Industrial Dynamics--A Major Breakthrough for Decision Makers

April 28, 2017 | Author: kakilinha | Category: N/A
Share Embed Donate


Short Description

Download Industrial Dynamics--A Major Breakthrough for Decision Makers...

Description

Harvard Business Review July-August 1958

INDUSTRIAL DYNAMICS a major breakthrough for decision makers

By Jay W. Forrester

clear or vague, but in any event they have a subtle and far-reaching impact on administrative thinking and decisions. A look at some promising new concepts of management should, I believe, convince even the skeptical executive that his job is developing into much more than an art, that conceptual skill will play an increasingly vital role in company success, and that management is fast becoming second to none as an exciting, dynamic, and intellectually demanding profession. •

Management is on the verge of a major breakthrough in understanding how industrial company suecess depends on the interaetion between the flows of information, materials, money, manpower, and capital equipment. The way these five flow systems interlock to amplify one another and to cause change and fluctuation will form a basis for anticipating the effects of decisions, policies, organizational forms, and investment choices. My aim in this article is to look ahead at the specific kinds of progress which will be achieved and at the concepts which will make this progress possible. While I shall suggest certain ways of thinking about management that should be helpful to executives today in working on inventory control, production scheduling, advertising, sales, and other problems, my primary concern here is not with techniques and preseriptions. Rather, I am interested in the development of a professional approach to management. Business leaders, like leaders in other areas, are influenced by an "image of the future." Their ideas about where they are going may he

To develop the status of a profession, management must discover the underlying principles which unify its separatti aspeets. It must develop a basic theory of behavior. It must learn how to convert experiences and particular case examples into a contribudon to this general theory. And, finally, it must be able to employ the basie principles of the tlieory as a useful practical guide for explaining and solving new problems as they arise. By accomplishing these aims, management will become; a true profession during the next generation. [

AUTHOR'S NOTE; The studies on which this article is hased were made possible through the support of the Sloan Research Fund of the School of Industrial Manage-

ment at Massaehusetts Institute of Technology, the Ford Foundation, and the use of the IBM 704 computer at the Massachusetts Institute of Technology Computation Center._



1



I

Toward a Theory

37

38

Hanmrd

Business

The task of management is to interrelate the flows of information, materials, manpower, money, and eapital tiquipment so as to achieve a higher standard of living, stability of employment, profit to the owners, and rewards appropriate to the suecess of the managers. Looked at in this way, its goals are rooted as deeply in the public interest as the hroad objeetives of the legal, medical, and engineering professions. In the past, with management considered more of an art than a profession, edueation and practice have heen highly fragmentized. Manufacturing, finance, distribution, organization, advertising, and researeh have too often been viewed as separate skills and not as part of a unified system. Too often management education eonsists of gathering eurrent industrial praetice and presenting it to the student as a sequence of unrelated subjeets. Similarly, in his work in industry the manager specializes within departments where his experience perpetuates the atmosphere of unrelated eompartmentalization. The next big step in management edueation will be the development of a basis for fitting together the many management functions into a meaningful whole. Around this eentral core specialized suhjeets and experience will take on more significance. Men can be developed more rapidly. They will he able to start from a point now accessible only through long training or fortuitous experienee. Sueh strides will far exceed in importanee reeent steps in using computing machines to execute clerical tasks or in applying operations research methods to isolated company problems. For we can cxpeet to gain, during the next 25 years, a far better understanding of the dynamic, ever-changing forces which shape the destiny of a company. This understanding will lead to better usage of available information, to improved understanding of advertising eifeetiveness and the dynamic behavior of the consumer market, and to company policies that keep paee with teclinological change. Beyond these aehievemcnts, there will he improvements in company organization resulting from a sounder basis for effeetive decentralization, from altering the relationships between line and staff tasks in the eompany, from the more effective utilization of scientific manpower, and from redueing the routine duties and enhancing the ereativity of managers. And executives will gain in "elairvoyanee.'' For example.

they will be able to anticipate clearly (as I shall illustrate later in the artiele): • How small changes in retail sales can lead to large swings in factory production. • How reducing clerical delays may fail to improve management decisions significantly. • How a factory manager may find himself unable to fill orders although at all times able to produce more goods than are being sold to consumers. • How an advertising policy can have a magnifying efEcct on production variations.

Tools of Progress The new management eoncepts will rest in part on reeent advanees in the data-processing industry, in part on military research (which has given us an improved understanding of deeision making and experience in analyzing and simulating the characteristics of complex systems), and largely on 20 years of research in information-feedback systems. Electronic Data Processing The performance of electronic computers has increased annually hy a factor of nearly 10 per year over the last deeade; in almost every year we have seen a tenfold increase in speed, memory eapaeity, or reliability. This represents a teclinological change greater than that effected in going from chemical to atomic explosives. Soeiety eannot ahsorh so big a change in a mere ten years. We therefore have a tremendous untapped baekiog of potential devices and applications. Vic\ving data processing in another way, we can trace the shifting frontiers of the field by dividing progress into five-year periods: •[From 1945 to 1950 was the period of clectronie research and the demonstration that machines having many thousand vactium tubes would indeed operate. C From 1950 to 1955 was the pioneering period in applying computers to the solution of scientific and engineering problems. « During the present period, 1955 to i960, electronic machines are being substituted for clerical effort in commercial organizations. C From i960 to 1965 we can expect to see the application of digital computers to physical process control. Already there are digital machine tool

Industrial Dynamics controls and the SAGE air defense system; we can look forward to dramatic improvements in processes, to reduction of capital investment requirements in the oil and chemical industries, and to civilian air traffic control systems, all based upon new kinds of information utilization for control purposes. c From 1965 to 1970 we should see all these developments converging into pioneering imprt>vements in the central management process. The routine, repetitive types of decisions will become more formalized, while management creativcncss will be directed to how decisions and policies should be made rather than to the actual repetitive making of such decisions. During the 1965-1970 period of widespread company testing of new management methods, electronic data processing will be an essential tool. ] Let me emphasize, however, tfiat it is no more the focal point of the future management profession than a slide rule is the essence of engineering. Decision Making Historically, military necessity has often led not only to new devices like aircraft and digital computers but also to new organizational forms and to a new understanding of social forces. These developments have then been adapted to civilian usage. Sueh new exploration is now happening in the military command (or management) function. As the pace of warfare lias quickened, there has of necessity been a shift of empbasis from the tactical decision (moment-by-monient direction of the battle) to strategic planning (preparing for possible eventualities, establishing policy, and determining in advance how tactical decisions will be made). The battle commander can no longer plot the course of his enemy on a chart and personally calculate the aiming point. In fact, with a ballistic missile he would have no time even to select his defensive weapon. Likewise in business: as the pace of technological change quickens, corporate management, even at the lower levels, must focus more and more on the strategic problems of running the business and less and less on the everyday operating problems. In the systems development for tbe military, it has been amply demonstrated that carefully selected formal rules can lead to tactical decisions that exGel those made by human judgment

39

under the pressure of time and with insufficient experience and practice. Furthermore, it has been found that men are just as adaptable to the more abstract strategic planning as they are to tactical decision making, once their outlook has been lifted to the broader and longer-range picture. Simulation Also from military research we have available the methods of simulation for determining the behavior of complex systems. These teehniques have reached the state of development where they can now be usefully applied to industrial organizations. Simulation is being used in the design of air defimsc systems and in engineering work. For example: In planning the development of a river basin, numbers in a digital con^;puter represent water volumes, flow rates, electric demand, and rainfall. A few seconds of computer time can solve a whole day of .systems operation. Dams can be located and designed for an optimum compromise between power generation, irrigation, navigation, and flood control. In business, simulation means setting up in a digital computer the conditions which describe company operations. On the basis of the descriptions and assumptions about the company, the computer then generates the resulting charts of information concerning finance, manpower, product movement, and so on. Different management policies and iiiarkct assumptions can be tested to determine their effect on company success. To use simulation studies will not require undue mathematical ability. To be sure, details of setting up a model -vill need to be handled by experts because there are special skills required and pitfalls to be avoided. However, the job of direeting the situations to be explored, judging the assumptions, and interpreting the results will be within tbe ability of the type of men we now see in management schools and executive development programs. i

Feedback Control Systems of information feedback control are fundamental to all life and human endeavor, from the slow pace of biological evolution to the launching of the latest satellite. A feedback control system exists w^henever the environment eauscs a decision which in turn affects the original environment. To illustrate:

40

Harvard Business Review • A thermostat receives temperature information, decides to start the furnace, and changes the temperature. • A person senses that he may fall, corrects hi.s balance, and thereby is able to stand erect. • In business, orders and inventory levels lead to manufacturing decisions which fill orders and correct inventories. • A profitable industry attracts competitors until, to use the economist's terms, the profit margin is reduced to equilibrium with other economic forces. • The competitive need for a new product leads to research and development expenditure that produces technological change.

ments, just as an electronic computer has certain characteristics as a system of parts. For instance, there is what the engineer might eall "amplification," caused by inxentory accumulation, filling of supply pipelines, and inept extrapolation of trends. There are delays in decisions, shipping, communications, and accounting. These all combine to cause production fluctuations, construction of excess plant capacity, creation of company-generated rather than eustomer-generated seasonal sales, and detrimental advertising policies. Without an awareness of basic informationflow principles, it is only through costly errors that managers can develop an effective intuitive judgment. For example:

All of these are information feedback control loops. The regenerative process is continuous, and new results lead to new decisions \vhicli keep the system in continuous motion. The study of feedback systems deals with the way information is used for the purpose of control. It helps us to understand how the amount of corrective action and the time delays in interconnected systems can lead to unstable fluctuation. Driving an automobile provides a good example:

In a company manufacturing consumer durables, it was discovered, after all data became available, that in a certain year retail sales had varied by 30%. In the same year the inventory and ordering practices in the distribution system and at the factory caused this small retail variation to be amplified to a four-to-one, or 400%, variation in factory produetion. The estimated avoidable costs of increasing and decreasing production and of carrying excess inventory were equal to the normal anticipated profit margin! (We shall see later how this could happen.)

The information and control loop extends from steering wheel, to auto, to street, to eye, to hand, and back to steering wheel. Suppose the driver were blindfolded and drove only by instructions from his front-seat companion. The resulting information delay and distortion would cause erratic driving. If the blindfolded diixer could get in.structions only on where ho had been from a companion who could sec only through the rear window, his drivijig would he even more erratic. Yet this is analogous to the situation in business. Top executives do not see the salesmen calling on customers, do not see the prospective buyers watching a TV commercial. They do not attend the board meetings of competitors. They do not have a clear view of the road ahead. The only thing they can tell with reasonable certainty (and even here there is sometimes doubt) is what happened to wages, sales, material costs, interest rates, and so on last year. Smoother Operations The quality of management control depentls upon what information executives use and for what they use it, as well as on their skills as administrators. As a system the company has certain characteristics which are completely independent of individual functions or depart-

Feedback theory explains how decisions, delays, and pretlictions can produce either good control or dramatically unstable operation. It relates sales pronn)tion to production swings, purchasing and pricing policies to inventory fluctuations, and typical life eycles of products to the need for researeh.

Production & Distribution How do the concepts and approaches 1 have been discussing apply to specitie business problems? Or, to be more precise, how do the delays, ampliiications, and oscillations in the circuital floiv of information in a company affect its operations? Now, in order to illustrate, I must eompromise one of the main ideas I have speeified; namely, that the system (meaning not tbe paperwork forms and proeedures, but the interrelationships between all the company operations) behaves aeeording to the characteristics of the whole antl not according to the characteristics of individual parts. For the sake of brevity, I shall have to limit myself to the production and distribution functions, forgetting research, engi-

Industrial Dynamics necring, sales, and, until a later point in the article, promotional effort. In a sense this is a subsystem, with its own eomplex of interrelationships tliat ean be studied as such, hut it is not the whole. Even \vitliin this ptoduetiondistrihution suhsystem, I shall have space to inelude only the Hows oi' intormation and materials. Omitted altogether are considerations of money, manpo\ver, and capital equipment. If I can make the prineiplos of the new approach clear, however, it should not he difficult for the reader to extend his own pieture to ineludc the areas that 1 must omit. Needed Information To l)cgin the study of our example, we need to know three kinds of information ahout the system: its organizational structure, the delays in decisions and actions, and the policies governing purchases and inventories. 1. Organizational Structure. lixiiiBiT i shows a typical organizational structure for the production and distribution functions in a hard goods industry like household appliances. If we examine the basic internal behavior of the system when customer orders are independently specilted, we will see that, even with very simple retail sales changes and with no other external disturbances afEecting the company, typical manufacturing and distribution practices can generate the types of husiness disturbances which are often blamed on conditions outside the company. For random, meaningless sales fluctuations can be converted into annual, seasonal production cycles. Advertising and price discount policies of an industry can create twoancl three-year sales cycles. Factory capacity, even though always exceeding retail sales, can seem to fall short of meeting demand with the result that production capacity is overcxpanded. Examining EXHIBIT I more closely, we see that the bottom box represents the retail level. Next EXHIBIT I. ORGANIZATION OF I'RODUCTION-DISTBIBUTION SYSTEM

41

above arc the distributors, and at the left the factory and factory warehouse. The dotted lines show the information How, here consisting of orders for goods ilowing upward. The solid lines show the shipment of goods flowing downward. 2. Delays in Decisions and Actions. To be able to determine some of the dynamic characteristics of tliis system we must also know the delays in the fiow of information and goods. The time delays are shown on the diagram in weeks and are reasonable values for a consumer durable product line. DeUi'ery of p,oodx to the eonsumer averages a week after the customer places an order. At the retail level, the accounting and purchasing delays average three weeks between the time of a sale and the time when that sale is reflected in an order sent out to ohtain a replacement. Mailing delay for the order is half a iveek. The distributor takes a week to process the order, and shipment of goods to the retailer takes another week. Similar delays exist between the distributor and the factory warehouse. The factory lead time averages six weeks between a deeision to change production rate and the time that factory output reaches the new level. Note that three levels of inventory exist — factory, distributor, and retailer. 3. Policy on Purchasing Orders and Inventories. To complete the initial description of our example, we need to know the policies followed in placing orders and maintaining inventory at each distribution level. Let us consider three principal types of orders: (a) orders which directly reflect sales, (b) orders to adjust inventories with changes in business volume, and (c) orders to fill the supply lines with in-process orders and shipments. Let us suppose further that orders are treated in the following ways: • After a sales analysis and purchasing delay (three, two, and one weeks for the three levels), orders to the next higher level of the system include the actual sales made hy the ordering level. • After a time for averaging out sales fluctuations (eight weeks), a gradual upward or downward adjustment is made in inventories as the rate of sales increases or deereases. • The orders in process (orders in the mail, uniilled orders at the supplier, and goods in transit) are proportional to the level of business aetivity and to the length of time required to fill an order. Both an increased sales volume imd an increased delivery lead time neeessarily result in increased total orders in the supply pipeline. The ordering rate will also depend on some presumption about future sales. Prediction methods that amount to extending forward (extrapolating)

42 Harvard Business the present sales trend AviU in general produce a more unstable and fluctuating system. For our example, however, we will use the conservative practice of basing the ordering rate on tiic assumption that sales are most likely to continue at their present level. Simulation Methods Before we ean determine how our system will function over a period of time, all of the ahove rather general descriptions of the system must he expressed in explieit quantitative form. Illustrations of these relationships are shown in F-',xHIBIX II.

The next step is to determine how the system as a whole behaves. To do so, we might use some pattern of consumer purchases as an input and obser\ e the resulting inventory and production changes. A very good test would be to sec what happens after a small sudden change in retail sales. The effeet on the company system EXHIBIT II.

ean be ohtaincd hy "simulation" methods — simulation being the technique of obtaining results from a model. For the industrial system, simulation consists of tracing through, step by step, the actual flow of orders, goods, and information, and observing the scries of new deeisions required. For example: Our production-distribution system might be simulated by a group of men around a table, one to represent retailers, another the postal service, another transportation, another the factory, and so on. Five minutes might represent a week, and in each time interval the proper purchase orders and deliveries would be made according to the rules illustrated in EXHIBIT II. Alternatively, the whole exercise can be done by one person in tabular form on paper. Better still, the entire sequence can be programed on a digital computer. Digital computer simulation was used to obtain the results in EXHIBIT HI. A sudden i o %

F O R M A L QUANTITATIVE S T A T E M E N T S O F T H E BF.LATIONSIIIPS IN E X H I B I T I

To determine the behavior of a system by simulating the performance of its xiarts requires that one describe exactly, and in detail, the characteristics which are to be included. The vahdity of the outcome of tbe system studies depends on tbe judjjment of what is pertinent to include in tbe system description. Tbe following examples sbow some of the kinds of relationships that arc needed in tbe study of the production-distribution functions of EXHIBIT I :

(1) Tbe inventory level at retail at the end of any time period is tbe inventory at the beginning of the period, plus goods received, minus goods tk-livcrcd during tbe period,

Sr.lil —

Lr V Dr., Tbis is to represent tbe aggregate of many retailers eacb selling a multiplicity of catalogue items in tbe product line. (4) We need to know tbe rate at wbieh tbe retail level purcbases from the distributor level. This reflects (a) tbe rate of customer purcbases from retail plus (b) tbe rate necessary to adjust tbe retail inventories plus (c) tbe adjusting ol" orders in process in tbe pipeline between retailers and distributors. Tbe matbematical statement is:

ibc relationsbip can be expressed as: Ir.k ^'- Ir.J -h AtRa.Jk ~ AtSr,)k

wbicb says that tbe inventory at retail, Ir, at time k is equal to tbe inventory one time period earlier at time j , plus tbe goods received in tbe time interval (tbe length of tbe time interval. At, multiplied by tbe rate at whicb goods arc received from the distributor, Ri, over tbe time interval j to k) minus tbe goods delivered (At multiplied by the rate of sales to customers, Sr, in tbe interval j to k). At must be a solution time interval tbat is sbort compared with any significant time delays wbicb are to be represented in tbe system. (At is bcrc 1/20 of a week, since tbircl-order exponential delays as sbort as balf a week — tlie mailing delays — arc used in tbe system description.) (2) Similarly, luifillcd orders equal previous unfilled orders plus new orders minus shipments; tbat is: Ur.t == Ur,J + AtN.Jt - AtSr.Jk wbcrc Ur represents unfilled orders at tbe retail level; Nr is tbe rate of receipt of new orders; and Sr is tbe rate of delivery of goods. (3) Sbipments by tbe retailers are dependent on various factors, bcre assumed to be tbe normal order-processing delay, d,,r, tbe level of unfilled orders, Ur, and tbe physical ability to till orders based on a ratio of actual inventory level, Ir, to a desirable level of inventory, Dr. Expressed mathematieally:

This says tbat tbe purehasing rate by retailers, Pr, in tbe next time interval, kl, is equal to tbe purchasing rate by customers, P,-, in tbe previous time interval, jk, as it is thougbt to be after data-processing delays (provided by anotber equation), plus an allowance for inventory and pipeline adjustments wbicb are made gradually over tbe length of time, dir. Tbe "desired owncrsbip," consisting of tbe desired inventory, Dr, plus tbe orders wbieh need to be in process to sustain tbe current business level (obtained by multiplying tbe total delay, d^, a variable time, in filling orders by tbe average rate of recent sales, Ar.Oi is: (Dr,K -f d,,,. Ar.v) The "actual ownership," represented by tbe present inventory, Ir, plus the actual orders in process, Gr, is: I,.,k + G.,K Some 40 rclationsbips of tbe above type serve to relate orders, sbipments, purcbases, mailing delays, shipping delays, factory lead time, and so forth. It is not tbe purpose of tliis article to do more than illustrate the nature of tbe relationsbips which arc required. Mucb study of any particular industrial system would be necessary to determine the relationsbips tbat meaningfully portray its bebavior. Line executives must set tbe stage for such a study, and specify assumptions to be used and results to be sought, but tbe detailed work is for tbe experts.

Indusirial Dynamics 43 EXHIBIT HI. RESPONSE

OF PRODUCTION-DISTRIBUTION INCREASE IN RETAIL

SYSTEM

T O A SUDDEN I O %

SALES

RETAILERS'ORDERS FROM CUSTOMERS DISTRIBUTORS" ORDERS FROM RETAILERS FACTORY WAREHOUSE ORDERS FROM DISTRIBUTORS -t-407... FACTORY PRODUCTION OUTPUT

WEEKS

increase in retail sales was introduced in January. The resulting fluctuations are shown in order rates, factory output, factory warehouse inventory, and unfilled orders. (Retail orders are here given as an input which is independent of what happens in the production-distrihution system itself. Retail orders are, of course, not actually independent since they are affected hy availability of the product and by advertising, a point to he discussed later.) Fluctuating Behavior Because of accounting, purchasing, and mailing delays, the increase in distributors' orders from retailers lags about a month in reaching the io% level. It is important to note, however, that the rise does not stop at io%. Instead it reaches a peak of i6% in March because of the new orders that have been added at the retail level (a) to increase inventories somewhat and (b) to raise the level of orders and goods in transit in the supply pipeline by io% to correspond to the io% increase in sales rate. These inventory and pipeline increments occur as "transient" or "nonrepeating" additions to the order rate and, when they have been satisfied, the re-

tailers' orders to the distributors drop back to the enduring io% increase. The factory warehouse orders from distributors show an even greater swing. The upturn in orders at the factory warehouse lags the upturn at the distributor level. The incoming order level for distributors is above retail sales for four months, rea
View more...

Comments

Copyright ©2017 KUPDF Inc.
SUPPORT KUPDF