Supporting machine-learning advancement of underwriting processes using online surveys

Traditionally, insurance companies rely on actuaries to calculate premiums. Actuaries are professionals who use mathematical and statistical modeling to assess risk and estimate the likelihood of a claim based on various variables like age and gender. They use this information to create actuarial tables which are then used by the insurance company's underwriting department to calculate premiums. However, with the advancement of technology, specifically Machine Learning (ML) and dedicated software, insurance companies can now calculate and write programs more efficiently.

Artificial Intelligence (AI), particularly ML, is a promising technology that has been successful in various aspects of the insurance process such as image detection, sentiment analysis, and anomaly detection. However, these models require accurate input to train them effectively. The common problem is that a large portion of the input comes from the applicant, making it susceptible to errors due to the declarative nature of the data. Therefore, using validation input that is different from user-stated data would provide significant value in improving the quality of future premium calculation models.

One of our clients approached us with a brief, asking us to support the process of obtaining actual image-based data to train these models and improve the calculation of insurance premiums. We understood that commonly used platforms such as M-Turk were not an option due to their low-quality output and standardized approach. Therefore, we proposed alternative solutions that would provide higher quality and more relevant data for training the models.

At TGM Research, we understand the importance of accurate and reliable data when it comes to Machine Learning (ML) systems. That's why we've developed a unique technology specifically for this custom research project, integrating a validation module that utilizes participants' webcams to capture high-quality selfies.

This validation process ensures that only participants who provide selfies that meet the client's specific requirements (such as removing glasses, hats, and ensuring proper lighting) can participate in the survey. This allows us to map a wide range of data points, including facial features and skin characteristics, alongside information about participants' lifestyles, such as whether they smoke or not.

We take privacy and data protection seriously at TGM Research. All data is anonymous and follows ESOMAR principles, with clear instructions about the use of submitted information. We ensure that no sensitive data is processed alongside information that could potentially identify participants, in compliance with all codes of conduct.

With this methodology and unique validation approach, we rolled out the project in 5 countries, including South Africa and India, where the quality of smartphones and camera resolution is a challenge.

We were able to develop a fast and smart custom solution for one Client - a solution that no other panel companies have, to provide data of a total N=2500 (500 valid anonymized selfies per country) with additional questionnaire input to be used for data training.