IHI.org - A resource from the Institute for Healthcare Improvement
Header Image






Optimize Inventory:
Use Pull Systems

In a "pull system" of production, work is only done at a particular step in the process if the next step in the process is demanding the work. Enough product is ordered or made to replenish what was just used. This is in contrast to most traditional "push systems" where work is done as long as inputs are available. A pull system is designed to match production quantities with a downstream need. Often this can result in lower inventories than a schedule-based production system. Pull systems are most beneficial in processes with short cycle times and high yields. Some features of effective pull systems are small lot sizes and container quantities, fast set-up times, and minimal rework and scrap.

The concepts, "Match Inventory to Predicted Demand" and "Use Pull Systems," seem to offer two conflicting approaches to optimizing inventory. Matching inventory to predicted demand is most appropriate when cycle times for production or assembly are large. In these situations, it is not acceptable to wait until an order is received to begin production. The use of pull systems is more appropriate when production delays are short. Often these two concepts are combined to develop an optimum inventory strategy.




Examples of Tests of this Change

The staffing of nurses in a hospital can be studied as an inventory problem. Each floor wants to maintain a staff of nurses available to meet routine and emergency needs 24 hours per day. Because of the variety of skills required, over-staffing is often the result. One hospital crossed-trained its nurses to be able to work in all areas of the hospital. They then created a hospital pool of nurses on each shift. Each floor would staff to meet routine needs based on patient census. Additional help for new patients and emergencies was obtained from the nursing pool. This pull-system approach resulted in a 15 percent reduction in nursing costs.


What others are saying
Post your comments about this item.
View All Comments
Post Your Comments