An interesting pattern often emerges when you graph the coefficient of variation (Cv) vs. volume of a process time series.
Often, but not always, you'll see a distribution where there are:
- A few high volume products that will have low demand or process volume variation as measured by having a low coefficient of variation (standard deviation/mean),
- and many other products that have erratic demand or activity.
- Occasionally you may see a few high volume and high Cv products (large one-time bulk orders for example).
The math is straight forward - gather up orders, shipments, consumption, receipts or other transactions of interest and calculate the time series average, standard deviation, and Cv. Then plot the average vs. the Cv.
This 'demand segmentation' is a fundamental analysis component of many process capacity and material handling design efforts. Traditional ABC-Pareto 80/20 can be misleading when only volume is considered. Typically, high Cv items don't make good candidates for kanban or point of use replenishment. Low volume 'steady eddy' C items can be very predictable and make perfect candidates to take off of MRP reordering.
