NOTES ON REAL-TIME CONVERSIONSAdvances in information technology have encouraged a most significant change in the conduct of certainmanagerial functions: Their conversion from a planning-based to a real-time orientation! Where this has occurred,there might be an occasion for a corresponding reorientation of management science constructs. If there is anygeneral rule that applies here, it’s that a transit from a planning-based to real-time posture involves some shift inemphasis from the analytical to the procedural aspects of transactional protocols. The analytical burden is easedmainly because converting a function to real-time obviates the projection problem. This means that an applicationthat once might have involved a model containing a relatively intricate collection of anticipatory components (say anextensive array of difference equations and stochastic nodes) might in real-time require only a relatively fairly leanand structurally much simpler dynamic model. Moreover, these simpler dynamic models may well provide their userswith solutions of generally greater precision and certainty (perhaps arriving at effectively deterministic conclusionswhere once only probabilistic solutions were possible). This is obviously not true of all applications that have beenconverted to real-time; there is, for example, much merit to the argument that programmed trading is a 'mindless'exercise. As a normative matter, adopting a real-time approach to a managerial function is defensible only to theextent that it can be expected to result in a net reduction of expected value of decision error. This means that eitherthe probability or likely severity of erroneous outcomes from a decision process must be reduced as a consequence ofthe transition to real-time. Because a planning-based approach requires commitments made in advance of the pointin time at which they are to apply, a primary source of decision errors rests with the failure of reality to align well withexpectations. A real-time orientation foreshortens the decision horizon, thus giving reality more opportunity to revealitself before committing to a particular course-of-action. As a result, the degree of disparity (variance) betweenexpected and actual conditions should be more containable, so restricting the expected magnitude (severity) of theadverse impact from any errors that are made. For example, the losses following an erroneous sales forecast willgenerally be tied more or less directly to the absolute value of the difference between anticipated and actual demand. Here, in essence, is the rationale behind the real-time features of the Virtual Corporation as an enterpriseconfigured with customer responsiveness foremost in mind. Towards this end, it will have exchanged the traditionalanticipatory-based mass production approach for a real-time “mass customization” posture predicated on severalinitiatives: (1). Setting up Quick Response arrangements with customers that represent the major (or representative)retail outlets for the corporation’s products; (2). Establishment of standing Kieretsu-type alliances with key suppliersas a basis for a transition to just-in-time inventory management; (3). Emulation of the “lean” manufacturing methodspioneered by Toyota. The characteristics of the Virtual Corporation point it towards product-market sectors wherethere’s a demand for ‘mass customization’. As the concept of the Virtual Corporation becomes more common, and asflexible manufacturing facilities become more capable, mass customization can be expected to intrude into more andmore markets where mass production now rules. The competitive reach of the Virtual Corporation is restricted, however, by the very thing which mostdistinguishes it; its customer-reliance. This largely excludes it from participation in innovation-driven industries,where success depends on being able to lead rather than follow customers. Automobile firms learned something aboutthe perils of customer-reliance (and reliance on survey data) from the Edsel, a vehicle inspired by what prospectivebuyers said they wanted a car to be. But IBM’s Thomas Watson, Jr. was maybe the best teacher in this regard: Heinsisted that IBM had to anticipate the customer. Particularly, he saw “a technological imperative at work” in thefield; improvements in computer systems would follow from engineering advances, and not new user-authoredrequirements. [Once Watson’s tutelage had passed, IBM apparently forgot this lesson…for example, it undertook tobuild customers exactly the laptop that customer surveys indicated they wanted; but most of these customers then wenton to actually buy the Compaq notebook”, a product inspired by engineers rather than prospective users.] Most interesting about the Virtual Corporation is it’s intent to soften the reliance on demand forecasts. Rather,given a constant stream of updated information on actual sales from retailers, production decisions can now bepredicated more on demonstrated than estimated demand. The upshot is that such firms have little need for complexalgorithmic constructs; very simple equilibrium-maintenance type models should be sufficient to support all but themost exotic QR arrangements. Much the same conclusion follows from the to just-in-time inventory management.Producers and suppliers need only very basic control models driven by actual draw-downs and deterministic activityprojections (based on actual contractual commitments). Just-in-time inventory management is also instructive of thesomewhat more elaborate procedural underpinning of real-time operations. De-emphasizing the algorithmic aspectsof decision exercises generally means a corresponding increase in reliance on heuristic or learning-related initiatives.For example, under the assumption that more current data is more relevant or ‘revealing’ of prospectiverealities, a basic heuristic directive would have decision authorities delaying any irrevocable commitments until thelast possible moment. The corollary would then be that any commitments that must be made should—to the fullestfeasible extent—be staged in order to exploit every available opportunity for in-vivo learning. Staging also reduces theexpected value of decision errors by limiting the amount that’s at risk at any point in time (with this effect mostpronounced where it’s most
View Full Document
Unlocking...