<![CDATA[Grammar accuracy remains a persistent challenge for vocational EFL learners, particularly in rural Indonesian contexts where opportunities to engage with authentic language use may be limited. This study examined changes in grammatical error frequency following Corpus of Contemporary American English (COCA)-assisted Data-Driven Learning (DDL) in procedural text writing. A pretest–posttest design was employed with 76 eleventh-grade vocational high school students in Rokan Hilir, Riau, Indonesia. Students completed grammar-oriented procedural writing tasks before and after the intervention. Grammatical performance was assessed through error counts across five categories: nouns, verbs, adjectives, articles, and prepositions. Descriptive statistics and a paired-samples t-test were used to examine pretest–posttest changes. The mean number of grammatical errors decreased from 35.80 (SD = 31.65) to 30.00 (SD = 27.48), representing a mean reduction of 6.00 errors. The paired-samples t-test indicated a statistically significant difference between the pretest and posttest, t(75) = 7.755, p = .001. At the category level, total errors decreased from 179 to 150, with the largest reductions occurring in article and verb errors. These findings indicate improved grammatical performance following the intervention. However, because the analysis was based on a pretest–posttest comparison and did not establish a valid between-group effect, the improvement cannot be attributed exclusively to COCA-assisted instruction. The findings suggest that COCA may serve as a supplementary resource for grammar and procedural writing instruction when combined with teacher guidance, explicit instruction, and sustained practice.]]>
