Abstract Study question Is it possible to create a model that combines follicle number and size to help physicians visualize the trigger requirements to predict oocyte utilization? Summary answer Findings support graphical follicle model as an innovative, simple, and practiced parameter for predicting clinical results in ovary stimulation treatment and facilitating personalized protocol adjustment. What is known already Previous studies have observed the number and size of follicles are two independent indicators of whether the oocyte is adequate and mature and are used to predict the trigger time in routine clinical practice. However, due to individualized differences, it currently relies mainly on physician experience. Study design, size, duration This retrospective study included 8405 patients started their first in vitro fertilization cycle with a GnRH-ant protocol including fresh and subsequent frozen-thawed cycles during 2016–2020. Participants/materials, setting, methods We classified follicles recorded on the HCG day by size. We then produced graphical models, and classified into Inverted-trapezoid (large follicles in dominant proportion), Polygon (moderate follicles in dominant proportion), Trapezoid (small follicles in dominant proportion), Rectangle (equivalent proportions of the three size categories). The Cochran–Mantel–Haenszel and Generalized Linear Model(GLM) were used to evaluate the difference among models in relation to maturity, fertilization, and number of viable embryos, as well as cumulative live birth rate. Main results and the role of chance In GLM analysis, after adjusting the confounders, there are differences between models of CLBR. The CLBR of the different models was higher in the Polygon and Inverted-trapezoid model than Trapezoid and Rectangle model (42.75%, 39.56%, vs. 37.38%, 28.57%, respectively; all P<0.05). For oocytes derived from very large follicles ( > 20 mm), the CLBR was lower than that of patients with ≤ 20 mm follicles [26.10% vs. 42.10%, OR = 1.74 (95% confidence interval 1.52–2.00), P<0.01] in Inverted-trapezoid model, but there was no difference between models. And the risk of ovarian hyperstimulation syndrome (OHSS) rate of patients with ≤ 20 mm follicles was lower than that of patients with > 20 mm follicles [8.64% vs. 17.89%, OR = 0.57 (95%CI: 0.49−0.65), P<0.01] in Inverted-trapezoid model. Patients who received an adjusted Gn dose(whether Decreased or Increased-dose protocol) showed no difference in CLBR among models (Polygon vs Inverted-trapezoid vs Trapezoid vs Rectangle model: 47.07% vs. 49.21% vs. 47.69% vs. 42.42%, P > 0.05); but when patients continued with the same starting dose in Fixed-dose protocol, the CLBR of the Polygon model was higher than that of other models (40.43% vs. 34.32, 31.13% vs. 26.46%, respectively; all P value <0.05). Limitations, reasons for caution This study is retrospective. Due to analysis of real-world data, the follicular corresponding outcome following measurement was not obtained. Besides, the majority of patients received 2 cleavage stage embryos transfer, this approach may impact the external validity of follicle model. Wider implications of the findings The follicle models can demonstrate ovarian response of patients. It is better for patients to reach the Inverted-trapezoid model with dominate follicles > 18 mm and <20 mm. Adjusting the protocol is critical to the outcome. But how to improve follicle models by regulating stimulation remains to be further studied. Trial registration number not applicable