Lot streaming is the process of splitting a given lot or job to allow the overlapping of successive operations in multi-stage production systems, thereby reducing the makespan of the corresponding schedule. Most of the researchers in this area concentrated on consistent sublot models. However, the makespan may increase under the constraint of consistent sublots. Recently, variable sublot models have received a lot of attentions. In this research, we develop a heuristic variable sublot procedure (model). This procedure determines machines, on which transfer sublot sizes are reallocated, and transfer sublot sizes for each machine according to machine processing times in the flow shop. The proposed procedure is tested and evaluated via simulation. Several factors, such as number of sublots, number of machine, processing time, and the ratio of (number of dominant machines / number of machines), are considered in different lot streaming models (equal sublots, consistent sublots, and variable sublots). The results show the variable sublot model outperforms consistent sublot and equal sublot models. When the number of sublots, number of machines, and processing time increase, corresponding percentage of improvement for different lot streaming models increases. In addition, the variable sublot model is more sensitive to the ratio factor than consistent and equal sublot models.