Insurance price optimization
Construction of a price optimization engine for non-life insurance leading to the improvement of the pricing strategy.
A key aspect of the operation of an insurance company is pricing. Only by proper pricing of insurance policies can a given insurance company maintain profitability. In case of low pricing, the company may become insolvent. In case of high pricing, the offer will have a non-market price, which will cause customers to choose offers from other companies.
Typically, pricing is associated with two types of models: the risk model and the demand model. Actuarial state-of-the-art uses statistical modeling methods (such as GAMs). However, in the constantly ongoing price war on the insurance market, the process of pricing must be continually improved. Thanks to data collected over many years, advanced machine learning algorithms can now be used in pricing. Their use allows for more complex phenomena to be included in the pricing process.
In our work in the field of actuarial data science, we were responsible for developing tools used in pricing. Primarily, we were involved in modeling - from data quality checks to modeling and implementation. Additionally, we were involved in R&D projects such as building an engine that optimizes insurance pricing and using denoised autoencoders to reduce the dimensionality of input data for modeling. In addition to programming, we also took care of the business management of the models, so that the implemented models met the customers' expectations.
By systematically monitoring the performance of the models we implemented, we were able to improve the positioning of insurance offers in online comparison websites.