Abstract Study question Is intrauterine pregnancy rate following a FET timed by the rsERT with hour precision higher than the conventional FET among patients with recurrent implantation failure? Summary answer Personalized embryo transfer (pET) guided by this newly optimized rsERT significantly improved pregnancy outcomes of patients with recurrent implantation failure (RIF). What is known already Asynchrony between blastocyst and endometrium is responsible for implantation failures. Hence, accurate prediction for endometrial window of implantation (WOI) would maximize the effectiveness of assisted reproductive technology (ART). Previously, we have established a predictive model for endometrial WOI (rsERT) by three-time points sampling from the same patient at 48-hour intervals during one menstrual cycle. To avoid multiple sampling and build a more efficient WOI prediction tool, estimation method with hour precision by single time point sampling is urgently needed. Study design, size, duration This was a two-phase strategy involving model optimization and benefit evaluation with a prospective controlled trial. The study was performed in a tertiary hospital setting between September 2018 and December 2021. In the first phase (September 2018-June 2020), 91 participants were recruited to optimize the rsERT. In the second phase (June 2020-December 2021), 176 patients with RIF were enrolled to valida the clinical efficacy of this newly optimized rsERT. Participants/materials, setting, methods First phase: endometrium samples were obtained from sampling on LH + 5, +7, +9 or P + 3, +5, +7 respectively. Samples were predicted by the rsERT. Only those in agreement with theoretical WOI timing and from whom successfully obtained intrauterine clinical pregnancy after subsequent FET were eligible. Second phase: 88 patients in the experimental group underwent pET timed by rsERT after sampling at LH + 7 or P + 5. 88 patients in the control group performed standard embryo transfer without sampling. Main results and the role of chance This newly optimized predictive model could provide hour-based results of endometrial WOI and have an average accuracy of 94.51% by using 10-fold cross validation. In the second phase, a total of 88 NGS were constructed for RNA-sequencing by using endometrial biopsy samples from RIF patients in the experimental group (n = 88). The results indicated WOI displacement in 40 of 88 (45.45%). Among them, advanced WOI occurred in two patients (2/40, 5.00%), and delayed WOI occurred in 38 patients (38/40, 95.00%). The baseline clinical parameters, including age, BMI, infertility duration, types of infertility, number of previous failed cycles, main aetiology of infertility, the prevalence of pre-implantation genetic screening/diagnosis (PGS/PGD), endometrial thickness, endometrial types, the percentage of transferred blastocysts, good-quality embryo transferred rate and number of transferred embryos were all comparable between the two groups (P > 0.05). The intrauterine pregnancy rate (IPR) and implantation rate (IR) of the experimental group were significantly improved compared to those of the control group (for IPR: 61.36% vs. 31.82%, RR 3.630, 95%CI 1.958-6.732, P = 0.000; for IR: 42.86% vs. 24.66%, RR 2.292, 95%CI 1.392-3.772, P = 0.001). Moreover, in the experimental group, pregnancy outcomes did not exhibit remarkable differences between the natural cycle and hormone replacement (HRT) cycle. Limitations, reasons for caution Caution is warranted due to the non-randomized design and limited sample size of the study. Besides, embryonic factors should be taken into consideration as well. Hence, a multi-center randomized controlled trial of rsERT combined with PGT is needed to further validate its clinical efficacy. Wider implications of the findings These data add a new alternative of predictive model for endometrial WOI among Chinese women. Trial registration number the National Key Research and Development Program of China