A next-generation diagnostic tool for bacterial vaginosis, consisting of quantitative and/or qualitative molecular criteria, has not yet been identified. The optimal diagnostic tool should not only diagnose bacterial vaginosis in diverse populations, but should also detect early signs of transition to dysbiosis. We evaluated a tool based on log10-transformed qPCR data for Lactobacillus crispatus, Lactobacillus iners, Lactobacillus jensenii, Lactobacillus gasseri, Lactobacillus vaginalis, Lactobacillus genus, Atopobium vaginae and Gardnerella vaginalis in vaginal specimens of 426 African women to detect dysbiosis and predict transition to dysbiosis. G. vaginalis (p = 0.204) and A. vaginae (p = 0.001) were more commonly present in women who evolved to an intermediate (Nugent 4-6) or bacterial vaginosis score (Nugent 7-10) compared to women who continued to have a normal Nugent score. The combination of G. vaginalis, A. vaginae and Lactobacillus genus counts performed best for diagnostic accuracy for bacterial vaginosis--sensitivity 93.4% and specificity 83.6%; and for predictive accuracy for bacterial vaginosis--sensitivity 79% and specificity 52%. L. crispatus combinations did not perform well. We conclude that a triple-G. vaginalis-A. vaginae-Lactobacillus genus-qPCR tool holds promise for research in sub-Saharan Africa or when developed as a next-generation clinical diagnostic modality for bacterial vaginosis, ideally engineered as a rapid assay.