Snowball effect is a positive mass and energy feedback acting on a process. This effect complicates process control by activating variables interactions and nonlinear phenomena inherent to the process. Traditional process control structures are strongly affected by snowball, which means that to reject relatively small perturbations large changes in the manipulated variables are required. In several cases, operative limits are surpassed and one or more variables cause a process fault. Present work proposes a control structure that can operate more efficiently in process with a tendency to develop snowball effect. The control system has a collaborative control structure, which uses a hierarchical structure with two levels of action in accordance with a dynamic hierarchy. Original PID controllers of traditional control structure are left in plant-floor level but a collaborative level is put over it, acting as an optimizer of process actions. The control proposal is illustrated using a typical buffer tank-separation process, commonly found in process industries.