INSURANCE PRODUCT IMPROVEMENT AND ESTIMATING THE SURCHARGE TO THE PURE PREMIUM

The idea of insurance can be traced back as far back in history as the era of Babylonian around 4000-3000 BCE. The first recorded example of an insurance product was found in The Hammurabi Code, where it is stated that if a trader pays a forgiveness fee to the loan issuer, he does not have to pay back the loan if the goods were stolen or lost in flooding. Insurance practices kept evolving in parallel around the word, until the concept of insurance that we know today took form. Insurance is defined as the transfer of risk from one party to another in exchange for a pre-specified price. Nowadays, the insurance industry is one of the biggest globally. Currently, the insurance business makes up a big part of the financial industry, it represents 2.9% of the total US gross domestic product (GDP) in 2021. Furthermore, on a micro-level, many financial institutions have established their own insurance company as an integrated part of their entity. BNP Paribas Cardif is an insurance company that operates on a global scale. It is a part of the French group BNP Paribas, and its head office for the Nordic countries is located in Gothenburg, Sweden. The main products sold in the local Nordic entity are CPI products (Creditor Protection Insurance). CPI products are sold to those customers who are taking out loans, mortgages or credit cards. The insurance will cover their outstanding balance or monthly instalment in certain events such as death, unemployment or hospitalisation for instance. As it is the case in almost all fields, a better product results in higher sales performance. It was found that product improvements related to quality have positive and significant correlation with higher corporate performance and profitability. The same idea can be applied to the insurance business. Each policy sold is a unit of our product and striving for a better product is the key to market growth and gaining more customers, as well as maintaining a good satisfaction level for existing customers. From a more quantitative perspective, when it comes to insurance products, there are two key performance indicators that reflect product quality and customer satisfaction: acceptance rate, and customer value. By acceptance rate, we mean the ratio of accepted claims to all reported claims, while customer value is the ratio of the amount of claims paid out to customers to the total written premium earned. The importance and relevance of acceptance rates comes from the fact that rejecting the majority of reported claims certainly has an adverse effect on the product quality from the clients perspective. As for the customer value, which can be referred to as loss ratio in traditional insurance, we can consider it as a measure of how much value an insurance firm gives back to the group of customers as a whole compared to what is earned. Therefore, one way to improve the quality of the insurance products is to enhance these two indicators. As we will see later, increasing the former will automatically improve the latter. High acceptance rates is one the main focus points of any insurer nowadays, and BNP Paribas Cardif (BNPPC) is no exception in this matter. In order to boost the acceptance rates and provide even higher customer value, BNPPC launched a project for this purpose. The IU boost project aims to increase acceptance rates through implementing several product improvements, such as making the eligibility criteria more customer-friendly and removing many exclusions from the terms and conditions. This shall leads to accepting more claims and guarantee a higher customer value of our products well-above the internal threshold set by the head office of the group. The project has been launched after noticing a considerable difference between the acceptance rates in BNPPC Nordics and BNPPC in other countries around the world, in particular in Europe. A study done by head office in Paris showed this disparity. The study showed that on average, for a product in the Nordics, the acceptance rate is around 8\% lower than its counterpart in European countries. Hence, BNPPC Nordic took the initiative to launch this project, with the main underlying goal of increasing the acceptance rates, as well as offer an improved standard product, exceeding those currently in the market in its quality and value for the end customer. Additionally, the improved product will have brief and simple terms and conditions, that can be checked readily. Another quite important benefit of this project, is that it will significantly facilitate the Claim Automation project, which is another on-going project at BNPPC to automatise the claim handling process. The project aims to have a fully automated digital customer journey for BNPPC clients during the claim phase. When the Claim Automation project is ready and actively used, BNPPC will receive information from the National Job Office in Sweden, that a person has registered themselves as unemployed. If the person exists in our database as a policyholder, and is covered for unemployment risk, this person's data will be passed through an AI scoring model. If the AI model determines that the policyholder is likely eligible to claim and receive the benefit, we will notify them and remind them to use their insurance policy and claim for unemployment benefit. This complete process will be automated, and it requires no active involvement from the customer until paying out the benefit. Most function allocation methods (FA) used to allocate the task between humans and automated machines and systems, take into consideration capabilities of humans and automation, and other factors such as cost, regulations, and health and safety issues. For BNPPC, when similar factors have been taken into account, the claim automation project proved to be quite promising. Furthermore, since BNPPC's vision is to grow at such a fast rate as a leading company in CPI insurance in the Nordic countries, the success of the Claim Automation project will mean the capacity to handle larger volumes of claims without the need to invest more resources in the claim department. Therefore, it is quite useful that the IU Boost project will produce a product with simple terms and conditions, as it will be easier to parameterise in the AI model. The main question at hand is the following: What is the impact of applying a certain product improvement on the acceptance rate, and eventually the pure premium, of a product? The challenges underlying this question are more operational, namely about understanding the data, rather than technical as we will see in the coming sections. Nonetheless, we propose a GLM approach that might relieve those operational difficulties in the future.