B.8 For further reading
1.11 FOR FURTHER READING
In this book we will deal with problems which lie at the boundary be- tween distribution logistics and production planning. An excellent book on manufacturing systems, including production planning and control, is [8].
0 An excellent text covering supply chain management with a wider scope (and, necessarily, sometimes a more shallow level) is [ 5 ] . Among other
REFERENCES 51
things. the reader will find there some treatment of re\-enue management and electronic commerce. For a text very rich in references to practical cases. see also [13].
0 TVe deal with distribution logistics from a n operatzons management per- spective. but we should keep in mind th at this dimension must be linked to a financial perspective: models integrating the two sides of the coin are illustrated in [ l a ] .
0 lye have pointed out that there is no best supply chain management approach: the strategy must be adapted to the specific firm and market at hand. a point which is very well illustrated in [6].
0 Readers interested in discrete-event simulation will find [9] very com- prehensive and readable.
0 A tutorial introduction t o stochastic programming models in manufac- turing can be found in [1]. For a comprehensive introduction to both models and solution methods, see. e.g.. [3].
REFERENCES
1. A. Alfieri and P. Brandimarte. Stochastic Programming Models for Man- ufacturing Applications. In A. hIatta and Q. Semeraro. editors. D~szgn of Advancrd Manufacturzng Systems. Springer. Dordrecht. 200.5.
2 C. Billington. G. Callioni, B. Crane, J . D . Ruark. J.U. Rapp. T. TVhite. and S.P. IVillems. Accelerating the Profitability of Hewlett-Packard's Supply Chains. Interfaces. 34:59-72. 2004.
3. J . R . Birge and F. Louveaux. Introductzon to Stochastac Proqrammzng.
Springer-Verlag. Yew Yoik. 1997.
4. P Braiidiniarte. Numerzcal Methods an Fmance and Economacs: A MAT- LAB-Based Introductzon (2nd Ed.). TViley. Yew York. 2006.
5. S. Chopra and P. hleindl. Supply Cham Munagrment. Strategy. Plannmg.
and Operafzon (2nd Ed.). Pearson Prentice Hall, Upper Saddle River. N J , 2004
6. 1I.L. Fisher. TVhat Is the Right Supply Chain for >our Products? Haraard Buszness Revzew, 75:105 -116. 1997.
7. P. Ghemawat and .J.L. Nueno. Zara: Fast Fashzon. case 9-703-497. Har- yard Business School Publishing. Boston. M A . 2003.
8. W. Hopp and M. Spearman. Factory Physics (2nd Ed.). hlcGraw-Hill.
New York. 2000.
9. A.M. Law and D . W . Kelton. Simulation Modeling and Analysis (3rd Ed.).
McGraw-Hill, New York, 1999.
10. H.L. Lee and C. Billingt,on. Material Management in Decentralized Sup- ply Chains. Operations Research, 41:835-847, 1993.
11. A. Raman and Z. Ton. Borders Group Inc., case 9-601-037. Harvard Business School Publishing, Boston, MA, 2003.
12. J.F. Shapiro. Modeling the Supply Chain. Duxbury/Thomson Learning, Pacific Grove, CA, 2001.
13. D. Simchi-Levi,
P.
Kaminsky, and E. Simchi-Levi. Designing and Manag- ing the Suppy Chain (2nd Ed.). IClcGraw-Hill/Irwin, New York, 2002.Network Design and Transp o rt a t i o n
In chapter 1 we have seen t h a t logistic networks can be shaped according t o several patterns: defining the structure of the network is a strategic task with a significant impact on the overall cost of the supply chain. and it resulls in constraints on its day-to-day operations. T h e main problem we deal with in this chapter is indeed the design of logistic networks. Actually. we should speak of network design problems. as there are many shades and nuances of this problem. In principle. designing a logistic network requires locating and sizing production plants. distribution centers, and retail stores. In practice.
we typically face a subset of those decisions. since some part of the network is given. To begin with. we rarely design a network from scratch; we may have t o redesign a n existing network in order to adapt it t o changing demand patterns or changing prices of inputs. Hence, we may have t o relocate some facilities.
to expand their capacities. or t o locate a few new ones. Furthermore. (i) when locating plants or large distribution centers. retail store locations are taken as given: on the contrary. (ii) in retail management we often have t o locate retail stores, i.e.. the last nodes in the network (e.g.. see [5]). The relevant criteria and constraints are quite different in the two problems. V'hen locating retail stores, a n important role is played by the logistic range, i.e.. the maximum distance a potential customer is willing to travel t o purchase a given item:
hence. distance may drive sales rather than just contributing a cost term t o the objective function. When locating a distribution center, the distance between the center and the retail stores is typically just an element t o evaluate the total transportation cost. SforeoTer. in many location problenis we take demand at final destination nodes as exogenously given. O n the contrary.
when locating retail stores. demand is a result of our decisions.
53
The design of a logistic network is typically considered a long-term. strate- gic problem. Indeed. building a large and expensive facility is certainly not a day-to-day decision. Nevertheless, recent trends, whereby third parties may offer logistic services. tend t o make the problem a bit more tactical and shorter-term. Obviously, building a plant and renting shelf space for the next four months are different decisions. In the latter case. we are changing the nature of costs from fixed ones to (relatively) variable ones. Flexibility is a requirement dictated by the faster and faster introduction of new products and ever changing market conditions. which may call for the almost continu- ous redesign of the supply chain. In any case. even if we are making strategic decisions. we need to represent their consequences on tactical decisions. such as transportation optimization. We need a sort of "anticipation function" in order to estimate the costs of tactical decisions that we will make next. subject to constraints enforced by strategic decisions: this estimate need not be overly precise. In strategic models. we cannot take detailed issues. such as opera- tional vehicle routing, into account: such decisions are the subject of chapter 8; by the same token. the optimal loading of a single vehicle is of no concern a t this level. Still, a suitably aggregate representation of transportation flows and their costs is needed when designing a network.
An interesting feature of logistic networks is the presence of intermediate nodes. such as distribution warehouses or transit points, between production plants and retail stores. Since such facilities represent a cost, there must be some good reason to introduce them. We discuss their functions in section 2.1. In particular. we point out their potential role in reducing the impact of demand uncertainty in section 2.1.1. whereas in section 2.1.2 we consider their role in optimizing transportation and in managing assortment.
Section 2.2 deals with classical linear programming models t o optimize transportation flows on a network, to locate facilities, and t o choose their ca- pacities. To keep computational effort limited, these models are static rather than dynamic, and we should wonder if such models are able to capture the interaction of flow routing and inventory management decisions. We cannot and should not mix detailed descriptions of both strategic and operational decisions in the same model: however, a suitable approximate model may be obtained by considering nonlinear cost functions. Then. t o avoid the burden of solving a large nonlinear mixed-integer programming model. we may approx- imate nonlinear costs by piecewise-linear functions, as described in section 2.3. Since some model formulations may be tough to solve. a huge amount of literature has been produced, based on heuristic approaches t o ease the computational burden or to make the solution process a bit more intuitive.
We will not consider this literature, for which we point out a few references at the end of the chapter; by the same token, we refrain from describing complex models accounting for some additional issues. Indeed, the astonishing progress in both hardware and optimization software libraries has paved the way t o the solution of large scale models. We believe that the main limitations of the modeling framework we describe here are not computational but, instead.
T H E ROLE OF INTERMED/ATE NODES IN A DISTRIBUTION NETWORK 55
lie in their limited ability to cope with denland uncertainty. as well as in the potential difficulty in understanding why we have obtained a certain optimal solution. Indeed. we should consider the models below as one tool within a complex decision support architecture: their role is to propose solutions.
which could be modified in order to comply with some further requirements and should be thoroughly checked by detailed simulation.
The chapter is complemented by two web sections. Section i I 7 . 2 . 4 deals with continuous-space location models. In fact. the previous sections assume that we have already identified sites for potential facilities. and we must make a choice between a discrete set of alternatives: in other cases. we would like t o find ideal positions of facilities. in continuous space. This ma!- be useful in the process of building alternatives. Section LV.2.5 illustrates peculiarities of retail store location problems. compared with plant and distribution center location models. This topic is usually covered in books on marketing rather than in books on logistics. Lye believe it is actually a borderline issue as it defines the “last mile” (i.e.% the last echelon) of the supply chain for consumer goods.
2.1 T H E ROLE OF INTERMEDIATE NODES IN A DISTRIBUTION