The global popularity of green roofs (GRs) rises as urban runoff becomes a primary environmental concern in both developed and developing countries. Although a growing number of studies have measured the runoff retention (RR) performance of GRs and investigated the underpinning factors, a systematic and quantitative understanding is lacking. This study applies a statistical approach on a dataset of 2375 original experimental samples associated with the RR performance of GRs observed across 21 countries, consolidated from 75 internationally peer-reviewed studies published in 2005–2020. The results show that the sampled RR rates (i.e., the proportion of rainfall retained on a per-event basis) range widely (0–100%), with an average of 62%. Rainfall intensity, substrate depth, GR surface coverage, climate type, vegetation type, and season type partially explain the variances in retention performance. Moreover, the effects of some factors (e.g., rainfall intensity) are not isolated but contingent on other factors (e.g., vegetative cover). This global synthesis shows few samples emanate from Africa, Central America, and Central Asia, highlighting the need of more GR research and applications in these regions. The average GR RR rate appears lower than some specified in green building standards, which implies the need to further improve the RR performance of GRs or combine GRs with other RR measures. Contingent effects of GR RR incluencing factors demonstrate the need to leverage design parameters and to account for local weather and climate characteristics in the optimization of GR performance.