Lean purists insist that it’s possible to manage any kind of inventory with some form of a pull system, which uses consumption-based triggers for replenishment. Alternatively, the material requirements planning and sales and operations planning folks typically prefer the zero-inventory benefits offered by a more traditional method involving replenishment that is triggered by planned orders only. As is so often the case, the truth lies somewhere in between: Certain operations do lend themselves to pull, while others are better managed with a push approach.
The challenge then becomes identifying what key elements of an item should be used to determine the best method for its replenishment. This evaluation is important because the infrastructure, skills required, and cost per order made for a push approach are much higher than for pull-based methods. Also, using pull-based replenishment is more efficient from an activity-based costing perspective.
The following case study illustrates the challenges inherent to subscribing too closely to either a push or pull mentality as a one-size-fits-all solution. It also shows that you don’t have to have a powerful enterprise resources planning (ERP) system to be successful.
A growing fulfilment and repair operations company supports many major providers of wired and wireless health-monitoring technologies. The firm’s business model is based on fulfilling orders from its medical-service-provider partners for devices and then serving reverse supply chain activities including replacement, repair, and recycling. Units are received, assessed, repaired, and put back into inventory to be resold, if practical; the rest are recycled. This practice of return-refurbish-resell helps the company avoid sending waste to landfills and having to push equipment to recycling facilities to recover value from the component parts.
Unfortunately, this business was facing a number of problems. First, it had too much inventory of the larger, high-dollar, electro-mechanical devices. Available warehouse space was far exceeded, and the organization even needed to use offices for storage. At the same time, shortages were reported periodically, which caused major headaches. Each type of returned device had a different rate of recovery to be put back into the supply chain. In addition, there was variation in the flow of returns for refurbishment, making planning for repair parts problematic.
Company leaders decided to have the firm’s information technology professionals develop an application to monitor usage for items historically shipped from week to week and summarize usage in the monthly buckets that they used for a biweekly planning cycle. Every other week, they would take a full physical inventory and then determine what orders to place, attempting to account for changes in future demand both upward and downward. Everything was managed with a push ordering method.
The main objectives were to establish a better method than just having what’s needed from week to week for the larger units and to put an end to surprise shortages of repair parts and packaging materials without accepting mountains of inventory. As mentioned previously, the firm had no ERP system, bills of material, or back-flushing capability. Instead, it relied on simple tools for inventory management, shipments, and packing lists. Repair parts usage was not tracked per repair, only when units were making their way back into inventory.
The push-pull dilemma
Employees put together a history of monthly usage and, with the help of sales data to understand the demand-side issues, forecast new launches for customers. It was determined that the top 20 units actually made up about 90 percent of spending and 90 percent of the space needed for storage. Next, a spreadsheet was created with weekly buckets to outline the findings. The company then was able to run a simple accumulated requirement over the horizon to plan for total demand throughout 16 weeks.
Projected supply was more complicated. In addition to reviewing on-hand inventory and planned purchase order receipts, employees had to account for units being returned from the field, repaired, and put back into inventory. There was a history of units being returned by week, but it took some digging to figure out what percentage of them was successfully repaired and returned into inventory. Once that was worked out, it was possible to consider by week the projected refurbishments from returned units going back into inventory. Finally, the organization ran simple spreadsheet calculations in order to understand accumulated available inventory compared with accumulated demand over the 16-week planning horizon.
It was then possible to perform a complete review of the requirements and adjust open purchase orders every other week in line with physical inventory cycles. For each item, a safety stock level was determined so that planners could evaluate projected stockout dates and schedule deliveries to minimize on-hand inventory within those parameters. This worked splendidly; inventories plummeted and stockouts were eliminated.
For the remaining 150-odd items, it turned out that not enough information existed to consider a push model. So, instead, they would be ordered one for one so that they would match up with launch schedules for the units themselves.
Another spreadsheet would be used to capture typical purchases throughout the last six months. An average weekly usage was derived and recorded in one column. The correct pull signal amount to trigger replenishment was calculated next. Eventually, each item in the spreadsheet contained information about item identity, source, average lead time in days, calculated average daily use, and related measures. The spreadsheet also included information about uncertainty for buffering purposes. Next, trigger levels were confirmed. Team members also fashioned pull signal cards that included source information and a record of reorders. These brightly colored cards were placed in inventory between the containers as a pull signal.
Not long after all of this was accomplished, the organization was able to easily identify items to be kept in a central supermarket, with only a small supply on the floor at the point of use. Within six weeks, everything was under pull system control, except for the small number of items that were better managed with a push method. After three months, the company realized an average on-hand reduction of more than 50 percent for its expensive, large items. At the same time, surprise shortages dropped from several per week to less than one per month. Clearly, a blended model was a wise choice.
Ron Crabtree, CIRM, SCOR-P, MLSSBB, is chief executive officer of MetaOps, a master MetaExpert, and an organizational transformation architect. He is the author or coauthor of five books about operational excellence and the online magazine at MetaOpsMagazine.com. Crabtree also teaches, presents, and consults. He
may be contacted at firstname.lastname@example.org.
To comment on this article, send a message to email@example.com.