Knowledge that Transforms

We consider a stochastic, capacitated production-inventory model in which the customer provides information about the expected timing of future orders to the supplier. We allow for randomness in customer order arrivals as well as the quantity demanded, but work under the assumption that the customer is making every effort to follow the schedule provided. We term this as a target reverting policy. This gives rise to an interesting nonstationary inventory control model at the supplier. After characterizing the optimal policy, we develop solution procedures to compute the optimal parameters. An extensive computational study provides insights into the behavior of this model at optimality. Further, comparing the cost of the optimal policy to the cost of simple policies that either ignore the customer's information or the capacity constraint, we are able to provide insights as to when these simplifications could be costly.

The bottleneck in a production-inventory network is commonly taken to be the facility that most limits flow through the network and thus the most highly utilized facility. A further connotation of “bottleneck,” however, is the facility that most constrains system-wide performance or the facility at which additional resources would have the greatest impact. Adopting this broader sense of the term, we look for fill-rate bottlenecks: facilities in a production-inventory network that most constrain the system-wide fill rate (the proportion of demands filled within a fixed delivery leadtime) or facilities at which either additional production capacity or additional inventory would have the greatest impact on the fill rate. We consider systems in which various components are produced through a series of stages holding intermediate inventories and are then assembled into finished goods to meet external demands. With each station in the network we associate precise measures of the station's propensity to constrain the fill rate. We call a station with a minimal measure a fill-rate bottleneck and justify this label both theoretically and numerically. Examples show that even the least utilized facility can be a fill-rate bottleneck. Unlike utilization, our bottleneck criteria capture information about process variability.

The publication of Volume 1, Number 1 of M&SOM marks the end of a process that began more than three years ago: a process whose goal was to create the premier journal in operations management. I will leave it to you, the reader of 1:1, to decide whether or not we have been successful “right out of the blocks” or if more issues must be published before we can rightfully claim to have succeeded. But we will suceed. I am committed to it. So are the distinguished group of senior editors and the esteemed members of M&SOM's Editorial Review Board, whose names are, and will remain, prominently featured in every issue. And most important, so are the authors who have submitted their manuscripts to us.

The relentless pursuit of increased product quality via continuous improvement is an important long-term strategy for achieving competitive advantage. However, manufacturers must still achieve high product quality in the short run. Hence, short-run quality improvement strategies are necessary, and, if possible, should complement (rather than substitute for) a longer run continuous improvement strategy with suppliers. We propose a novel short-run quality improvement strategy with suppliers which is based on a combinatorial optimization model for determining the optimal matching of raw material batches (from suppliers) with process variable parameters (of the manufacturer) by exploiting their interaction effects. Although the model can be difficult to solve, we develop sufficient conditions for a special case where the attainment of an optimal solution is straightforward. We demonstrate the use of the proposed approach in an actual pharmaceutical process. Using data from the pharmaceutical case, we develop several simulation scenarios where matching is shown (predicted) to greatly improve process yield. We also discuss the use of matching to illuminate various strategies for long-term continuous improvement of supplier and manufacturer processes in general.

We study the problem of allocating stocked fibers to made-to-order cables with the goals of satisfying due dates and reducing the costs of scrap, setup, and fiber circulation. These goals are achieved by generating remnant fibers either long enough to satisfy future orders or short enough to scrap with little waste. They are also achieved by manufacturing concatenations, in which multiple cable orders are satisfied by the production of a single cable that is afterwards cut into the constituent cables ordered. We use a function that values fibers according to length, and which can be viewed as an approximation to the optimal value function of an underlying dynamic programming problem. The daily policy that arises under this approximation is an integer program with a simple linear objective function that uses changes in fiber value to take into account the multi-period consequences of decisions. We describe our successful implementation of this integer program in the factory, summarizing our computational experience as well as realized operational improvements.

In preparing a review I recently discovered an important reference for a key result in Graves (1999). Wecker (1979) had previously derived the variance for demand over a deterministic lead-time for an IMA (0, 1, 1) demand process. I develop effectively the same result, given as Equation (8) in Graves (1999). I state this as the variance of the inventory random variable for an inventory system that is subject to an IMA (0, 1, 1) demand process, a deterministic replenishment lead-time and an adaptive base-stock control policy given by (7). But given these assumptions, the variance of the inventory is the same as the variance of the demand over the lead-time. Although the Wecker manuscript has not been published, Eppen and Martin (1988) reference it and use the key result as part of their research.

Paced or Synchronous assembly lines are a popular class of assembly systems consisting of a series of assembly stations arranged in tandem. Every job (or order) visits all assembly stations in the same sequence and spends the same amount of time (known as the production cycle) at each station. Industries such as aircraft, fire-engine, and automobile assembly have production cycles of a few hours and are labor intensive. In spite of increased automation in such industries, human capital remains the most expensive and important contributor to a flexible production system. In this article we formulate the cross-training problem on a paced assembly line with m stations (mCT). We assume that each worker possesses a number of skills referred to as a skill vector. Our objective is to schedule a set of work orders through the assembly system so as to minimize the size of the required workforce and/or the workforce cross-training costs. We analyze the complexity of mCT and identify polynomially solvable cases. A variety of lower bounds is developed based on optimization techniques. These lower bounds are used to develop a branch and bound algorithm as well as to evaluate our heuristics. A computational experiment reports the performance of all algorithms. Using these algorithms, we examine how the formation of skill vectors affects the workforce size and draw guidelines for cross-training programs in organizations with labor intensive assembly operations.

In this paper, we address the product-family design problem of a firm in a market in which customers choose products based on some measure of product performance. By developing products as a family, the firm can reduce the cost of developing individual product variants due to the reuse of a common product platform. Such a platform, designed in an aggregate-planning phase that precedes the development of individual product variants, is itself expensive to develop. Hence, its costs must be weighted against the benefits of its reuse in a family. We offer a model for capturing costs of product development when the family consists of variants based on a common platform. It is shown that the model can be converted into a network-optimization problem, and the optimal product-family can be identified under fairly general conditions by determining the shortest path of its network formulation. We also analytically examine the effect of alternative product designs on product-family composition, and discuss the implications of investing in new-product technology. Finally, we illustrate our model and managerial insights with an application from the electronics industry.

This paper considers serial production-transportation systems. In recent years, researchers have developed a fairly simple functional equation that characterizes optimal system behavior, under the assumption of constant leadtimes. We show that the equation covers a variety of stochastic-leadtime systems as well. Still, many basic managerial issues remain obscure: When should stock be held at upstream stages? Which system attributes drive overall performance, and how? To address these questions, we develop and analyze several heuristic methods, inspired by observation of common practice and numerical experiments. One of these heuristics yields a bound on the optimal average cost. We also study a set of numerical examples, to gain insight into the nature of the optimal solution and to evaluate the heuristics.

We argue that industries and industry segments are characterized by a clockspeed that gauges the velocity of change in the external business environment and sets the pace of their firms' internal operations. Using data from the electronics industry, we develop and validate an integrated metric for clockspeed that takes into account both demand- and supply-side factors. We show that after controlling for product complexity and other factors, higher industry clockspeed is associated with faster execution in product development and manufacturing (e.g., shorter development time, quicker stabilization of production) and more frequent changes in organizational structure. Our findings on the effects of clockspeed can help researchers studying other industries, and our results provide benchmarks against which practitioners can compare and classify their own organizations.