Inability in correctly predicting heavy rain events is primarily due to two reasons: lack of full understanding its physical mechanism and negligence of its predictability limit. How to deal with its predictability limit is the focus of this review paper, which is especially important to enhance the value of numerical weather prediction products to better serve end-users. Based mainly on authors' own or directly involved researches and experiences, many applications of ensemble methodology to heavy rain research and prediction are brielfy overviewed. Speciifcally speaking, the following four general areas are discussed: (1) ensemble prediction system including initial condition and model/physics perturbations, optimal ensemble size, model resolution, data assimilation, and various "virtual" ensembles; (2) forecast methods including ensemble anomaly forecasting, reforecasting analog ensemble, and storm track clustering; (3) forecast post-processing and calibration including ensemble mean, performance ranking and best member, weighted ensemble mean, probability-matched ensemble mean, and ensemble of dynamic factors; and (4) weather system analysis and model initial condition improvement including perturbation difference analysis, ensemble sensitivity, and targeted observation. It is expected that this review will inspire actions from both operation and research communities: many proven-to-be effective methods described in this paper could be adopted in routine weather forecasting practice by operational meteorologists to improve their forecast and service; research community could have a new starting point with new ideas and a clearer direction for future science and technology development including the improvement of current existing operational ensemble prediction systems in years to come.