<![CDATA[Background: The persistently low numeracy performance of Indonesian elementary students, as indicated by the PISA 2022 results and the national Education Report Card, highlights the need for contextual mathematics instruction that connects mathematical concepts with authentic learning experiences. Aims: This study investigated the effectiveness of project-based STEM learning through vertical garden design in enhancing fifth-grade students’ numeracy and mathematical problem-solving compared with the Direct Instruction model. Method: A quasi-experimental study employing a nonequivalent control group design was conducted with 46 fifth-grade students, comprising an experimental group (n = 23) and a control group (n = 23). Data were collected using validated and reliable numeracy and mathematical problem-solving tests and analyzed through descriptive statistics, normalized gain (N-Gain), paired-sample t-test, Wilcoxon signed-rank test, independent-sample t-test, and Cohen’s effect size. Results: Project-based STEM learning significantly improved students’ numeracy (p < 0.001) and mathematical problem-solving (p < 0.001). The experimental group achieved significantly greater numeracy improvement (N-Gain = 0.68) than the control group (N-Gain = 0.52), with a large effect size (d = 0.914). In contrast, no significant between-group difference was found in mathematical problem-solving improvement (p = 0.428; d = 0.236), likely due to a ceiling effect. Conclusion: Project-based STEM learning through vertical garden design effectively enhances elementary students’ numeracy and mathematical problem-solving. Its strongest contribution lies in improving contextual numeracy, while its comparative advantage over Direct Instruction depends on the specific mathematical competency being developed.]]>
