This study aimed to examine the effects data-driven learning approach on EFL learners’ academic speaking performance and engagement. It also aimed at exploring learners’ perceptions of the DDL intervention. The study used a quasi-experimental research design that employed an interrupted time series design with single-group participants. The participants were fourth-year EFL major students at Mekdela Amba University, Ethiopia. In the intervention, which lasted for eight weeks, the participants were taught target language features for speaking via the data-driven learning approach. Tests, self-report surveys, questionnaires, and the students' reflective journals were used to collect data. One-way repeated measures ANOVA and one-sample t-test were used to analyze the quantitative data, while thematic analysis was employed for the reflective journals. Consequently, the findings of the one-way repeated measures ANOVA results (F (2.248, 31.470) = 381.503; p < 0.05) indicated that the data-driven learning approach had a statistically significant effect on participants’ academic speaking performance. Similarly, the results obtained from the reflective journals also indicated that incorporating authentic linguistic features in focused noticing and production activities during the intervention contributed to improving participants’ academic speaking performance. Moreover, the one sample t-test results about the participants’ engagement in the three dimensions: the behavioral engagement result (t(14)=8.454, p< 0.05); cognitive engagement result (t(14)=8. 508, p< 0.05), and emotional engagement result (t(14)= 13.759, p< 0.05), showed that the participants’ engagement in speaking during the DDL intervention was promising. Regarding participants’ perception, the findings showed that participants felt data-driven learning was beneficial to their speaking skills development and had positive attitudes to the utilization of the DDL approach in speaking instruction.