"We would stand a better chance of success if we gutted a farm animal and read its entrails."

Tony Rizzo wrote this article in 1997 and you can find it here on John Caspari's site: http://casparija.home.comcast.net/~casparija/dweb/l118.htm.
I thought it would be worthwhile to reproduce this article here. It caused a lot of discussion way back then (in 1997) and maybe it will again. Tony is an engineer but he understands systems very well!
John Caspari ran simulations to test Tony's theory about farm animals and reading entrails. That will make an interesting post too.

THE FUNDAMENTALLY FLAWED THINKING BEHIND activity based costing

By: Tony Rizzo
tocguy@lucent.m
tocguy@home.com

(C) Tony Rizzo, 1997

I want to begin not by discussing activity based costing (please note that the very name of it is unworthy of capital letters) but by discussing a technical subject, Design Of Experiments (DOE). If your educational background is not in a technical field, and if you feel that this discussion is likely to be too technical and too difficult for you to understand, then you're probably correct. Stop reading now, and expose your organization to the debilitating effects of serious, random mistakes. However, if you want to really understand why every company that is managed according to the philosophy of activity based costing can never achieve performance levels beyond the mediocre, then please keep reading.

We begin with an introductory discussion of DOE.

The subject of DOE is of great interest to experimentalists, for a very practical reason. Most engineers, statisticians, and scientists who perform physical experiments are forced to work without the luxury of the mathematical models that theoreticians develop and use. This is so, not because the experimentalists are any less capable than are the theoreticians, but because the experimentalists frequently venture into the realm of the mathematically intractable. None of us is capable of developing fundamental, mathematical models of the problems typically tackled by experimentalists. For example, process development engineers often face having to optimize manufacturing processes for which no mathematical model can ever be developed from fundamental principles.

Consider the case of an engineer trying to develop a new paint formulation. If he/she tried to do so from fundamental principles, the poor sod would have to derive equations that predicted, say, the number of orange peel dimples in the paint, as a function of the concentration of solvent, the concentration of pigment, the temperature of the paint, the temperature of the room, the pressure setting of the sprayer, the temperature of the work, and the relative humidity of the environment. This is the sort of problem for which no one can derive a mathematical model from the fundamental laws of physics. Yet, our poor engineer still has a job to do. He/she still has to develop and optimize the paint formulation. This is where DOE comes in.

DOE was developed by mathematical statisticians for precisely the purpose of solving mathematically intractable problems. DOE is an experimental approach, not a theoretical one. As such, DOE must fulfill several important requirements, chief among which is the ability to help the experimentalist to extract a maximum amount of information from a minimum amount of experimental data. To achieve this clearly difficult objective, DOE offers several important concepts. These are, the response function, the main effect, and the interaction effect. Let's discuss the first of these, the response function.

The response function is probably the easiest of these four concepts to understand. It is the thing that is of real interest to us. To our paint-optimizing engineer, the response function might be the number of orange peel dimples formed by the paint, or it might be the tendency of the paint to crack or form bubbles. Whichever of the paint's many features the engineer is trying to optimize, that is the response function. As he/she strives to determine optimum settings for the many independent variables, such as chemical concentrations or pressure settings, the engineer determines success or failure relative to the response function. Think of the response function, therefore, as the goal of the optimization effort.

In most engineering efforts of practical value, the engineer is faced with optimizing not a single response function but several. For example, it would be of little value for our engineer to create a paint formulation that formed absolutely zero orange peel dimples, under all imaginable circumstances, if the price for such a paint formulation were a paint that cracked and peeled readily. Therefore, most practical optimization efforts must optimize not one but a combination of response functions.

In business, of course, we have response functions that interest us too. They are profitability, return on investment, expenses, cash flow, etc. For those of us who favor the surrogate measurements suggested by the Theory of Constraints (TOC), the response functions would be Throughput, Inventory (including investment), and Operating Expense: T, I, and OE (Throughput is a measure of the money that comes into a business and sticks with it. It is calculated as the price that we get for a sale, minus the fully variable cost of making the sale). Thus, our situation is very similar to that of our manufacturing engineer. We face having to optimize not a single response function but several. For the sake of this discussion, let's assume that we need to optimize the three TOC response functions: T, I, and OE.

We really have cause to pity our poor manufacturing engineer. He/she is faced with deciding a specific setting for each of very many independent variables. For example, should the concentration of solvent be set at 35 percent or should it be set at 47 percent? Should the sprayer's pressure setting be 30 psi, or should it be 80 psi? These are the questions that our engineer must answer with respect to the independent variables. Performing the optimization successfully means that the engineer is able to identify precisely the settings that cause the performance of the paint to be optimum, as measured by the full set of response functions.

In business, of course, our task is even more daunting. Not only do we have to deal with our own set of independent variables, we have to deal with many more independent variables. In comparison with most of the technical systems that engineers might try to optimize, most business systems present us with a relative infinity of independent variables. Business systems, also known as organizations, are highly complex systems. Yet, despite all the complexity, we face a very similar optimization problem. We need to determine optimum settings for the many independent organizational variables.

What are these independent organizational variables? Here are just a few: the capacity of the purchasing department, the capacity of the sales department, the capacity of the mechanical engineering group, the capacity of the electrical engineering group, the capacity of the manufacturing engineering group, the capacity of the software engineering group, the capacity of our manufacturing lines, the capacity of our distribution system, the capacity of our financial department, the capacity of our shipping department, the capacity of our receiving department, the capacity of our internal computer network, the capacity of.... In other words, our independent variables include (but are not limited to) all the functions that some among us call "cost drivers."

The task that we seem to face consists of having to find the optimum setting for each of these, i.e., the set of settings that results in the optimum performance of our organizational system, as measured by the combination of response functions, T, I, and OE. Clearly, our business optimizations are significantly more difficult than the optimization of any technical system. How, then, can we possibly make any measurable progress toward this lofty goal? Let's see how our manufacturing engineer goes about it. We may find our conclusions somewhat surprising.

Since our hypothetical engineer understands DOE, he/she is likely to try to identify the few variables that have a significant main effect on the response functions. What is a main effect? It's easy, really. The main effect of an independent variable, on a response function, equals the average change in the function, as the variable changes from its low setting to its high setting. Let's illustrate this with an analogy.

Imagine that you are standing on a hillside. You can walk in the North-South direction, or you can walk in the East-West direction. Each direction is an independent variable. The finite distance that you are able to walk in each direction is the range of the corresponding independent variable. The main effect is the average change in elevation, as you walk in each direction.

An interesting aspect of DOE, however, is that one does not change the independent variables one at a time, simply because one would have to perform too many experiments for most practical problems. Instead, our hypothetical manufacturing engineer would change all the variables at the same time, during the course of a manageable number of experiments. This is tantamount to your walking in a Northwest-Southeast direction and then allocating part of the change in elevation to movement in the North-South direction and part to movement in the East-West direction The allocations are the main effects of the independent variables.

There! You should be feeling more comfortable right now. I've used a word that is familiar to you. I've mentioned allocations. In fact, this is what activity based costing attempts to do. It attempts to allocate pieces of the overall costs of an organization to the many independent variables (activities) associated with a particular product. Again, in terms of our surrogate TOC response functions, activity based costing tries to allocate pieces of overall operating expenses (OE) and investment (I) to the activities required to make and sell a product.

The strategy behind activity based costing is to search for the "big cost drivers," i.e., those activities that are required for the manufacture and sale of a product and that have significant allocations of overall cost associated with them. These, supposedly, are the areas where management should focus, to achieve significant cost reductions. It sounds wonderful.

Part of this overall strategy is to identify those independent variables (activities) that at the time have a main effect of zero on the T response function. These, supposedly, are the variables whose settings we can alter without affecting the T response function. In other words, an important part of this strategy is to identify where we can, either, cut costs by cutting resources, or at least keep costs under control by shifting resources, without having an adverse impact on T. Activity based costing claims to be able to show us which independent variables have a zero (or at best a negligible) main effect on T, thus letting us reduce I and OE without risk of damaging the business. Consequently, with activity based costing we should be able to utilize all our resources fully and optimally at all times. Those resources for which we don't have work today.... Well, they represent an opportunity to save costs.

There is one fundamental flaw in the thinking behind this strategy. To understand the flaw, we first need to understand another DOE concept, the interaction effect. We'll need the elevation analogy again. But, this time let's use it in a more metaphorical manner, just for the fun of it.

Let's pretend that you are in a very hostile environment, in the middle of a blinding data-storm. Numbers are flying past you at hurricane speeds. Financial figures stab your eyes mercilessly, leaving you nearly blind. All that you can see are a single altitude gauge and a compass, which you hold in your hands.

The economic wind leaves you breathless, gasping for information about your environment, information that can guide you to safer, higher ground. If you need anything, you need higher ground. You need elevation. Fortunately, you remember your activity based costing strategy. So, you begin your assessment of your environment. You start by walking two hundred yards in a Northerly direction. After your brief walk, you look at your altimeter, and you observe that your elevation has increased by only ten yards. In other words, the main effect of your two-hundred-yard movement in a Northerly direction is only ten yards of elevation. You walk back to your original spot, somewhat disappointed.

But, you are determined to achieve a much higher elevation, somehow. So, you continue your assessment of your environment. From your original spot, you walk two hundred yards to the West. When you get there, you look at your altimeter again, and you observe that your elevation has increased by thirty yards. In other words, the main effect of your walk in a Westerly direction is thirty yards of elevation. This is good. Life is a little easier here.

Still, there is always room for improvement. If you could achieve just a bit more elevation, life would be even better. Then, you remember that earlier, when you walked toward the North, you gained ten yards of elevation. Why not walk to the North again, and gain a total of forty yards? Sure! So, you head North, confidently, only to find that when your little stroll is over you've lost fifty yards of elevation. This means that the main effect of your Northerly walk is now NEGATIVE FIFTY yards. What happened?

You've fallen victim to an interaction. The main effect of walking in a Northerly direction is ten yards of elevation, when you start walking from your original spot. But you began your recent Northerly walk not from your original spot but from a spot two hundred yards to the west of it. Your original spot was South and East of a crater. As you walked North from your original spot, you walked with the crater on your left, missing the crater. As you walked West from your original spot, you walked with the crater on your right, again, missing the crater. But, when you walked North from the second, more Westerly spot, you walked into the crater.

This is the nature of interactions. The main effect of one independent variable is, itself, changed by another independent variable. In this illustration, the main effect of walking in a Northerly direction is POSITIVE TEN yards of elevation, when the Northerly walk begins at the original spot. But the main effect of a Northerly walk is NEGATIVE FIFTY yards of elevation, when the Northerly walk begins two hundred yards to the West of the original spot.

What does this have to do with the business systems that we call organizations? Plenty! Interactions abound in every complex system. The number of interactions increases geometrically with the complexity of the system, and our organizations are exceedingly complex systems. Without knowledge of the way that the independent variables of our organizational systems interact with each other, we walk in a hostile environment filled with frighteningly deep craters.

For example, it may appear that cutting two mechanical engineering jobs from two hundred such jobs is likely to have no noticeable effect on our precious T response function. But what would be the effect of this seemingly harmless move, if one of the two mechanical engineers also happened to maintain the computing environment for twenty of his friends? Further, can we be aware of all such interactions in our organizational systems?

Of course, we cannot know all the powerful interactions that exist in our organizations, even if we try to identify them and estimate their effects. For the engineers and scientists who regularly hunt interactions in systems of even moderate complexity, this is a most challenging task. Their weapons include specially designed experiments, such as high-resolution fractional factorial experiments and D-optimal experiments. But the experiments are only the beginning. The data from these are then analyzed with a most special brand of mathematics, so that the valuable information might be extracted, condensed, and purified. Still, the process is not always successful, for a number of reasons. Chief among these is the significant level of statistical noise generated by the systems under study, noise that abounds in the more complex systems that are our organizations.

Consequently, we must question even the notion of an independent variable, when we speak of organizations. The safer assumption is, in fact, that there are no independent variables. There are only interacting variables, linked to each other by rows of invisible dominos. If we tweak one, we risk upsetting many that we never intended to change. We cannot know what powerful interactions lie before us, like so many land mines.

This is why the so-called strategy of activity based costing is every bit as dangerous as it is flawed. By suggesting that we can identify all the independent variables that we can tweak without having an adverse effect on the performance of the organization, activity based costing causes us to ignore the very existence of countless, unidentifyable, powerful interactions. It says to us, "Go with confidence! Walk West, then walk North. Life will be better when you get there." Well, maybe life will be better. But there's an equal chance that we'll have our financial legs blown off on the way.

But, this is only part of the damage inflicted upon our organizations by this non-strategy called activity based costing. Activity based costing often blocks us from exploiting even the one interaction that we know exists and that we know to be powerful and favorable. It is the interaction that exists between excess capacity and marketing.

Specifically, excess capacity has a zero main effect on the T response, in the absence of an increase in marketing activity. After all, if we don't sell the excess capacity, then we cannot convert it to T (throughput). Similarly, an increase in marketing activity has a zero main effect on the T response, in the absence of excess capacity. If we don't have the capacity with which to fill the additional sales commitments, then, again, there can be no increase in T from an increase in marketing effort. But, together the two variables have a significant, positive interaction effect. The right focus on marketing, when the organization has truly excess capacity, always improves the bottom line significantly. This is the highly beneficial interaction that many of us need to exploit, when sales are in a slump, and our manufacturing capacity exceeds our customers' demand for our current products. More of us would do so, were it not for activity based costing and the thinking that it triggers in us.

Activity based costing often blocks us from exploiting this interaction, by causing us to think in terms of improving local measurements. For example, when we have excess manufacturing capacity, either, in an entire plant or in sections of a plant, activity based costing causes many of us to think only in terms of improving the cost of the activity, often by simply cutting the capacity of a department. It misleads us into thinking that we are making real contributions to the company's bottom line, by cutting functions and jobs that appear to be unneeded. Thus, many of us think not in terms of looking for new markets into which we can sell the excess capacity but in terms of eliminating the cost associated with the excess capacity. We settle for saving pennies, when we might be able to earn millions. Yet, this is still not a complete account of the damage caused by the backward-facing gun known as activity based costing.

Just as activity based costing often blocks us from converting to additional throughput the excess capacity caused by a sales slump, it also blocks us from enjoying a steady sequence of growth spurts. It does this, by preventing us from tolerating any excess capacity, ever. Yet, excess capacity is an absolutely necessary condition for business growth. How many additional products can we possibly sell, if we don't have the spare capacity with which to produce them? None!

Oh, sure! We can always invest, if we see a new customer need. We can always buy a new manufacturing line or build a new plant, once we recognize a new market. But will the opportunity just sit there and wait patiently, while we build our new plant? Will our competitors take a time-out, while we take our time preparing? No! This spineless approach to business growth can never produce the kind of top-line and bottom-line growth that most stockholders demand. This "me too" approach to business growth can yield only the most mediocre of results; in the electronics industry, this slug-like response to the marketplace would position us like so many pounds of red meat in the midst of a hungry wolf-pack. We'd last just as long, too.

Truly excess capacity enables a quick response. It must come first. Certainly, it isn't sufficient by itself, for breathtaking speed, but it is a most necessary condition. Yet, activity based costing essentially blocks us from even tolerating excess capacity. By doing so, it blocks many of us from possessing the speed required to even survive in many markets. The damage, in lost throughput and lost profit, is incalculable.

Therefore, we absolutely must ask ourselves a most critical question: How can this so-called decision making tool, which is based on such fundamentally flawed thinking, which exposes our organizational systems to such extraordinary risk, and which often blocks us from exploiting even our own common sense, ever provide managers and executives with the kind of accurate information required to steer a corporate convoy safely through the treacherous territory of international business? It cannot! We would stand a better chance of success if we gutted a farm animal and read its entrails.

A BETTER PARADIGM

Our problem is that we try to tweak things in the first place. The Master, W. Edwards Deming, knew this to be a disastrous approach. He saw the effects that this fallacious thinking had on the quality of the output of complex manufacturing systems, for this was (and unfortunately continues to be) the approach of many manufacturing engineers. Hence, Deming spoke of having and of using a system of profound knowledge. By this, he meant that thorough knowledge of the system under study was a most necessary condition. He was right!

As we've seen, we cannot adjust complex systems constantly by tweaking one thing or even a few things at time, no matter how hard we try. The results are simply too unpredictable. Therefore, our very paradigm is at fault -- it is this paradigm that gave birth to the activity based costing monster in the first place. We need a new paradigm, one that doesn't rely on the constant tweaking of anything.

We need to design organizational systems of a different sort. We need to design organizational systems that drift naturally toward a state of optimal performance and that are inherently stable in that state.

There are many such systems in nature and in business, not to mention all the technical systems created by engineers and scientists. Even the human body is a good example of such a stable system. Consider the body's ability to maintain a nearly constant temperature, despite drastic changes in environmental temperature. Consider, too, the body's ability to maintain the concentrations of key minerals within surprisingly narrow ranges -- a process known as homeostasis. Of course, there are the many social systems as well, which demonstrate significant levels of stability. Consider the way that the market sets the prices of products. If a product is in short supply, then buyers bid the price up. The higher prices attract additional suppliers who, in turn, increase the supply of the product. In response to the greater supply, the price of the product decreases. Thus, the price of no product ever increases without bound. Instead, this system guarantees that the price remains stable. It is this kind of stability that we need to design into our organizational systems. The question, of course, is "How?"

The answer is not easy. There are no silver bullets here. But there is a logical process that we can follow. In the words of Dr. Stephen R. Covey, "Begin with the end in mind." The details...? Well, we have to leave something for the next article, don't we?

(C) Tony Rizzo, 1997. tocguy@lucent.com tocguy@home.com

This article may be reproduced only in its entirety. Any reproduction must include the author's name. This article may be published in formal publications, either in print or in electronic form, without written permission from the author.

Technorati Tags: Accounting, Cost, Cost Accounting, Costing, Finance, Focus, Shareholder Value Add, SVA, Theory of Constraints, TOC, ABC