The success of mobile applications is intrinsically linked to the quality of their User Interfaces (UIs), yet an open gap remains in the systematic integration of app store metadata with semantic interface components. This study addresses this challenge by employing the Design Science Research Methodology to develop and analyze two comprehensive and complementary artifacts: the Automated Insights Dataset (AID), containing 48 technical and market metadata types from 6,400 applications, and the User Interface Depth Dataset (UID), which features a detailed manual mapping of 50 UI component types and 1,948 screenshots from 400 high-quality apps. Moving beyond descriptive statistics, this research performs a multidimensional analysis that uncovers latent design patterns and correlations between interface elements, application categories, and visual identities (characteristic colors). Furthermore, we demonstrate the practical utility of these datasets through a predictive modeling experiment using Natural Language Processing, which successfully infers UI composition from textual descriptions with accuracy levels exceeding 90% in controlled evaluations. The results provide a robust empirical foundation for data-driven design, offering actionable insights for researchers and practitioners to ground their decisions on real-world market evidence and established design conventions.
Kuspil et al. (Wed,) studied this question.