With the boom of digital landscape architecture, the evaluation of landscape efficacy has evolved from pictorial description to model analysis. Many scholars have explored the mathematical model behind traditional geometric laws. Structural equation modelling (SEM) provides a scientific measurement tool for landscape performance, thanks to its simple path, simultaneous equations, and in-depth variable measurement. Inspired by the theories of the SEM, this paper establishes a theoretical model for landscape efficacy based on five unobservable factors, namely, spatial density, vegetation coverage, recreational travel, scenic integration, and topological depth. Next, typical new Chinese style (NCS) commercial streets were selected as samples, and the initial data on the five factors were extracted and normalized by depthmapX, Grasshopper, and SPSS. With the aid of SPSS Amos, the path between each unobserved element and landscape efficacy was simulated, quantified, and fitted by our model. The fitting results were verified and analysed through multiple regression. A total of 200 typical NCS commercial streets were selected as samples to verify the established model. The results show that topological depth (path coefficient: 0.98) has the greatest impact on landscape efficacy, followed in turn by spatial density (0.86), vegetation coverage (0.65), scenic integration (0.59), and recreational travel (0.50). The empirical analysis proves that the SEM can effectively evaluate the landscape efficacy of the NCS commercial streets. The research results provide a good reference for the quantification of the coupling relationship between unobserved factors in built environment.