Stringent exhaust emission and fuel consumption regulations impose the need for new solutions for further development of internal combustion engines. With this in mind, a refined control of the combustion process in each cylinder can represent a useful and affordable way to limit cycle-to-cycle and cylinder-to-cylinder variation reducing CO2 emission. In this paper, a twin-cylinder turbocharged Port Fuel Injection–Spark Ignition engine is experimentally and numerically characterized under different operating conditions in order to investigate the influence of cycle-to-cycle variation and cylinder-to-cylinder variability on the combustion and performance. Significant differences in the combustion behavior between cylinders were found, mainly due to a non-uniform effective in-cylinder air/fuel (A/F) ratio. For each cylinder, the coefficients of variation (CoVs) of selected combustion parameters are used to quantify the cyclic dispersion. Experimental-derived CoV correlations representative of the engine behavior are developed, validated against the measurements in various speed/load points and then coupled to an advanced 1D model of the whole engine. The latter is employed to reproduce the experimental findings, taking into account the effects of cycle-to-cycle variation. Once validated, the whole model is applied to optimize single cylinder operation, mainly acting on the spark timing and fuel injection, with the aim to reduce the specific fuel consumption and cyclic dispersion.