This study aims to find out: (1) The effect of the application of the DLDPF model on improving TDBI writing proficiency; (2) Differences between student learning outcomes with the DLDPF model and without DLDPF; and (3) Student obstacles in the application of the DLDPF model. The design of this study is a pseudo-experiment with a non-equivalent comparison group design scheme. The researcher used a nonprobability sampling technique in the form of total sampling. The instruments used in this study are tests and questionnaires. The test instrument is in the form of an essay test with four aspects of assessment criteria (content, organization, grammar, and vocabulary). There are also document study instruments and semi-open questionnaires with the Likert Scale. The results of the study revealed the following findings. First, the results of the paired sample test showed a sig. (2-tailed) of 0.00 < 0.05 which means that there is a significant influence of the application of the DLDPF model on the improvement of TDBI writing proficiency. Second, the results of the independent sample t-test showed a sig. (2-tailed) of 0.00 < 0.05 which means that the application of the DLDPF model has a significant impact on improving TDBI writing proficiency. In addition, the experimental class has an N-gain score of 0.59 which means it has a moderate level of improvement. The control class has an N-gain score of 0.24 which means it has a low level of improvement. In other words, it shows that there is a difference between student learning outcomes with the DLDPF model and student learning outcomes with no DLDPF model. And third, respondents showed few obstacles in the application of the DLDPF model, including: (1) Most of the respondents (70.53%) did not experience linguistic barriers; and (2) Most respondents (70%) did not experience nonlinguistic barriers.
