Click to view animation illustrating the effect of batch size on throughput. This animation compares delivery performance of 3 different batch sizes (10, 5, 1) running though a sequential process.
At the top, our batch size is 10, work is handed off to the downstream stage when 10 items are completed. At first view this may appear more efficient, the backlog is quickly emptied, stuff is being worked on. But what about deliveries? The first delivery (release) is on day 31, while at the bottom (batch size 1) the final delivery completed on day 23.
A batch size of 1 is sometimes called “single piece flow,” and while it delivers in the shortest time it also attracts a greater coordination cost. In practice, optimum batch size is a trade off between cost of delay and transaction cost. Reinertsen points out that this is a U shaped optimisation meaning that extreme values will be furthest from optimum, and optimum value does not require precision.
The animation finishes with a Gantt and cumulative flow charts showing the behaviour over time of the 3 different scenarios. This animation assumes no defects in the process and everything is right first time through. But what happens when defects occur?