I recently read an article that drove me crazy.
In discussing “work-in-process” it talked about “pull systems” and how they are essential to lean success. It also discussed that pull systems are instrumental for producing only what the customer needs and when they need it … i.e., Just-In-Time.
So far, so good.
The article went on to say that the way to implement a pull system is to minimize work-in-process inventory. It then gave an example of how reducing production lot sizes would achieve the desired results. The implication was that you could reduce lot sizes and operate with the same production processes that currently exist to achieve JIT.
That’s a big problem—and one we learned the hard way 35 years ago at the Wiremold Co. In the early days of experimenting with JIT, we did what that article recommended. We reduced production lot sizes. And inventory did go down—slightly—from 3.4 turns to 3.6 turns. But so did our customer service level—from 98% on-time delivery to less than 50%.
But we persisted. And within 24 months, profit went from record highs to just above break-even. It was at that low point that Art Byrne joined Wiremold as our CEO and taught us how to properly implement JIT (it wasn’t called lean until several years later).
Manufacturing companies normally maintain three types of inventory: raw materials, work-in-process and finished goods. If you look at the flow of products, it looks like this:
What is at the center of that flow? The Factory. The Engine. That’s where value is created and profit is made. In a batch-production world, the engine is broken because there is so much non-value-adding activity taking place.
When Art first toured our main factory before joining us, he told me “There are twice as many people as you need.” Everyone was “working” so my “non-lean eyes” didn’t see the waste that he saw.
Wrestling with MRP
Our JIT efforts at Wiremold failed because we didn’t know how our MRP system really worked to provide production and inventories based on the batch world that we lived in.
Inventory levels and production-batch sizes are managed by a company’s MRP system (or the MRP module of their ERP system). Raw-material inventory batch sizes are influenced by the company’s purchasing department. If vendors offer discounts for larger quantities, purchasing agents will generally take advantage of those terms. They see their mission as getting the lowest price—period.
Work-in-process batch sizes are determined by an “economic order quantity” (EOQ) calculation that is imbedded deep within the MRP system. EOQ is expressed as a formula that tries to balance the cost of acquiring products with the cost of holding them. A major component of the cost of acquisition is set-up time—the time it takes to set up a machine when switching from product A to product B. The cost of holding inventory is comprised of the cost of damage, obsolescence, acquiring and maintaining space, insurance, the company’s “cost of capital” (the financial cost), etc. The financial cost element is a complex calculation that blends the company’s cost of borrowed capital (i.e., debt) with the cost of its equity capital (which only people on Wall Street claim to understand). Naturally, the cost of each of those factors is different for every company.
When software vendors develop MRP systems, it is impossible for them to know the “holding cost” for every company that might buy its system. But without this number, EOQ cannot be calculated, and the system won’t work. So, they insert a “place holder” with the intent that each customer will replace it with their own number. In my experience, the place holder most often used is 24% per year (i.e. 2% per month). Unfortunately, the vendors don’t tell the customer this. In the past 30+ years I have asked many accountants and production planners if:
- They realize that their MRP system is dependent on an EOQ calculation,
- They realize the factors that go into it
- If the answer to 1 and 2 are “yes,” did they replace the “place holder” with their company’s costs.
I have never received a “yes” answer to this last question.
As I explained this to the chief financial officer of a manufacturer of fashion goods, when I mentioned the 24% “cost of holding” factor, he was stunned. Because of the nature of their products, their cost of obsolescence alone was about 50%. He immediately understood the value of reducing set-up time, which would result in smaller batches and reduced obsolescence cost.
Finished goods inventory is very dependent upon the type of products and customers a company has. One of the Wiremold companies sold exclusively to OEMs. Its products were made-to-order and shipped as soon as they were made. Another company sold to electrical distributors who ordered from a catalogue and expected their orders to be shipped immediately. That company had to rely on carrying finished goods inventory of everything we sold “just-in-case” it was requested.
To determine how much to make of any product, MRP systems need to know what the future demand for those products might be. Which brings us to the subject of “forecasting.” If you are a strict “make to order” company, you don’t have to guess how much to make at any given time—although you may need to forecast demand for long-lead-time raw materials or adding capacity. However, if you are not “make to order,” then your MRP system looks to its forecast module. This module uses sales history to project future demand by taking into account past sales quantities, trends, seasonality, etc. It also gives marketing personnel the ability to have manual input for certain planned events, such as sales promotions.
All forecasts are wrong. We just don’t know by how much and in what direction. So, to compensate for forecast error, plus potential operational problems, the system calculates a “safety stock” level for each item. The more volatile the historical demand, the more safety stock you need. Another factor that drives what safety stock should be is the level of on-time delivery you want to achieve. Our system was set at 98%, so we had a lot of inventory.
I once did an experiment to see how sensitive the calculation was to the on-time delivery factor. We set up a “test company” in our system and copied our real data into it. I then changed the on-time delivery factor to 98.1% and let the system calculate the planned inventory. Then again for 98.2%, and 98.3%, all the way up to 100%. With each one-tenth of a percent increase, the amount of required inventory went up geometrically. And it was all in additional “safety stock” because the system tries to mitigate problems of forecast error and operational inconsistency.
So, if you want to never fail in fulfilling a customer order (i.e. 100% customer service) you need a lot of safety stock. We would have had to build lots of new warehouse space to store it all.
That exercise clearly demonstrated the benefit of lean’s pull system “sell one, make one” principle.
The Engine: Flow, Pull, Takt Time
At Wiremold, our ignorance of how to achieve JIT put us in big trouble with our customers.
Production batch sizes were what they were because of the time to set up machines. High set-up times justified big batches based on our cost of holding inventory.
After Art became our CEO, I learned that if we wanted to service our customers “just-in-time,” we needed to fix the engine. Where to start? Since set-up time drives big batches, we started there. For the first kaizen event that I was on, we had five days to reduce the set-up time on a punch press from 90 minutes to 10 minutes. Long-story-short, we achieved a 5-minute and 5-second result. And spent only $100 in capital to achieve it. We ultimately reduced the set-up times for every machine that we had by 90% or more.
Some say that lean obsoletes EOQ. I say it validates the math of EOQ. The smaller the set-up time, the smaller the “batch” that the EOQ formula will calculate. But, eventually the batch size becomes irrelevant as you move to a pull system, using kanban, and produce based on real, not forecasted, demand.
As we worked on reducing machine set-up times, we concurrently worked on creating flow. The first thing that we did to facilitate this was to eliminate the functional departments (e.g. press, rolling mills, paint, assembly, etc.) and reorganize the factory by “product families” (called value streams in today’s lean lexicon). We then established flow lines within those families by moving machines and people closer together. Every machine was eventually moved—sometimes more than once. We had many machines that hadn’t been moved since the day they were installed decades ago.
Since the MRP production schedule is based on a forecast, workers are disconnected from what the customer is currently buying. When they make something, they assume it’s because the customer ordered it. They don’t realize that once it goes into a WIP storeroom it can sit there sleeping for a long time. To dispel that belief, we immediately dismantled those storerooms, which occupied an entire building, and redistributed the
WIP back to the plant. If it was a manufactured part, it went back to those that made it. If it was a purchased part, it went to the people that bought it. We had areas in the factory where the inventory was stacked so high you could not see the machines or the people running them. The shock of that had the desired effect. We then shut off MRP’s calculation of batch production orders and gave the factory orders, using Kanban cards, based on actual sales, which were initially fulfilled using the WIP inventory that already existed. Only when that inventory was depleted did they make something with the cell set up to operate at takt time.
What did we do about Raw Materials and Finished Goods? At the beginning: LEAVE THEM ALONE! Fixing the factory is going to create chaos and we needed to protect the customer from that. As flow and pull were established, we required less safety stock and then began to reduce the front and back end of the material flow. Ultimately, we worked with vendors to establish pull systems with them. And we worked with customers to teach them to buy from us on a pull system. By spreading JIT knowledge throughout the value chain, our partners were able to reap the benefits of Lean also. Note: It’s easier to accomplish this with vendors than customers.
JIT is only one element of implementing a comprehensive Lean Strategy. It involved much more than is described above: improving machine and machine tool maintenance (TPM), implementing kanban and heijunka, 5S, reducing the number of unique job descriptions (in an union factory) to create a flexible workforce, establishing meaningful KPI’s that are maintained within the product families, creating an effective suggestion system, revising our hiring process, changing our marketing approach to sell value instead of products, changing sales terms to discourage batch buying, improving product development to reduce the time from concept to launch from years to months, eliminating standard cost accounting and changing financial statements to the “Plain English P&L,” and other changes, that are beyond the scope of this article. (For more information about eliminating standard cost accounting and changing financial statements to the “Plain English P&L”, see the book “Real Numbers: Management Accounting in a Lean Organization,” by Orest J Fiume and Jean Cunningham.)
As a result of fixing our engine we improved inventory turns at our distributor-oriented business from 3.4 to 18. In addition, we improved the turns at our OEM-oriented company to more than 30. The benefits: we re-established excellent customer service and freed up 50% of our production space due to quick set-up and flow lines resulting in less inventory. In addition, the amount of cash generated from less inventory was redeployed and used to buy our first five acquisitions. Thus, we traded one asset, that cost us money to hold, for others that contributed additional profit.
So, don’t fall for the advice that you can achieve JIT by just reducing batch sizes. It’s not that easy and will get you into a world of trouble. The right way to achieve it is to first fix your engine and everything else follows that.
Orest "Orry" Fiume is the retired CFO of the Wiremold Co.
For more information about the many other changes made during Wiremold’s Lean transformation, see “Better Thinking, Better Results,” by Bob Emiliani, et al.
