A leading financial institution was suffering from data quality issues pertaining to its internally generated price index. The company stored transactional records in a data warehouse used to monitor and forecast housing prices. As a part of the index calculation process the company had a specific requirement to segregate distressed sales records reflecting abnormal market conditions. Since not all distressed sales were easily identifiable within the data warehouse, the company decided to implement complex recognition logic aimed at identifying and segregating distressed sales records prior to generating the price index.
Working as a liaison between the company’s business and technology stakeholders, Baker Tilly effectively managed a team to design, construct, test and deploy the complex recognition logic by following a proven software development lifecycle methodology. First, Baker Tilly worked with the company’s stakeholders to capture the business requirements of the desired solution. From there, Baker Tilly worked with the technology team to prepare functional requirements to enable design and construction by the development team. Following design and construction, a robust approach to system and user-acceptance testing was deployed. The testing approach involved high-level “black-box” testing focusing on the inputs and outputs of the system as well as detailed “white-box” testing focusing on the successful execution of each step of the development code. Upon completion of each round of testing, defects were effectively remedied and adequate regression testing was performed to ensure system continuity. The deployment activities included scheduling multiple teams and analyzing the impacts on inter-related systems to ensure minimal system downtime.
Upon deployment of the recognition logic, the company realized significant improvement in the accuracy of its price index. The increased accuracy enabled the company to obtain a better picture of the overall health of the housing market as well as pursue plans to publish the index externally. Additionally, the company was so pleased with the results that it plans to incorporate the information generated by the solution to improve other forecasting models and research initiatives.