<![CDATA[Pronunciation, a key indicator of speaking ability in the Common European Framework of Reference (CEFR), is crucial for accurate communication of technical terms; while native‑like pronunciation is unnecessary, minimizing errors prevents misunderstandings. This study examined the effectiveness of the ELSA Speak mobile application—designed specifically for English pronunciation—in the English program of Prambanan Vocational High School, Yogyakarta, where it has been used for more than two years but has not been formally evaluated. A purposive sample of 76 students (out of a total of 380) was selected and the research applied the CIPP evaluation model (Context, Input, Process, Product) together with a quasi‑experimental pre‑test/post‑test design. Four research questions guided the inquiry: (1) the needs of current students and alumni for English communication skills; (2) the resources, planning, and readiness of school management, teachers, and learners for implementing ELSA Speak; (3) the challenges encountered by students, alumni, and teachers during use; and (4) the degree of pronunciation improvement attributable to the app. Contextual analysis revealed a strong demand for English, especially Received Pronunciation, to enhance career prospects. Input evaluation identified barriers such as limited internet access, mother‑tongue interference, and phonetic complexity. Process results showed statistically significant gains in pronunciation (t = 7.885, p < 0.001) and increased learner engagement (t = 4.88, p < 0.001). Product outcomes highlighted notable improvements in word‑level intonation and stress patterns. The school addressed input challenges through offline materials, targeted teacher training, and clear usage guidelines, which contributed to the observed performance gains and positive student feedback. The study concludes that ELSA Speak markedly enhances pronunciation for vocational learners and recommends its continued adoption with structured support. Future research should involve larger, multi‑major samples to confirm generalizability]]>
