Web-based problem-solving, a compound ability of critical thinking, creative thinking, reasoning thinking and information-searching abilities, has been recognised as an important competence for elementary school students. Some researchers have reported the possible correlations between problem-solving competence and information searching ability; however, few studies have proposed what factors might affect students' web-based problem-solving performance. In this study, a five-phase web-based learning approach is proposed; moreover, a web-based problem-solving model was developed to investigate the factors that might affect students' web-based problem-solving performance. To evaluate the effectiveness of the proposed approach, an experiment was conducted by engaging 201 fifth- and sixth-graders from three elementary schools in a series of web-based problem-solving activities. Structural equation modelling was employed to analyse the antecedents affecting the students' web-based problem-solving performance. From the analysis results, it was found that task-technology fit could be the major factor affecting the students' intention to learn on the Web, as it highly affected their web-based problem-solving performance.