The acceleration of urbanization has resulted in the increase of urban surface runoff. Bio-swale is a promising stormwater control measure that has been proven to be hydrologically effective on urban surface runoff. Column studies were conducted to determine the optimal bio-swale composition. Results demonstrated that water reduction was proportional to inflow decrease. Columns that planted border privet and Ophiopogon japonicus showed a larger water quantity reduction compared with that of planted boxwood and ryegrass, glossy privet and Chlorophytum comosum ‘Variegatum’ in vegetation tests, which was the same as the order of measured transpiration capacity of the plants. Water reduction rate increases dramatically with decreasing planting soil thickness. By contrast, no significant change occurs once the thickness of the artificial filler layer is altered. The bio-swale column with a high-infiltration rate artificial filler produced a good hydrological control effect. Sand was found to be the optimal media among the selected media compositions. Although the inclusion of an additional ponding depth affected total water reduction, it produced a stable outflow. SPSS software was used to assess the relationship between water reduction rate and its influence. On the one hand, water reduction rate increased linearly with increasing water inflow, soil thickness, and ponding depth. On the other, water reduction rate grew linearly with increasing plant factor and artificial filler infiltration rate. The multiple linear regression model revealing the relationship between the water reduction effect, and its influencing factors were obtained via the stepwise regression method in the SPSS software.