cannot be achieved without a proper under- standing of the concept of a system. Deming emphasized that fact over and over, and it is at the heart of his idea of profound knowledge. Yet the way most managers were trained to think implicitly ignores the system concept. In an economy that is global in scale, and one in which technological and product life- cycles turn several times inside many compa- nies decision radius (time to make a deci- sion), it is necessary to understand three critical and interrelated concepts: a system, emergent properties, and complexity. We need to explore each of these concepts: (1) System: a set of different elements so connected or related as to perform a unique function not performable by the elements alone (Rechtin and Maier, 1997). (2) Emergent properties: those functions, attrib- utes or behaviors, good or bad, which would not exist except for the operation of a system. (3) Complexity: something that is composed of interconnected or interwoven elements that function as a system to produce emergent properties. A syst em A system is an aggregation of elements, but not just any aggregation will do. Several require- ments must be met before we can call some- thing a system. First, there must be emergent properties, i.e. it must provide something that is not available with the parts alone. That requires interconnection and interrelationships. Deming was fond of pointing out that it is not sufcient to gather together all of the best automobile parts in one place. That collection does not make a system. To become a system they must be designed to work together, or the connections that make the system called auto- mobile, which provides the emergent property called transportation, will be missing and the parts will remain a useless pile of materials. A good system both is purposeful and has value to some customer or user. A farm tractor is a system similar to an automobile: it has four wheels, engine, seat, transmission and (today) an air-conditioner. But if you do not have to plow a eld, and do have to transport a family around town, then a tractor might be purpose- ful, but does not have value to you. Lacking either purposefulness or value renders the system little more than an object of curiosity. Most good systems also are feedback con- trolled systems. In order to maintain its cur- rent state, and improve in the future, some 54 The TQM Magazi ne Vol ume 11 Number 1 1999 pp. 5457 MCB Uni versi t y Press ISSN 0954-478X Techniques What s t his syst ems st uf f, anyhow? H. J ames Harrington J oseph J . Carr and Robert P. Reid The aut hors H. James Harringt on i s an Int ernat i onal Qual i t y Advi sor Pri nci pal at Ernst & Young LLP, San Jose, Cal i f orni a, USA. Joseph J. Carr i s Seni or Syst ems Engi neer, Naval Ai r Syst ems Command, Pat uxent Ri ver, Maryl and, USA. Robert P. Reid i s an i nt ernat i onal l y know n i ndependent consul t ant on organi zat i onal ef f ect i veness. He i s a ski l l ed t eacher of TQM and many ot her rel at ed t opi cs. He i s based at Kel l er Graduat e School and Uni versi t y of Maryl and, USA. Keywords Demi ng, Syst ems devel opment , Syst ems t heory Abst ract Di scusses a si mpl e overvi ew of syst ems and syst ems t heory. Descri bes w hat a syst em i s and does. Suggest s seven pri nci pl es t o use w hen desi gni ng or eval uat i ng a syst em i ncl udi ng t i ps such as: organi ze t o reduce compl ex- i t y, do not over-opt i mi ze t he syst em, and l eave det ai l s of managi ng sub-el ement s t o t he speci al i st s w ho know how. Il l ust rat es w i t h graphi c exampl es. form of feedback is required, but the feedback is also a source of complexity and has both advantages and risks. Emergent propert ies It is the emergent properties that give a system life, and it is both the nature of the elements and their interrelationships that give rise to the emergent properties. To give rise to the family transportation emergent property an automobile must have certain components, and they must be congured in a way that gives them the correct interrelationship. For example, you might have the best tires and the best axle, but if they are at right angles to each other (Figure 1), then their interrelationship does not allow the emergent property needed to work in an automobile system. Which emergent properties are useful depends on the nature of the system. But here is the dark side of the force: not every emer- gent property is usefulsome are either wasteful or dangerous. Some systems do unexpected things, show surprising behavior, or result in unintended consequences. Complexit y A system produces emergent properties, and that infers the existence of the interconnections and interrelationships that make complexity. It is in the complexity of the system that both the opportunities and the dangers lie. The com- plexity gives rise to the emergent properties that dene the system and make it valuable, but it is also the source of the most serious, and least expected, problems. It is in the nature of complexity that most people lose the systems bubble. Complexity was not even studied seriously or widely until the 1970s, and when it was studied some surprising ideas emerged. It was discovered that complexity is on the cusp between stabili- ty and chaos, and that only a small amount of perturbation is needed to push the system one way or the other. In recent years, some people have attempt- ed to apply chaos and complexity theory to management systems. It is noted that com- plexity gives rise to new systems that did not exist before. Some people seem to believe that this leads to a kind of mystical creativity that can lead to ultimate success. Although there was some good material published, there is also a lot of evidence that many of the authors did not understand the nature of complexity. To be sure, a complex system can erupt into chaos, and from that chaos can emerge new stable structures. Order out of chaos is the buzz word used to describe this very real process, but what is rarely addressed is the fact that you might not like the new stable state. In other words, there is no guarantee that the new stable order is better than the old. It is in the best interests of all aboard if an airliner is in a stable ight regime. Or is it? Not all stable states are equal. Ask any pilot whether a at spin is a desirable state. Stable, yes, but desirable, no! Bankruptcy and death are also examples of stable states. Research demonstrates that systems with an order of complexity as small as three ele- ments, with two interconnections per ele- ment, can produce chaotic behavior. In terms of a product or organization, this means unexpected behavior (good or bad), unin- tended consequences, and unpredictability. Not all such systems will produce chaos, but the possibility certainly exists. How we t hink about syst ems Most managers today are singularly ill equipped to deal with complex systems, whether that system is their organization or the products it produces. The root of the problem is the way we were trained to think about problems. Our basis for solving problems, and thinking about sys- tems, is reductionism and analysis. In other words, break the system down into smaller elements that can easily be analyzed, rather than the larger entity that cannot. There is a small problem with that approach: the very interrelationships and connections that make the system behave as a system are lost in the breaking down. You can analyze tires, engines and transmissions forever and not come up with the system automobile or its emergent property of transportation. How good would your physician be if he or she had an innite knowledge of the cells of the human body, but had not a clue how they worked together or how they malfunctioned? Without using an overarching way of look- ing at the system, there is little possibility of understanding it. And without understanding a system there is very little possibility of reengineering it to do anything useful. One of the serious errors made by the reengineering movement was the notion that they could 55 What s t hi s syst ems st uf f, anyhow ? H. James Harri ngt on, Joseph J. Carr and Robert P. Rei d The TQM Magazi ne Vol ume 11 Number 1 1999 5457 Figure 1 Int errel at i onshi p of component s build a new system tabula rasa, without regard for the old system. Lacking understanding of the old system (its function, purpose, design, operation and value to the stakeholders) makes it very difcult, if not impossible, to architect a new system. What t o do? The modern business organization is simply too complex to succumb to simple analysis, and the more complex it is, the more likely it is to discover a complexity that can degenerate into an undesirable at spin state. Fortu- nately, there are some principles that will help when evaluating an organizational system, or designing a new one: Deneand managetheinterfaces. An inter- face is a point where two elements come together and exchange something. In other words, it concerns the interconnections and interrelationships that make up a system. The most important key is to realize that it is in designing and managing those interfaces that one gains the greatest leverage, and faces the greatest risks. Organizeto reducecomplexity. If the real way things get done is considerably differ- ent from the right way found in a policy manual, then it is a sure bet that the inter- faces are mismanaged and/or are too com- plex. People will nd a simplied way to work around a bad system, but they should not have to do so. Furthermore, if they use work arounds then there is an increased probability of a at spin behav- ior emerging: stable, but destructive. It is well recognized that a poor design, i.e. one that is too complex, may well lead to greater, undesirable complexity when infor- mal attempts to simplify things are made. Design it simple, stupid (DISS). A high degree of complexity is necessary in many of our organizations, but that does not mean that the relationship of elements of the overall system must be complex. The goal is to create an organization in which the ele- ments have a high order of internal com- plexity, and a low order of external complex- ity. Each element, in other words, must be as independent as possible. A robust and responsive organization will result. Makeinterfaces clear and well dened. Inter- faces that are ambiguous, gray, subject to interpretation will result in variation in behavior. High complexity and ill-dened interfaces can be described in a single word: disaster. Do not over-optimizethesystem. It is a natur- al tendency for managers to require all elements to optimize their performance. There is a simple rule that applies to all complex systems: not all elements of a computer system can be optimized; at least one element must be sub-optimized to optimize the entire system. This advice ies in the face of what many people believe to be true. However, consid- er this true situation. A company required all departments to optimize by attaining the lowest costs possible. The travel ofce was required to obtain the lowest air fares possible. The optimization of the travel ofce meant that expensive engineers, marketing people, and executives either had to endure many hours of layover at airports, or took ights that spent all day to arrive at a destination because of all the intermediate stops. The travel ofce opti- mization, done according to the general rule for the entire organization, cost a huge amount of money before it was corrected. Or consider the urge to optimize the big three of program management: cost, schedule and performance. At least one of these must be sub-optimized if the entire program is to be optimized. Whoever came up with the notion of cost as an indepen- dent variable did not understand that the big three are interdependent, and none is truly independent. Among the big three, there are no independent variables because of their mutual dependence. Leavedetails of managing sub-elements to the specialists who know how. A good higher level manager understands that it is the interfaces that concern them, and not the day-to-day operation of the elements of the system. There are signicant differences between the thinking required at higher and lower levels of complexity within the organization (Rechtin and Maier, 1997). Manageperturbations and changeas a system. The mark of a good system is that it is robust to challenges, perturbations and change. Systems, including organizations, will respond to these threats in one of three manners (Figure 2). The trajectory to the changed state in Figure 2a is that which is implicitly, if erroneously, modeled by all too many people. Something happens, then by some process of a miraculous immaculate metamorphosis the new state is achieved. Real systems behave differently. The approach in Figure 2b shows what happens in many organizations. The perturbation occurs and, for one reason or many, the organization jumps into chaos. Massive disruption occurs, and there is even danger of the organization settling into a new state that is a at spin rather than something desirable. This situation is called subcritically damped change. 56 What s t hi s syst ems st uf f, anyhow ? H. James Harri ngt on, Joseph J. Carr and Robert P. Rei d The TQM Magazi ne Vol ume 11 Number 1 1999 5457 Figure 2c depicts a situation that occurs in many organizations. There is so much resis- tance and foot dragging that the change to the system does not occur on time or as planned. This situation is called overcritically damped change. The ideal practical situation is depicted in Figure 2d. This is critically damped change. So how do we achieve this approach? In prac- tice, we do not. Systems tend to act like Fig- ures 2b or 2c, rather than 2d. According to Rechtin and Maier (1997), it is preferable to implement several intermediate stable states on the way to the nal stable state. Each time change is applied, there will be some familiar anchor in the old system that people can live with. Physicists and engineers who study feed- back controlled systems call this approach quasi-static change (Figure 3). They know that such systems may go into either chaos or some at spin regime if perturbed too much at one time. However, if the system is per- turbed a little bit at a time, and the minor perturbations that result from each change are allowed to die out before the next change is made, the end state can be reached with no massive and destructive upheavals. Done otherwise, you will get to a new stable state, to be sure, but there is no telling what that state will be like. Ref erence Recht i n, E. and Mai er, M.W. (1997), The Art of Syst ems Archi t ect i ng, CRC Press, Boca Rat on, FL. 57 What s t hi s syst ems st uf f, anyhow ? H. James Harri ngt on, Joseph J. Carr and Robert P. Rei d The TQM Magazi ne Vol ume 11 Number 1 1999 5457 OLD STATE SOMETHING HAPPENS NEW STATE Figure 2a Unreal i st i c t raj ect ory t o changed st at e OLD STATE SOMETHING HAPPENS NEW STATE MASSIVE DISRUPTION OCCURS! Figure 2b Subcri t i cal l y damped change SOMETHING HAPPENS OLD STATE NEW STATE PROGRESS TAKES TOO LONG Figure 2c Overcri t i cal l y damped change OLD STATE SOMETHING HAPPENS NEW STATE ABOUT RIGHT Figure 2d Cri t i cal l y damped change OLD STATE SOMETHING HAPPENS NEW STATE USE MULTIPLE INTERMEDIATE STABLE STATES Figure 3 Quasi -st at i c change Comment ary To understand quality weneed to understand systems. This is a great start.