural, lawful, independent of the will of the policyholder and sufficiently common to be subject to calculation of its probability of occurrence without being almost certain.
The digitalization of the relationship between insurers and policyholders significantly increases the frequency of data collection. The matching capacities of data collected with external data increase the diversity of the available information and the machine learning methods make it possible to identify the most discriminatory variables in order to predict behavior or the occurrence of a given risk. In effect, the vision of risks tends to be refined and price variations may occur:
– infra-annual variations related to changes in the consumption habits of policyholders during the year. (For example, seasonality of automobile use or occupation of a secondary residence.) Externally, pricing and, by extension, insurance to the act, based on data from connected objects, could be conceived as of future events.
– refined segmentation of guarantees within a contract and ability of the policyholder to underwrite only a part of the guarantees offered. For example, home insurance contracts often include lump-sum guarantees that may be unsuitable for young clients.
Examples Of Application In Different Insurance Activities
This section aims to highlight existing examples, examples under development or potential examples of the application of big data relevant in insurance. These examples, while varied, are in no way an exhaustive list of possible applications of big data in insurance, and it is up to each insurer to analyze the potential contribution:
– of its own data, as regards both already used data and data not yet valued or data which may be collected but is yet to be.
– of external data accessible through open data, the purchase of data from external partners or implementation of sensors within the framework of a connected insurance offer.
The matching of the various data makes it possible to generate new data that may be of significant interest through feature engineering. They can also help implement a real data-driven approach limiting the cognitive biases on the explanatory character of the different variables in the explanation of a given behavior.
These contributions will be amplified by the cross-matching of big data with other emerging innovations, including:
– blockchain and smart contracts.
– artificial intelligence and the emergence of bots and chatbots.
Other innovations will contribute to the exponential generation of new data sources, including:
– the development of connected objects.
– the emergence of platforms for the collection of personal data.
Finally, if the data generated so far are mainly due to mature markets, the contribution of emerging countries is set to explode soon.
Savings In Life Insurance
In the current low-interest-rate context, insurers offering savings products such as life insurance are naturally interested in achieving an in-depth understanding of the behavioral dynamics of unit-linked investment of their policyholders. Data science, through the reuse of both internal and external data, and the use of machine learning algorithms can make it possible to obtain such an understanding. The added-value brought about by the matching of external data, notably relating to the economic environment, varies according to the cyclical nature of the various predicted acts.
Beyond the precise use that allows an understanding of the historical behavior of policyholders, it should be possible to easily reuse the implemented algorithms. For example, the past decade has witnessed an almost continuous drop in interest rates and an unprecedented financial crisis. In effect, the future cyclical behavior of policyholders may potentially differ from past behaviors. The regular reuse of algorithms makes it possible to capture behavioral changes, evolutions which will not fail to occur with the rise in interest rates expected.
