Mathematics teaching in Indonesian vocational secondary schools (SMK) remains dominated by teacher-centred routines that leave many students with shallow conceptual understanding and weak learning readiness. This study compared a flipped-classroom package with conventional instruction in a quasi-experimental pre-test-post-test control-group design: two intact Grade XI classes at SMK Negeri 1 Painan were allocated to flipped (n = 30) or conventional (n = 31) conditions during a unit on circles and transformations. Because allocation was at class level and not random, class-specific influences are confounded with condition, and estimates are adjusted group differences, not causal effects. Conceptual understanding was measured with a double-marked rubric-scored essay test and learning readiness with a Likert questionnaire covering physical, psychological, cognitive, and environmental dimensions. A MANCOVA with both pre-tests as covariates returned a significant multivariate difference, Wilks' Λ = .49, F(2, 56) = 29.13, p < .001, partial η² = .51. Follow-up ANCOVAs showed adjusted differences of 9.75 points on conceptual understanding, 95% CI [6.09, 13.40], partial η² = .33, and 8.24 points on learning readiness, 95% CI [5.32, 11.16], partial η² = .36, both surviving Holm correction. Normalized gains were medium in the flipped class and low in the conventional class, and estimates held under Welch and rank-based tests. With one class per condition, one teacher, and one school, these results describe a difference in this setting and cannot separate the model from novelty, access, and compliance.