Software Development of Business Intelligence Dashboards: Empirical Study of COSMIC ISO/IEC 19761 Size-Based Effort Estimation Using Machine Learning Techniques

Measuring the functional size of business intelligence (BI) and digital marketing analytics dashboard development can provide quantitative input for estimating the required time and cost of software development efforts. However, digital analytics practitioners currently lack practical guidance for planning dashboard development using measurement-based inputs that support reliable estimation models. Therefore, this study proposes the first Common Software Measurement International Consortium (COSMIC) ISO/IEC 19761 framework to measure the functional size of software user requirements of Power BI, Data Studio, and Tableau. An empirical study was conducted to measure the functional sizes of two datasets based on COSMIC. These functional sizes were considered inputs to several machine learning (ML) models to predict the effort required for BI dashboard development. The outcomes of these ML models showed that projects with consistent, low-variability effort should be prioritized when collecting or selecting data for training effort-estimation models and recommended the use of Stacking Regressor, Linear Regression, and Voting Regressor models for similar datasets because these models recorded the highest standardized accuracy for predicting BI development efforts with high generalization performance. These findings demonstrate the importance of applying COSMIC as a software engineering standard to BI and digital marketing analytics projects by offering measurements that allow for decisions informed by data rather than intuition or guesswork.