Automated Validation of Insurance Applications against Calculation Specifications
Advaita Datar, Amey Zare, Asia A, R Venkatesh, Shrawan Kumar, Ulka Shrotri
TL;DR
The paper addresses the heavy manual effort required to QA modern Policy Administration Systems (PAS) against legacy Calculation Specification (CS) sheets. It introduces BVCS, a schema-driven, low-code/no-code framework that automates validation through three steps: schema generation from CS sheets, data collection from PAS via parameterized scripts, and comparison of CS-generated outputs with PAS results, all presented in an interactive dashboard. Experimental evaluation across 12 CS sheets and 254 policies demonstrates ~92% average effort savings over manual validation, with batch-mode execution showing ~91% savings and scalable validation for bulk data. The work offers a practical method to improve accuracy and efficiency in insurance QA during digital transformation and suggests applicability to other financial domains.
Abstract
Insurance companies rely on their Legacy Insurance System (LIS) to govern day-to-day operations. These LIS operate as per the companys business rules that are formally specified in Calculation Specification (CS) sheets. To meet ever-changing business demands, insurance companies are increasingly transforming their outdated LIS to modern Policy Administration Systems (PAS). Quality Assurance (QA) of such PAS involves manual validation of calculations implementation against the corresponding CS sheets from the LIS. This manual QA approach is effort-intensive and error-prone, which may fail to detect inconsistencies in PAS implementations and ultimately result in monetary loss. To address this challenge, we propose a novel low-code no-code technique to automatically validate PAS implementation against CS sheets. Our technique has been evaluated on a digital transformation project of a large insurance company on 12 real-world calculations through 254 policies. The evaluation resulted in effort savings of approximately 92 percent against the conventional manual validation approach.
