In July of 2008, I stepped out of an engineering leadership role and into an operational role. The transition was exciting and overwhelming. While I had been in and around manufacturing for a little more than a decade, I had never operated as a part of supply chain or production management. The closest I had really been to this side of the business was as a program manager, but even that role was predominantly engineering-based. My initial marching orders were simple: Turn the business around. For the month of July, our on-time delivery (OTD) was 64 percent, our scrap per earned hour (total dollars scrapped divided by total hours put to stock or earned hours) was $15, and our quality failures were hovering around 40,000 parts per million (PPM). While margin was not terrible, all other indicators were. If you did not catch the date above—July 2008—we were starting this initiative just before the great global recession.

Not really knowing where to start, my manufacturing engineering and Six Sigma training surfaced. After all, we are talking about manufacturing issues. Issues that I was trained to define, measure, analyze, improve, and control (DMAIC). I began interviewing and chasing known issues, looking for the “big problems.” I figured that a definition of the problem (or problems, in this case) would be a good first step. While there were many issues that floated to the top, a few things really stood out. First, we ran out of material all the time. Stock outage seemed to be the primary reason we were late on deliveries. We also scrapped a lot of material due to expiration of shelf life. Our technicians would go to pull material from stock just to find out it was expired. We would then miss the shipment and subsequently scrap the expired materials.

Define

My first step was to interview the production manager, Ron Kuegler, who had been with us for less than a year but already had the most widespread knowledge of our operations. I knew he would be key but I did not know he had already found all the bodies, so to speak. Through our conversation and interviews on the floor we learned about a change that our previous supply chain manager had made to our process—back flush accounting.

Back flushing is an accounting method for handling material transactions within a material requirements planning (MRP) system. More traditional cost accounting models require a transaction to issue raw material from a warehouse location to the production floor or work-in-process location. Then, after the product being manufactured is complete, another transaction is required to move the now finished assembly into a finished goods inventory location, typically back in the warehouse. This method allows a manufacturing company to keep tight controls of inventory (see figure 1).

Back flushing, however, is a streamlined approach where the raw material is never issued to the shop floor or work in procress even though the material is physically moved from the warehouse to production. Instead, the MRP system will continue to show the material still in raw material inventory. Then when the finished assembly is transferred to finished goods, the MRP system automatically pulls all materials on the manufacturing bill of materials out of raw materials inventory and issues them to finished goods inventory (see figure 2).

Back flushing has its place in manufacturing and can significantly reduce manual transaction. Specifically, this method works well in an environment where cycle times are very short. However, for this method to work, our inventory control and the material movement logic within the MRP system must be very robust. When back flushing was fully introduced in our system, the inventory accuracy was very low and the logic in MRP was not developed. 

Upon further investigation, we also found that cycle counting was not being performed. Inventory was received and delivered directly to the work cell, no matter how large the quantity. In addition, we had only 15 inventory locations in MRP—many millions of dollars and thousands of piece parts in one of 15 system locations. The primary warehouse location in our system was “floor.” A part could be on any shelf, or in most cases on a pallet on the floor, anywhere within the 2,500 square ft. warehouse. For example, if we were to look up part No. 1234 in MRP, it would tell us that eight units were available and in the warehouse at location “floor,” or maybe just “warehouse.” The problem being, we only know that part No. 1234 is somewhere in the 2,500 square ft. warehouse with thousands of other parts. While we are happy to know what part of the building to look in, it does not help much beyond that. Rather, we prefer that the MRP system inform us that the part is located in the warehouse in row AB on shelf four. With specific row and shelf locations, we can also easily count materials and record the accuracy. This problem was a primary reason why cycle counting was not conducted. 

Problem No. 1: The inventory management systems and controls were not properly developed. 

Measure 

Now we had a definition of a problem, but how bad was it, really? We all quickly agreed that a physical inventory was needed in order to measure the depth of our problem. What unfolded during the next month was even more interesting.

At about 8:30 p.m. on Saturday, in the middle of August, I hung up my cell phone from a depressing conversation with Kuegler. I had just learned the outcome of our physical inventory. At that moment we finally understood what we had been handed, an inflated balance sheet and true inventory accuracy of 40 percent. In one weekend, we identified a significant amount of inventory we did not know we had and even more that we could not find. It turns out our inventory was overstated by 6 percent (i.e., we thought we had $100 in inventory but we really only had $94).

What the overstatement meant is that, since our last inventory count, we had shipped finished products to customers and assumed that we shipped less material then we did. Meaning that we shipped a widget we believed cost $10 to manufacture and we sold it for $12, giving us a $2 gross profit, when it really cost us $10.60 ($10 × 106 percent = the 6 percent overstated inventory). Therefore, our actual gross profit was only $1.40. While this does not look that bad when referenced to $10 or even $100, the reality is that this error affected millions of dollars. For each million dollars of inventory, there was an error of $60,000. As terrible as all that sounds, there was some good news. Most of these errors in inventory had occurred between January and July of 2008. That is good news because the inflated balance sheet and profits had not carried from year to year. We could still correct the error within the year it occurred. 

Analyze

We knew what Problem No. 1 was and we knew how bad the problem was, but we still did not completely understand why the inventory was in such bad shape or what was contributing to the condition. To find out, we took a standard industrial engineering approach and mapped our processes. We learned right away that our processes were insufficient. There was no automation in the system at all, no designated locations for materials, all material movements were transacted manually and no standards for storage or location existed. The warehouse had a location called “floor” and that is where most items went, on a pallet on the floor. What we did have were good employees that were doing their best to keep the place organized and tidy, but we didn’t give them any of the tools required to do the job right. Even the shelving in the warehouse was insufficient for the job. The existing shelves were mismatched, damaged, and homemade.

What we also realized was that most of our inventory was not in the warehouse. Now it wasn't really an issue for the inventory to be physically located outside the warehouse, but it wasn't even controlled by the warehouse. For accounting purposes, we would simply expense raw material to the department as overhead. We would receive a shipment of 50,000 bushings and they were immediately issued to the floor. That might be a good practice under the right conditions, say, if you actually use 50,000 bushings a day, but we still have most of those bushings. 

Another problem emerged during the analysis: We had parts set up with date-driven order points. By this stage in the project, the recession was spooling up and we were still ordering parts based on date. That seems silly but we had no other triggers to order. We were not cycle counting our inventory and many of our materials were issued directly to the floor. As far as MRP was concerned, the material did not exist after a date in time. It was assumed to be consumed. 

Improve 

It was time to start improving the inventory system, but we had a problem. We were entering the improve phase of DMAIC and did not have a supply chain or warehouse manager. Our subject matter experts were consultants, both external and internal (other divisions of the company), but more important, there was no one on our team that was directly responsible for the inventory. Yet, as if by fate, the week we started the improve phase, the right candidate made his way to our conference room for an interview, Ron Allgood, a seasoned supply chain manager with extensive warehouse experience. We delayed our improve phase a bit and three weeks later we had a subject matter expert who would ultimately be responsible for controlling the improvements. I cannot stress enough how important this was. 

With Allgood on board and hitting on all cylinders, the improve phase went quick. We wrote all new procedures with incorporated process flowcharts. We designed and built a brand new warehouse with new shelving and racks, expanded space, and most important, very specific inventory locations. We even went as far as placing items in the warehouse by touch times, meaning that the more times per month the item was touched by a human, the closer it was placed to the front of the warehouse. Two more weeks and we had the warehouse up and running; however, it still took a couple months to get all the inventory cycle counted and moved into the new locations. The biggest part of this effort was moving inventory off the shop floor and into the warehouse (both physically and by using point-of-service warehouse locations).

We also came to terms with back flushing vs. traditional work-in-process accounting systems. We all agree that a streamlined back-flush process automated by our MRP system is an ideal setup but we have yet to work out the required logic and controls to successfully execute a pure back-flush method.

Instead, we have settled on a hybrid of the two methods. Inventory is physically moved to the production floor and remains in raw material inventory but we move the raw material to a new raw material location in the MRP system that identifies it as on the production floor; we save transactions by making this particular move as a batch process. We then back flush the majority of the materials from raw inventory to finished goods when the finished product moves to finished goods inventory. This is not our final state of operation (I doubt there will ever be a final state). The improvements to our system and warehouse continue.

A few months ago, we transitioned to using barcodes and hand-held scanners for all material movements. We, also, are working on adding labor collection to the same barcode system. Further, we are in the midst of transitioning from our more antiquated MRP system to an SAP system. With the transition to SAP, we will gain much greater opportunity for inventory management and maybe more back flushing with less (preferably none) manual transactions in the system.

Control

As we closed out this project in December 2008, we knew that our systems had to be utilized to control our inventory going forward. Cycle counting had to be resumed and standard reports for aged and obsolete inventory had to be strengthened.

Starting with cycle counting, we were fortunate to learn that our MRP system already included a cycle counting module that had not been utilized. After a few conversations with the software company we upgraded to the latest version and were off and running with cycle counting. With the introduction of barcodes and hand-held scanners, our accounting and supply chain departments quickly beefed up our reporting and set out to monitor the new system.

During the past year, we have tweaked our processes and procedures where necessary, but for the most part, it continues to operate in the same condition we achieved in December 2008, averaging 98-percent accuracy using the zero-variance cycle counting method. Stock outages are now unheard of and an entirely new method for controlling time sensitive (shelf life) material has been developed and is working beautifully, with no issues for more than a year.

However, the biggest outcome from this DMAIC project is the contribution that it has had on our on-time delivery. This is a topic for another article, but this project was the cornerstone in our 18-month journey from 60-percent to 98-percent on-time delivery. March of 2010 marks our sixth consecutive month of on-time delivery above 98 percent.