The aim of this study was to retrospectively evaluate the accuracy and reliability of a skeletal cephalometric analysis based on the divine proportion (Sassouni's analysis) in predicting "divinely proportioned" faces in the treatment of dentofacial deformities. Preoperative and postoperative frontal photographs of 50 patients were analyzed, and the following 5 proportions were measured: (1) TR-AL:AL-ME, (2) ME-AL:AL-LC, (3) LC-CH:CH-ME, (4) LC-AL:AL-CH, and (5) ME-CH:CH-AL (TR indicates trichion; AL, alar rim; ME, menton; LC, lateral canthus; CH, cheilion). For each proportion, the effect of the treatment on the absolute difference between the proportions and the golden number (Φ) (1.618) was statistically analyzed. Analyses were adjusted for sex, age, type of dentofacial deformity (anteroposterior and/or vertical), and surgical technique (1-jaw vs 2-jaw surgery with or without a genioplasty). In all of the cases, the facial proportions analyzed were not significantly changed by the surgical treatment except for the proportion ME-CH:CH-AL (lower facial third), which was significantly moved toward 1.618 (P < 0.01). The current study demonstrated that (1) facial divine proportion can be closely approximated and predicted only by a specific cephalometric analysis in the lower facial third and (2) parameters such as age, sex, type of dentofacial deformities, type of surgery, and so on, could potentially and significantly influence the final results toward the divine proportion.