In order to optimize the utilization of influent carbon source, a feeding pattern so-called "coefficient of influent flow rate" was adopted in a pilot-scale step-feed A/O process treating domestic wastewater. The effects of influent flow distribution ratio on nitrogen removal efficiency were investigated when the reactor was operated at different loading rates and COD/TKN ratio of 3, 5, 7, 9, 11, 13, respectively. The experiment results indicated that the wasting of nitrification capacity occurred when high loading rate and low C/N ratio (COD/TKN <5) were applied, while ammonia removal efficiency was decreased obviously at the same time. When the system was operated at high loading rate and high C/N ratios, the insufficient nitrification occurred in first stage, which resulted in the absence of electron accepter in the downstream anoxic zone. Consequently, the total nitrogen (TN) removal efficiency was decreased even though the COD/TKN was higher than 9. However, due to the unlimited nitrification, the increased C/N ratio led to an enhancement of TN removal efficiency when the system was operated at low influent loading rate. When the influent COD/TKN was kept at 13 around, relatively low effluent TN concentration less than 2 mg/L, and a highest TN removal efficiency of 97.6% were achieved, respectively. For the feeding pattern selected for the study, the conclusions obtained from the experiment results showed that the influent flow coefficient method could use carbon source sufficiently and decrease the influent flow rate of last stage when the wastewater with higher C/N(C/N > alpha) was fed. However, a completely nitrification should be promised in each stage during this period. When the wastewater with low C/N ratio (C/N < alpha), due to the limited carbon source, C/N is the key parameter for TN removal efficiency. From the point view to favor the growth of nitrifiers and satisfy the ammonia effluent standard, the balanced loading of nitrifiers in each stage strategy maybe substitute the feeding pattern proposed in this study as an optimal choice.