A high failure rate in physics subjects is caused by students' inability to understand the material. The physics learning process certainly requires an appropriate learning approach. The learning analytics approach is one of the innovations in learning. It allows students to analyze problems. Therefore, this research aimed to determine the learning analytics approach's effect on improving students' multiple representation skills. This research is experimental. The validation of the lesson plan obtained excellent results based on two material experts and four practitioners. Furthermore, the empirical test was carried out on 504 students. The research design was the pretest-posttest control group design. The control and experimental groups were determined using cluster random sampling. The findings of this research include (1) the normality test results that are higher than 0.05, which means that the research data was normally distributed. (2) The homogeneity test results were higher than 0.05, which indicated that the data was homogeneous. (3) The paired sample T-test obtained a value lower than 0.05, which means that there was an influence of the multiple representation approach. (4) The N-gain value in the experimental class was higher than the control class. Lastly, (5) only 47.2% of students used graphical and mathematical representation skills. Based on these findings, the effect of the learning analytics approach on students' multiple representation skills was fairly good with moderate criteria. For further research, learning products or models can be developed to improve multiple representation skills focused on a combination of graphics and mathematics.