Learning Analytics-Supported Flipped Classroom and EFL Students' Mastery of Conditional Sentence Writing: Evidence from a Senior High School in Indonesia

Abstract

Conditional sentences remain one of the most persistent sources of difficulty for EFL learners, and conventional grammar instruction rarely gives teachers a way to intervene before that difficulty hardens into error. This study tested whether pairing a flipped classroom model with learning analytics drawn from Google Classroom could raise senior high school students' mastery of conditional sentence writing beyond what conventional instruction produces. Using a quasi-experimental, unequal control-group design, two intact eleventh-grade classes at SMAN 1 Bangsal, Indonesia (n = 36 each) were assigned to an experimental group taught through a Learning Analytics-Supported Flipped Classroom (LASFC) model and a control group taught conventionally, across an eight-week intervention. Writing ability was measured through a pre-test and post-test scored on four dimensions: grammatical accuracy, contextual appropriateness, structural complexity, and unity in writing. Welch's t-test showed a statistically significant post-test advantage for the experimental group, M = 83.33 (SD = 8.91), over the control group, M = 79.22 (SD = 3.32), t(44.554) = 2.595, p = .013, with a medium effect size (Cohen's d = 0.61). A secondary analysis of normalized gain produced an apparently contradictory result: the experimental group's gain (⟨g⟩ = 0.16) was lower than the control group's (⟨g⟩ = 0.53). This study argues that the discrepancy is best explained by a ceiling effect tied to the experimental group's markedly higher pre-test score (M = 80.22 versus M = 56.22), not by any weakness in the intervention. The findings support the use of learning-analytics data—material-access records, quiz scores, and assignment-completion timestamps—as a diagnostic layer within flipped instruction, and they caution against relying on normalized gain alone when comparing groups with unequal starting points.