Perceived Performance in Web Applications : Comparing User, Developer, and AI Evaluations of Loading Strategies

Sammanfattning

Background. Perceived performance is a critical aspect of user experience (UX) in modern web applications, particularly during loading states. While various loading strategies such as spinners, skeleton screens, and progressive rendering are widely used, their actual impact on user perception is not fully understood. Objectives. This study aims to investigate how different loading strategies influence perceived performance and interaction quality, and to examine how developer assumptions and AI-based evaluations align with actual user preferences. Methods. A controlled web-based prototype was developed to simulate five loading strategies under identical conditions. A total of 20 end users evaluated their experience across multiple tasks, while 5 developers and 4 AI models provided predictions regarding user preferences. Data was collected through structured rating scales and qualitative feedback. Results. The results show that progressive rendering was consistently rated as the most effective loading strategy across all user experience metrics. In contrast, developers and most AI models predicted that skeleton screens would be preferred. This reveals a clear mismatch between predicted and actual user preferences. Conclusions. The findings demonstrate that perceived performance is strongly influenced by continuous feedback and incremental content delivery rather than static placeholders. The study highlights the importance of empirical validation in UX design and suggests that commonly assumed best practices may not always alignwith real user experience.