ANALYSIS OF METHODS AND MODELS FOR ESTIMATING THE DURATION OF BANKING SOFTWARE DEVELOPMENT PROJECTS/авторів LIUDMYLA LATANSKA, ANTON POLOVYNKA

Abstract

This paper addresses the issue of inaccurate duration estimation in software development projects, with a particular focus on projects implemented in the banking sector. The relevance of the topic is substantiated by the high failure rate of such projects, a significant portion of which are not completed within the planned timeframe, resulting in substantial financial losses and reduced project management efficiency. The key causes of schedule deviations are analyzed, including the specific characteristics of banking software, such as regulatory constraints, architectural complexity, integration of numerous external services, high security requirements, and multi-level approval processes on the part of the client. It is clarified that traditional models, such as COCOMO and ISBSG, often fail to reflect the actual distribution of empirical data and industry-specific factors, which reduces the accuracy of estimation. The results of multivariate normality testing of project duration and effort distribution for banking software projects, based on a dataset of 482 real projects implemented between 2014 and 2024, are presented using Mardia’s criteria. The data were found to deviate from the normal distribution. Based on the same dataset, six COCOMO models and three ISBSG models were built – both for the full dataset and for subgroups categorized by project type (single-team and multi-team). A comparative analysis of model quality showed that none of the resulting models achieved acceptable estimation accuracy based on MMRE and Pred(0.25) metrics. The study justifies the need to improve estimation approaches by developing specialized nonlinear regression models using data transformation techniques. These models would enhance estimation reliability and enable interval-based forecasting. The results are of practical value for analysts, project managers, and planning specialists involved in banking software development.