Cyclic gas injection (CGI), also known as huff-n-puff process, has been demonstrated as the most effective and promising IOR solution in shale reservoirs. Such process involves many operating parameters that can affect the recovery performance in different degrees. As soaking time and pressure depletion rate (PDR) are the two crucial factors, this study aims to further investigate their roles in the oil recovery process of CGI experimentally and numerically. A total of ten series of the N2 huff-n-puff tests were performed on oil-saturated Eagle Ford shale plugs in a matrix-fracture system. With the injection pressure (Pin) of 1,000 psi, two groups of tests were conducted: 1) under a constant PDR in blowout and five soaking times changed from 0.25-hr to 48-hr; 2) under a constant soaking time of 12-hr and four pressure depletion times changed from 0.05-hr to 48-hr. With the Pin of 5,000 psi, the CGI tests were performed under 12-hr soaking time and blowout pressure depletion to study the effect of Pin on recovery efficiency. Lab-scale simulation models were built and tuned to history match the experimental data. The experimental results show that recovery factor (RF) from a single cycle increases with soaking time within a certain range, and a longer time has no effect on improving oil recovery. For 1,000 psi Pin, during the soaking phase, the system pressure declined rapidly at the first 3 hr and then became stabilized. The pressure drop was relatively low (10 psi) in the first cycle compared with the following cycles. It tended to increase with the number of cycles performed, which was caused by the decrease of oil saturation thus more gas can be injected into the core sample. It reveals that a soaking period is necessary to the oil recovery process. On the other hand, increasing the pressure depletion rate can raise the incremental RF from each cycle. The blowout condition showed the highest cumulative RF. The performance of N2 huff-n-puff with 5,000 psi Pin outperforms the case of 1,000 psi, which can increase the ultimate RF with less cycles. The well-tuned simulation models were used to analyze and optimize the CGI recovery process.