A few days ago, I announced that Economy of Scale was dead. Those who took the time to read the comments may have noted that Eric Husman disagreed. That is worth paying attention to because Eric is a very smart guy. He blows things up in the desert by day and swills ale while thinking very deep thoughts by night, all the while keeping up with his truly significant other, Kathleen Fasanella, which is no small feat. Eric does not let me get away with much, which is good because I make most of this stuff up as I go along and Superfactory readers need to have someone like Eric keeping me honest.
For the record, after reading and learning from Eric’s comments, I can assert that Economy of Scale is as healthy as ever at a macro level, and is only as dead or alive at the micro level as management chooses it to be. In my previous post, I stated that, "When set up time = 0, it does not matter whether the company makes 10,000 of 1 product; or 1 each of 10,000 products." That is very true. However, whether it is 10,000 of one product or one each of 10,000 products, the plant must make 10,000 of something to produce economically. The rent on the building is fixed and must be leveraged over as much volume as the building can support if it is to be minimized. So far as Economy of Scale is applied to facility capacity utilization, it is as valid as ever.
Where Economy of Scale has died in some companies is in its micro application – EOQ theory. Eric described the math of Economy of Scale with some gibberish that only an ale swilling pyromaniac engineer can follow – something along the lines of "C(X1+X2+…+Xn) < C(X1) + C(X2) + … + C(Xn), where C is cost, Xi is the ith part up to n of them, and the left side is making many of them in some unified way (mass or lean production), while the right side is making each independently." Doing my best to keep up, I think this is true if the cost on the right side of the equation is the cost of making each item in a separate factory. If they are all made in the same factory, then I believe the math proves the irrelevance (not the death) of Economy of Scale at the machine level.
More important, I believe that there is an extraordinarily important management lesson in this. EOQ is nothing more than Economy of Scale applied to a particular item on a particular machine. It is math that seeks to find the balance point at which the inventory cost of making more than you need equals the savings you get from running more than you need. Instead of the Economy of Scale at the macro level of a plant, it is Economy of Scale at a micro level – the machine.
Traditionally, Western management thinking has been that the cost of making too many and carrying them in inventory is little more than the cost of money. The cash tied up in inventory could be in some investment somewhere making 5% interest, so the right side of the equation was only 5% of the cost of the excess production. With such minimal costs associated with over-production, the benefit of spreading the cost of the setup over large batches always carried the day.
What killed it was not math or logic, however, nor was it a new theory of economics. It was killed by a unilateral, arbitrary decision that became a self-fulfilling prophecy. Either Kiichiro or Eiji Toyoda, or both, simply decided that the correct answer was always going to be the opposite of conventional thinking, and that making exactly what was needed was always the right answer to the EOQ arithmetic. Then they tasked Taichi Ohno with the responsibility for making that decision turn out to be the economically correct one.
That unilateral, arbitrary decision was most certainly driven by a severe cash shortage, rather than a ‘Eureka!’ moment of economic thought. Producing a great big batch is not an option if you don’t have the money to pay for the batch. Regardless, they simply decided that EOQ theory was not valid, then worked like crazy to make that decision economically correct. Shingo contributed the principles of SMED, went a long way toward making the Toyoda boys look like economic geniuses.
Eric points out that it is SMED, not ZMED (Zero Minute Exchange of Dies), which is true, but irrelevant. The setup time does not have to be zero to effectively toss lot sizing out the window. Set up time just has to be some low number. At the same time Toyota was driving setup times down, they were also proving that the cost of inventory is a whole lot more than the cost of money. It includes handling costs, floor space, and much of the cost of poor quality. Producing too many drives a lot of excess cost, so the setup time does not have to be zero to make producing one at a time, or at least exactly to demand a good bargain. To use Eric’s equation, the numbers to the right of the ">" include a lot of inventory driven costs that are not on the left side, so the setup cost added to the left side has to be pretty high to make batch production a good idea. As Toyota drove the setup cost down on the left, and simultaneously proved that the costs of inventory on the right are pretty high, they beat EOQ theory to oblivion.
The proponents of micro economy of scale – the EOQ theorists – proposed that making many of something at a time was the only way to leverage high fixed costs. Then they put in place factories twice as big as they needed to be to store all of that inventory, material handling departments, MRP systems and quality inspectors – all fixed costs – that helped to make their theory come true.
Toyota defied the theory, then drove the setup times and the fixed costs driven by inventory down – and made their theory come true.
Here is the point: Eric once told me that a weakness in some of the economists thinking is that they do not usually take management into account. They bandy terms like "cost" around as though it is a pure figure determined by natural circumstances and it is the same among all producers in a country or a region. In fact, as just about everyone pursuing lean manufacturing knows, the ‘cost’ of a car is one number if you are in Detroit, while the ‘cost’ of a car is something a whole lot lower if you are in Georgetown, Kentucky. Even in the same plant, a traditional accountant would say the cost of a car is one number, while a lean accountant would come up with a completely different number.
Management determines what the cost of everything is both by their manufacturing practices and how they do their accounting; and they can generate costs that make batch manufacturing appear to be economical, or single piece manufacturing economical. The laws and principles of economics are not like the laws and principles of physics. Physics is real, while economics is an idea – a will o’ the wisp. The relevance of Economy of Scale of production, or lack of relevance, is more a function of management than economic principles.
The same is true of offshore outsourcing. Management can run manufacturing and calculate costs in such a manner as to prove that making things in China is more economical, like the folks at at Furniture Brands did, or they can manage the factories and calculate costs in such a manner as to prove that keeping manufacturing of the same products in the U.S. is more economical, like the management of Stanley Furniture. Which do the economists say is right? Both, which pretty much proves the irrelevance of economic theory to manufacturing management.
So is Economy of Scale dead in manufacturing? Where management says it’s dead, it’s dead; where management says it’s alive, it’s alive. Lean manufacturing is largely a matter of management deciding as arbitrarily as the Toyoda boys did, that Economy of Scale – at least the micro version of it known as EOQ – is dead, then driving the company to do whatever it takes to kill it and keep them from becoming liars.
In our previous musings on this topic, both Eric and I cited tooling costs as relevant to the validity of Scale or EOQ. Upon further pondering, I would suggest that it has little or no bearing on the issue. Tools usually have a life of X number of cuts or strokes. It really doesn’t matter if those cuts or strokes come one at a time, or the tool is left in the machine and cuts quite a few at a time – the cost of the tool is $X/Y pieces. In other words, if the tool life is 250,000 strokes, then the life of the tool is not lengthened or shortened by having it stroke 500 a day for two years, or 2500 every Thursday for two years, or 10,000 on the second Tuesday of each month for two years.
What determines the economy of the tooling is not the lot size but the sales volume relative to the tool cost. If the company makes and sells 250,000, they economized on the tool. If they made or sold less, they lost out on the tool investment. The production pattern, however, has little bearing on the math.
If the volume needed to economize on all of the tooling exceeds plant capacity – for example they have 10 tools for 10 products and they need to make 250,000 of each product to pay for the tooling over the next two years, after which the products will be obsolete – then their total capacity had better be 2.5 million over the next two years or they obviously will not make it.
Conversely, if the company decides to offer fewer products because there will not be enough capacity for an broader range of product offerings to pay for all the tooling the broad range requires, it is a stretch to relate it to economy of scale. It is a cost/price decision wrapped up more in marketing strategy than in economic theory.
Final note – Eric, if you would tell me the brand of ale you drink, I could down a pint or two before I write and perhaps save both of us the effort and confusion of me having to write multiple times on the same topic to get things right.
Eric H says
In order of preference:
1) Homebrew, usually IPA, sometimes cream or imperial stout, but I rarely get around to making it
2) High Desert Brewing Company IPA
3) Sierra Nevada Pale Ale
4) Sam Adams Boston Ale
Something else I would add to this: I said that I think Lean manufacturers are better able to take advantage of economy of scale than Sloanists. I believe this for two reasons: first, the relevant discussion of batch sizing that I think you have adequately covered. The Wiki entry on Just In Time has some relevant math on optimum EOQ for the special case in which the cost of ordering a unit (exchange time) is small.
Second, a lean manufacturer should be able to see increasing gains to scale compared even to mass manufacturing because for them,
C(Xn) < C(X(n-1)) << C(X1) in any facility that practices continuous improvement. That means that the cost of producing the nth widget is less than the cost of producing the (n-1)th widget, and both of these are far less than the cost of the first widget. It is not true of mass manufacturers, where the design engineers hand it over to the production engineers, and they finalize the production approach before the first widget (X1) even rolls out the door. If they have done their jobs correctly (in their eyes, not mine), then C(Xn) = C(X1) If the lean manufacturer also gets the improvement from scale (amortization of design and other costs over all units), they realize far greater gains than even the mass manufacturer with scale economies alone. Furthermore, the lean manufacturer applies lessons learned to design and other fixed costs (for example, better organization means the same product can be manufactured in a smaller factory, engineering is done faster and with fewer startup problems, etc.), so he sees more advantage the longer he is in business. The mass producer can't figure this out, so they keep expanding factory size, design and production engineering staffs, and specialized machines and see less advantage the longer they are in business. While we have been talking about JIT and SMED as methods to reduce production costs, I also understood Ohno to have introduced those as a means to force improvements in quality. That has benefits on both the cost and value sides of the equation. The quality improvements resulted in a reputation for low lifecycle costs, for which consumers are willing to pay a premium. The whole approach seems to sell itself, doesn't it Bill? ----------------------- A point I would like to clarify: it seems all too prevalent among economists (especially amateur ones like me) to view the firm as a production function. Raw materials in, manufactured product out, and whoever has the highest profit must have the best organization. They have no excuse for doing so in the post-Oliver Williamson era (he's a prolific writer who has been influential since at least the early 70s). I have been reading his _Economic Institutions of Capitalism_ lately. His approach is to view firms as governance mechanisms rather than production functions, and differential organization is a prime area of consideration. Fascinating reading - it even features a short review of Toyota's subcontracting practices, but it was written way back in 1983, when relatively little was known or appreciated about Toyota's methods.