DIRECT DISTRIBUTION OR STAGED DISTRIBUTION?

The last three decades have been marked by strong transformations in supply relations between industry and retail. There are several reports on these different supply relationships (Christopher, 2000). Several industries have restructured their distribution networks to meet the continued retail demand for lower inventories and higher service levels (Hoek, 1998a and 1998b). The management initiatives that culminated in the restructuring of these operations are known by different names: Efficient Consumer Response – ECR, Quick Response – QR (Fiorito and May, 1995), Vendor Managed Inventory – VMI (Waller and Johnson, 1999), Continuous Replenishment – ​​CR (Ellinger and Taylor, 1999) and Continuous Replenishment Program – CRP (Liz, 1999 and Andraski, 1994).

In general lines, these initiatives are based on a common objective: to reduce dependence on sales forecasts in the industry and inventory levels in retail, with a simultaneous increase in service levels (Lee and Padmanabhan, 1997). According to Vergin and Barr (1999), cooperation and sharing of information on final consumer demand would allow for this objective (Kiely, 1998). However, the literature is not conclusive with regard to the homogeneity of the results achieved from the perspectives of industry and retail. While in retail the restructuring of operations in the industry allowed the reduction of inventories, in industry there were cases of reduction and increase of the same (Harrison and Voss, 1990 and Romero, 1991).

It is possible that the benefits “promised” by these initiatives regarding inventory levels (and also total costs) in the industry depend on the adherence between the strategic positioning of its distribution network, the characteristics of the product and demand and what is expected from retail performance indicators (Johnson and Stice, 1993 and Jones, 1991).

Wanke and Zinn (2004) provide an example in this regard from the VMI initiative, which can be used to illustrate the choice between centralization and decentralization of distribution under different trade-offs between inventory turnover in industry and delivery time required by retailers. . At VMI, industry stock can be located in a state-of-the-art distribution center and delivery to retailers is virtually instantaneous.

The key question is how the industry should replenish retail inventory. The choices are an industry-centric inventory combined with a longer lead time for retail (and lower turnover) or a decentralized inventory, close to retail, with faster delivery (and higher turnover). In essence, what is being sought is to answer whether this management initiative (VMI) should be structured via direct distribution (stock centralized in the industry or in a single distribution center) or via staggered distribution (stock decentralized in a local distribution center). What is the most suitable type of distribution?

Qualitatively, the impact of distribution type on key performance indicators in industry and retail is relatively well documented, and empirical evidence on the direction of the main effects of direct distribution and staggered distribution is diverse (Evers, 1999; Leew et al. Goor, 1999; Evers and Beier, 1998; Tallon, 1993; Amstel and Amstel, 1985).

Some of these impacts have already been mentioned in books over twenty years ago. For example, Bowersox et al. (1980) state that staggered distribution implies higher inventory levels for the industry, being preferable when the products have low added cost and there is the possibility of consolidating transport between the industry and the distribution center (Jayaraman, 1998 and Carter and Ferrin, 1996). Direct distribution from the industry tends to occur with high added cost products, especially if the volumes are high and there is proximity to retail (Bowersox and Closs, 1996). The high added cost can also inhibit intermediaries interested in keeping stocks, leading the industry to direct distribution to the final consumer (Lambert and Stock, 1998).

According to Levy and Weitz (1998), the choice of the type of retail distribution must simultaneously consider the total cost associated with each alternative and customer service, that is, having the product in the store when the final consumer wants to buy it. . Staggered distribution allows retailers to operate with less inventory as a result of more frequent deliveries from the distribution center. In addition, a better balance between leftovers and shortages can result from reviewing, whenever necessary, the quantities requested from the distribution center (Berman and Evans, 1998).

Direct distribution, because it takes retail time to receive and process orders, can lead to less frequent resupply and consolidation of shipments. According to Levy and Weitz (1998), direct retail distribution is also favored by geographic proximity.

It can be seen that, from the perspectives of industry and retail, the choice of distribution type is indifferent when considering the criteria distance between origin and destination and volume of purchases: greater distances and smaller volumes, staggered distribution with consolidation via distribution center; shorter distances and higher volumes, direct distribution from industry to retail. When the analysis criterion is the stock level in industry and retail, staggered distribution implies higher stock levels for the first and lower stock levels for the second, and direct distribution, vice versa.

THE NETWORK UNDER ANALYSIS

The question that arises now is, considering the prism of the industry, under what conditions direct distribution and staggered distribution could be used in a complementary way, and not exclusive, to serve the markets and in what proportions. More specifically, we want to answer how these proportions could be adjusted to minimize total costs from a better balance between safety stock levels (greater in staggered distribution) and transportation expenses (greater in direct distribution). What tools (methodologies, models and systems) are available to help the industry evaluate this decision?

The answer to these questions will take as a starting point a simple network, composed of a factory, a Central DC, a Local DC and a universe of customers to be served (market). These customers are offered a distribution policy that is quite common in Brazilian companies: receive CIF directly from the Central CD with a longer lead time or withdraw FOB from the Local CD with a shorter lead time. For the industry, this distribution policy boils down to: spending more on transport in direct CD Central-Mercado distribution and saving on safety stocks; or spend less on staggered distribution, just transfer the Central DC to the Local DC, and lock up capital in safety stocks at the Local DC.

Figure 1 not only illustrates this simple logistic network, but also presents the main variables involved in the decision to determine the optimal proportion between direct distribution (Wc) and staggered distribution (W1). It should be noted that W1 and Wc must add up to 100%.

ALTERNATIVE METHODOLOGIES TO SOLVE THE PROBLEM

There are two approaches to trying to solve this problem. The most classic and well-known is associated with mathematical programming techniques in the field of operational research, emphasizing transport and installation costs and relegating inventory costs to the background. More recent works such as those by Jayaraman (1998) and Croxton and Zinn (2005)* have attempted to establish bridges between mathematical programming models and inventory levels, even though disregarding the demand correlation between the branches or legs of the logistics network.

The less widespread, but suitable for resizing already existing logistics networks, is derived from the different ways of calculating the portfolio effect, that is, the expected percentage of reduction in safety stocks from the centralization of service. In this approach, transport costs are relegated to the background and predominate analyzes of how the correlation coefficients between the different legs of the logistics network can contribute to the consolidation of safety stocks in a smaller number of facilities. Mahmoud (1992) tried to establish a bridge between the reduction of safety stocks and the impact on other components of the logistics system, such as, for example, total costs.

This description is not intended to exhaust all the particularities of these two alternative methodologies, but to draw attention to the fact that they can indicate conflicting results, even when considering a relatively simple network like the one shown in Figure 1.

For example, Figure 2 illustrates the typical solution pattern for the mathematical programming models when considering that the Central DC and the Local DC must remain open and that the cost of transport via the Local DC (part of the path is FOB) is necessarily lower than the cost of transport from CD Central (CIF). All distribution would have to be done by the Local CD (W1 = 100%) and there would be no room for direct distribution. But at what cost of holding safety stocks on both DCs?

Figure 3 shows the typical decision pattern when analyzing the problem from the perspective of the portfolio effect. The duplicity of safety stocks in both DCs is the crux of the matter, which would lead to the consolidation of safety stocks in the DC closest to the source of supply, that is, it would lead to the centralization of stocks in the Central DC that is closer to the factory. In other words, W1 = 0% and consequently Wc = 100%. But at what level of spending on transport?

As illustrated in Figure 4, one sees the need to integrate both approaches in order to obtain more satisfactory answers in terms of distribution planning. Not just direct distribution, not just staggered distribution, but a mix between these two types. Several researchers at universities and consulting companies have been continuously working towards a greater integration of these approaches into a single methodology, which, if achieved, could generate significant benefits in terms of saving human and material resources spent on these issues, also implying significant evolution in the way distribution networks are planned today.

Figure 4 shows some of the elements needed to make this integration possible, even if applied to the simple network discussed in this article. They are: (a) the determination of the total cost equation that contains the main trade-off of the analysis (in this case, the expenses with transports versus the safety stock levels in the network) and (b) the determination of how it behaves the correlation coefficient between Local DC and market demands for Central DC.

Demand from the Local DC to the Central DC depends, among other variables, on cycle stock coverage (COB), in terms of days of consumption, that the commercial or marketing area deems reasonable to keep at the Local DC. From time to time, this coverage is consumed by market service and the need for a replacement lot, proportional to this coverage, is generated for the Central DC.

The following section presents the analytical result for this problem, that is, the formula to determine the optimal value of W1, in addition to sensitivity analyzes with the main parameters for decision making. Before proceeding, we highlight the main assumptions adopted to derive this result: the Central DC is equidistant from the factory and the market, safety stocks are determined by the service factor k of normal distribution, and orders are placed electronically from the Local DC to the Central DC, generating negligible fixed costs related to batch resupply. In addition, Local DC coverage is a market variable to be defined by the commercial area.

MAIN RESULTS

Figure 5 presents the formula for the optimal percentage of staggered distribution, given the simple network shown in Figure 1 and the assumptions previously indicated. The bars around the W1 Optimum formula indicate that the modulus of the expression is being considered and it should be remembered that there are no real solutions (negative root) when –a2*a42+a12< 0.

Figures 6 and 7 present sensitivity analyzes for the main variables of the problem, considering a set of real data, with the coverage of the Local DC defined by the commercial area. Below is a brief discussion of the impact of each of these variables on the optimal percentage of staggered distribution (optimal W1).

• Demand variation coefficient (CVd). As illustrated in Figure 6, the coefficient of variation of demand presents the “most interesting” trade-off with respect to the optimal percentage of staggered distribution. The greater this coefficient, that is, the greater the demand uncertainty, the smaller the percentage of staggered distribution and the greater the percentage of direct distribution. It just happens after a certain point. Small values ​​of the demand variation coefficient (up to 0,5) favor staggered distribution. These values ​​indicate the extent to which it is worthwhile to keep safety stocks at an intermediate facility to save on transportation costs with direct delivery.
• Unit cost of holding inventories (Ci). As it should be, the higher the costs of holding stocks, the lower the propensity for staggered distribution.
• Direct distribution unit costs per day (Cd) and staggered distribution unit costs per day (Ce). Like the previous trade-off, these are also “easy” to interpret. Higher direct distribution costs, higher percentage of staggered distribution. On the other hand, the higher the staggered distribution costs, the lower the staggered distribution percentage.
• Ratio between lead-time variances (y). The greater the uncertainty of the Central DC in serving the market compared to its service to the Local DC, the greater the percentage of staggered distribution should be, since the safety stocks necessary to ensure a certain level of service will be proportionally smaller in the Local DC than in the Central CD.
• Ratio between lead-time averages (x). The closer the Local DC is to the Central DC and the more distant both DCs are from the market, the Local DC loses importance as an element for minimizing total costs and, consequently, the lower the percentage of staggered distribution. A curious fact is that in extreme situations and keeping everything else constant, the values ​​of W1 and Wc converge, respectively, to 40% and 60%, that is, 40% for FOB delivery via Local DC and 60% for direct delivery to from CD Central. In other words, for Local DCs that are “satellites” of a Central DC far from the market, an optimal floor of approximately 40% would be verified for the analyzed data.
• Service factor (k). The higher the desired level of service in terms of product availability, the lower the percentage of staggered distribution, basically due to two reasons: (a) safety stocks start to become excessive in the distribution network in relation to transportation expenses and (b), from the point of view of service level, the role of a Local DC is to contribute, above all, with the dimension of delivery time, and not with the dimension of product availability (which depends on the k factor).

CONCLUSION

In summary, the results show that the coexistence of direct distribution with staggered distribution may be the best policy to be adopted by industries most of the time. Two variables are worth mentioning: the demand variation coefficient and the proximity between the Local DC and the Central DC in relation to the distance of the Central DC from the market. Regarding the first variable, a “watershed” is established on what would be a high or low demand uncertainty: for demand variation coefficient values ​​around 0,5, the optimal percentage of staggered distribution stops increasing (CVd < 0,5) and starts to systematically decrease (CVd > 0,5). Regarding the second variable, the percentage of staggered distribution is strongly influenced by whether the Local DC is a satellite installation to the Central CD or to the market, with this percentage being greater the closer the Local DC is to the market.

In this article, the importance of developing new approaches for planning the type of distribution, whether direct or staggered, and what is the optimal percentage of both, was discussed. For this purpose, the perspective of industry costs and a simple distribution network in which several variables were considered simultaneously were considered. The analytical solution for the optimal percentage of staggered distribution was determined and several sensitivity analyzes were conducted, with the objective of highlighting the relative importance of the main variables of the problem and understanding the trade-offs involved in decision making.

BIBLIOGRAPHY

AMSTEL, M.; AMSTEL, W.. 1985, “Economic Trade-offs in Physical Distribution”, Intl. Journal of Physical Distribution & Materials Management, v. 17, n.7, pp.15-54.

ANDRASKI, J.. 1994, “Foundations for Successful Continuous Replenishment Programs”, Intl. Journal of Logistics Management, v. 5, no. 1, pp. 1-8.

BERMAN, B.; EVANS, J.. 1998, Retail Management: A Strategic Approach. Prentice Hall.

BOWERSOX, DJ; COOPER, M.; LAMBERT, D.; TAYLOR, D.. 1980, Management in Marketing Channels. 1st ed., New York, McGraw-Hill.

BOWERSOX, DJ; CLOSS, DJ. 1996, Logistical Management – ​​The Integrated Supply Chain Process, 1st ed., McGraw-Hill.

CARTER, JR; FERRIN, BG. 1996, “Transportation Costs and Inventory Management: Why Transportation Cost Matter”, P&IM Journal, Third Quarter, pp. 58-62.

CHRISTOPHER, M.. 2000, “The Agile Supply Chain – Competing in Volatile Markets”, Industrial Marketing Management, v. 29, no. 1, pp. 37-44.

CROXTON, KL; ZINN, W.. 2005, “Inventory considerations in network design”. Journal of Business Logistics, vol. 26, no. 1, pp.149-168.

ELLINGER, A.; TAYLOR, J.. 1999, “Automatic Replenishment Programs and Level of Involvement: Performance Implications”, Intl. Journal of Logistics Management, pp. 25-36.

EVERS, PT; BEIER, FJ. 1998, “Operational Aspects of Inventory Consolidation Decision Making”, Journal of Business Logistics, v. 19, no. 1, pp. 173-189.

EVERY, PT. 1999, “The Effect of Lead Times on Safety Stocks”, P&IM Journal, Second Quarter, v. 40, no. 2, pp. 6-10.

FIORITO, SS; MAY, EG. 1995, “Quick Response in Retailing”, Intl. Journal of Retail and Distribution Management, v. 23, no. 5, pp. 12-21.

HARRISON, A.; VOSS, C.. 1990, “Issues in setting up JIT supply”, Intl. Journal of Operations and Production Management, v. 10, no. 2, pp. 84-93.

HOEK, RI. 1998a, “Logistics and Virtual Integration”, Intl. Journal of Physical Distribution & Logistics Management, v. 28, no. 7, pp. 508-523.

HOEK, RI. 1998b, “Reconfiguring the Supply Chain to Implement Postponed Manufacturing”, Intl. Journal of Logistics Management, v. 9, no. 1, pp. 95-110.

JAYARAMAN, V.. 1998, “Transportation, Facility Location and Inventory Issues in Distribution Network Design”, Intl. Journal of Operations & Production Management, v. 18, no. 5, pp. 471-494.

JOHNSON, GH; STICE, JD. 1993, “Not Quite Just-in-Time Inventories”, The National Public Accountant, v. 38, no. 3, pp. 26-29.

JONES, DJ, 1991, “JIT and EOQ Model: Odd Couple No More”, Management Accounting, v. 72, no. 8, pp. 54-57.

KIELY, DA. 1998, “Synchronizing Supply Chain Operations with Consumer Demand Using Customer Data”, The Journal of Business Forecasting Methods & Systems, Winter, pp.3-9.

LAMBERT, DM; STOCK, JR; ELLRAM, LM. 1998, Fundamentals of Logistics Management. 1st ed., New York, Irwin-McGraw Hill.

LEE, HL; PADMANABHAN, V.. 1997, “The Bullwhip Effect in Supply Chains”, Sloan Management Review, Spring, pp. 93-102.

LEEUW, SD; GOOR, AR. 1999, “The Selection of Distribution Control Techniques”, The Intl. Journal of Logistics Management, v. 10, no. 1, pp. 97-112.

LEVY, M.; WEITZ, B.. 1998, Retailing Management. 3rd ed., New York, McGraw Hill.

LIZ, P.. 1999, “CRP Investment Pays Off in Many Ways”, Drug Store News, v. 21, no. 2, p.26.

MAHMOUD, MM. 1992, “Optimal inventory consolidation schemes: a portfolio effects analysis”. Journal of Business Logistics, vol. 13, no. 1, pp.193-214.

ROMERO, B.. 1991, “The Other Side of Supply Management”, P&IM Journal, v. 32, no. 4, pp. 1-3.

TALLON, W.. 1993, “The Impact of Inventory Centralization on Aggregate Safety Stock”, Journal of Business Logistics, v. 14, no. 1, pp. 185-196.

VERGIN, R; BARR, K.. 1999, “Building Competitiveness in the Grocery Supply Chain Through Continuous Replenishment Planning”, Industrial Marketing Management, v. 28, pp. 145-153.

WALLER, M.; JOHNSON, M.. 1999, “Vendor Managed Inventory in the Retail Supply Chain”, Journal of Business Logistics, v.20, n.1, pp. 183-198.

WANKE, P.; ZINN, W.. 2004, “Strategic Logistics Decision Making”, Intl. Journal of Physical Distribution and Logistics Management, v. 34, no. 6, pp. 466-478.
*See the article by Walter Zinn and Keely Croxton “Considerations about inventory in the design of logistics networks”, 1st and 2nd parts, in editions 133 (December 2006) and 134 (January 2007) of Tecnologística – now also available on the website: www.tecnologistica.com.br/site/lista_anteriores.asp