The pricing strategy in the field of insurance is very important, if the premium is very low compared to the competition, high-risk clients will be attracted; therefore, there will be a greater number of reported claims, having the effect of reducing profit margins. On the contrary, if the premium is very high in relation to the competition, there will be no higher sales. The main factor for defining prices is risk prediction , hence the importance of the estimated risk being as accurate as possible.

But what is customer risk? Well, it comes to be the probability that something bad or harmful will happen to the client or to the good that is insured. It can involve illness, accidents, property damage, among others.

How is risk calculated? Actuaries do this by using statistics and linear regression, methods used until the 1980's, when a model with advantages over linear regression called the Generalized Linear Model (GLM) was established.

Although GLM is the most used model in insurance companies, at present and thanks to all the technological advances in the field of Data Science, the use of Machine Learning has become popular to obtain rates based on both the frequency and severity of the accidents. Using models such as Random Forests and Boosted Trees is intended to improve predictions and ensure less error in price estimation.

Claims predictions can be made based on the company's historical data, but can be supplemented by using statistical data from the general population to calculate risk, especially with data from the field of Health. For example, seasonal diseases in an area such as dengue, which is present in approximately 80% of the Ecuadorian territory (warm and humid sectors), coincide in that the epidemic cycles are associated with the rainy season.

Data science and insurance

The estimation of risks, prices and claims are not the only applications of Data Science in insurance, the following use cases are also available:

Insurance fraud brings many losses in this industry, and can not only come from customers but also from interested parties such as brokers or service providers (for example, clinics or hospitals).

Predictive models based on historical data of fraudulent activities can be used, in this way we can detect ongoing fraud. We may also be able to identify suspicious links and activity before fraud occurs.

In today's world, where people increasingly require products that adapt to their lifestyle, insurance companies face the challenge of maintaining fluid communication with their customers to understand their needs.

The analysis of demographic data, preferences, interests, hobbies, lifestyles, fashion activities, among others; they can help companies set up personalized offers, rewards programs, or new products.

For example, interest in pet care has now grown, this could be an opportunity to offer veterinary medical plans, including including them within family plans.

Customer Lifetime Value is a forecast of the value a customer represents to a company. It is based on the difference between the projected profits and the projected expenses of the future relationship between client and company.

The direct application of the CLV is what allows deciding the amount that can be spent to get a new client, this amount must be less than the CLV for the business to be profitable.

Predictions can be made based on historical data from the same company, as well as demographic, interest and preference analyses, as in the previous point.

In conclusion, the application of statistics in the field of insurance is not something new; therefore, it is very logical that Machine Learning tools are currently used to improve these processes, mainly for the prediction of risks and claims, thus allowing a premium to be established with greater accuracy.

Additionally, Data Science can be used to detect and prevent fraud, perform personalized marketing and establish a clearer Customer Lifetime Value (CLV); These applications are not only used in the insurance field, but can be used in any type of industry. In this sense, the practical application would be to identify possible use cases for your line of business and be the first to implement initiatives focused on data analysis.