Analytics and Automation of Ifrs 9 Modeling and Reporting: a Case Study of Kenyan Banks

Abstract

Becoming IFRS 9 continues to prove a challenge for Kenyan banks because of a lack of mature credit scoring system which most models rely on. A 2018 report by ICPAK report shows that 90% of banks rely on enterprise spreadsheet models built on MS Excel for their financial modeling and reporting functions. The primary disadvantages of building a financial model on excel is lack of error control, lack of reusability, little to no automation of common tasks, poor integration with existing data sources, limited scalability and poor maintainability. An error-prone output potentially results in the misrepresentation of a bank’s financial position. Existing automated implementations, such as the one proposed by Volarevic and Varovic (2018), rely heavily on credit scoring for Probability of Default estimations and are not suitable for local markets. ERP providers also offer automated IFRS 9 modeling solutions. However, these are vendor-specific and have high implementation costs. For instance, Surecomp a financial services solutions provider, offers a cloud-based IFRS 9 solution called IMEX at $300,000 as a flat-rate, one-time payment (Capterra, 2022). This study proposes an automated IFRS 9 model built on Alteryx that uses multi-state Markov (MSM) probability analysis to estimate the Probability of Default (PD). The probability modeling approach used relies only on historical loan information making it suitable in the local context. It also uses a low-code development platform to ensure ease of development, use and maintenance while addressing the pain points of spreadsheet modeling. The solution integrates with an existing database instance for automated data input. It also integrates with a reporting and visualization platform that summarizes the key drivers of Expected Credit Loss and inform management decisions or overlays. The model was built for one of the banks using anonymized loan data. For testing, results from the model of the model were compared to the output of the reported financials. An allowable variance of 2% was applied as advised by the bank. The results observed were congruent with the reported financials. The highest Expected Credit Loss variance was observed in the Credit Card sector at 0.45%. An evaluation of the packaged solution (Alteryx app) was then done by administering the System Usability Measurement Inventory (SUMI) questionnaire to potential users. Generally, the system received positive reviews on the following metrics – Efficiency, Learnability, Affect and Control. The system was also evaluated on improving efficiency of the current approach.