Supply Network Design for Manufacturing Industry Pranav Rajvanshy,
[email protected] Student - One Year Full Time MBA (PGPX), IIM, Ahmedabad Vivekanand B Khanapuri,
[email protected] Associate Professor, NITIE, Mumbai
Abstract Purpose Supply chain is a key strategy area in multiple industries including among others manufacturing, FMCG, retail and logistics sector. An efficient supply chain is not only fast in responding to market dynamics but also extracts cost redundancies out of business. Companies incur substantial costs in form of COGS at various stages of value chain starting from raw material procurement to delivery of Finished Goods to distributors. This paper is an attempt towards designing a cost effective and responsive supply network of vendors, factories, warehouses and distributors in manufacturing / FMCG sector. The associated costs of the entire supply network could be as high as 80%-90% of overall revenue of company and, therefore, represent a prime area for cost optimization. Methodology We present here a comprehensive and quantitative approach (using Operations Research model) to design a supply network, which minimizes the overall cost of serving the distributors. Key cost elements that have been considered include procurement costs (raw material and packing material), manufacturing costs (labour, power, fuel etc.), warehousing costs (storage and handling at warehouses) and finished goods distribution costs (factory to warehouse; warehouse to distributor; factory to distributor). Unlike in the conventional distribution system, emphasis has been also laid on identifying direct dispatch potential i.e. serving the distributors directly from factory and thereby reducing the associated warehousing costs. Workings have been shown for a simplified network to keep the data manageable. This model can, however, be easily scaled up for real life situations. Similarly, while the model can also be expanded to find out cost optimal green field factory and warehouse locations; for want of building the basic concept of efficient design such scenarios have been kept out of scope of current study. Findings The approach comes up with a supply network that serves the customers at minimum cost. Key deliverables include cost optimal production schedules at factories; primary and secondary freight lanes selection and dispatch schedules (quantity and frequency) from factories to either warehouses or distributors.
1
Conclusion An efficient supply network gives the company an enduring competitive advantage in terms of cost to serve and service levels. The study reflects in quantitative terms the benefits that could accrue by redesigning the product flow across the supply network. Key Words: Supply chain management, Cost optimal network design, Direct dispatch Introduction Cost optimal supply networks are a source of competitive advantage to manufacturing companies. This holds especially for those firms that have a wide array of products consumed by a sizeable customer base spread over a large geography. Network design thus becomes a critical aspect of the overall business strategy of such companies. While the design problem could be geared either towards cost or service leadership, the manufacturing companies focus mostly on optimizing their network’s overall cost. Focus on cost optimization has led to various innovations in network design and operation. One of such innovations is direct sales from factories to distributors and has been focused upon in this paper. Supply networks of most of the manufacturing companies especially those from FMCG (Fast Moving Consumer Goods) sector are composed of product flow through multiple echelons that include stakeholders such as raw material suppliers, production plants, warehouses and distribution points. These stakeholders are spread out geographically resulting in multiple supply network nodes (factories, warehouses and distributors) and network linkages (transportation lanes). Such large supply networks not only present challenges towards optimal design but also towards efficient execution of day-to-day operational activities. For example, production planning and dispatch scheduling decisions for the entire product line of companies are taken on a daily basis to ensure a time responsive and cost optimal product flow from factories to distributors via warehouses. With increasing competition, rising inflation and relatively slower economic growth many companies have intensified their focus on designing more cost optimal and innovative supply structures. A large number of initiatives have so far focused on benchmarking best business practices of Industry; identifying cost optimal production centers and warehouses; availing tax incentives offered by governments; reducing inventory levels in the system etc. However, still most of the companies have a rather ‘sequential’ theme to their supply chains i.e. finished goods flowing from factories to warehouses to distributors. A major area in network design is, therefore, to rethink the traditional roles of various stakeholders and thereby develop innovative solutions aimed at reducing supply cost structures and time taken to cater to market demand. A typical supply network design problem could be classified under two categories: 1. Defining product flow through existing network to minimize distribution cost; 2. Optimizing the existing network of facilities to minimize distribution cost. The main contribution of this study is that it looks at both the aforesaid categories and integrates the concept of direct sales from factories in the overall network design. Since the problem involves selection (opening / closure) of production lines and factories, we formulate a mixed integer programming model. The model minimizes total distribution cost through selection of cost competitive production centers; [VK1]identification of optimal routes for both primary and secondary distribution; and exploration 2
of direct dispatch opportunities offered by a vast network of production facilities. In formulating the model, locations of warehouses and distributors have been considered fixed in this study. Literature Review Pursuit of cost excellence and on time performance has led to increased collaboration amongst various actors of a supply network (Chopra and Meindl 2004). Firms have been increasingly focusing on minimizing the costs related to their entire distribution network. Thus, locationallocation problems in a multi echelon and multi facility supply network have been researched intensively in last few years. A typical facility location problem is about taking simultaneous decisions regarding design and control of product flow in a generic distribution network (Manzini et al. 2006). Facility location problems can be further classified under two heads – Single facility location problems – Such problems consider cases where a single facility is to be located to serve a set of customers. Young and Hwan (2003) show the working for one such case. Multiple facility location problems – These problems consider cases where a set of potential locations are to be assessed for setting up a certain number of facilities to serve customers. Ghiani et al. (2002) have presented an example on such problems. Network design problems with multiple echelons and production facilities fall under the ambit of multiple facility location problems. Ambrosino and Scutella (2005) proposed an integrated distribution network model that integrates stakeholders both vertically and horizontally to optimize distribution costs. Crainic and Laporte (1997) discussed various Operations Research models to resolve issues in transportation planning and network design. Beamon (1998) put forward deterministic and stochastic models in area of multi stage supply chain design analysis with decisions being made regarding location, allocation, routing and inventory to meet customer demand at minimal cost. Tiwari et al. (2012) proposed a heuristic approach based on HOT (Highly Optimized Tolerance) algorithm for a single-source, multi-product, multi-stage supply chain network design problem. This approach minimizes the cost of supply chain distribution in which products flow sequentially from factories to distributors. Lee et al. (2010) suggested an approach to combine location-allocation problem and multi depot vehicle routing problem with an objective to minimize overall distribution costs. They suggested a heuristic algorithm with LP relaxation to solve the combined designed problem of location-allocation and routing. Wu et al. (2002) proposed a decomposition-based method for solving the location routing problem with multiple depots, multiple fleet types, and limited number of vehicles for each different vehicle type. The location-routing problem has been worked out for a two echelon network (including warehouses and distribution centres) to find the optimal number and locations of the distribution centres along with the vehicle schedules and distribution routes to minimize total system costs. Riccardo and Elisa (2008) proposed models to design multi-period and multi-stage location allocation problems. The key element amongst all these problems has been to optimize the cost by designing the network of facilities (production plants, warehouses and distributor points) and defining the product flow from source to demand centers. The product flow has been considered to be vertically sequential i.e. from one stage of supply network to the next stage. Optimization efforts in these models have so far largely focused on three main areas – 1) Identifying cost optimal production centers; 2) Optimizing primary distribution network involving factories and 3
warehouses; and 3) Optimizing the secondary distribution network involving warehouses and distributors. In this paper we formulate a mixed integer programming model to optimize overall cost along with integration of direct dispatches in network design. We extend the model to cover opportunities of supplying the customers directly from the production centres and thus bypassing the intermediate warehousing stage. The Model The mixed integer programming model presented here is an example of Location Allocation Problem for a multi-product, multi-facility supply network with three stages including factories, warehouses and distribution centres. One of the major design questions for companies with growing business is to simultaneously identify new factory locations along with cost optimization of existing network. The problem statement for our current paper, thus, involves making following decisions – 1. Number and location of factories – Optimal number and location of factories with sufficient capacity to meet customer demand. The model chooses existing from production sites (new production sites can be added to the choice set of locations without any impact on model) 2. Number of production lines per factory – Different production lines could be installed for different products on the basis of manufacturing process and technology requirements. Decision is to be made on number of lines (and hence the type of products) that could be operated in a factory. 3. Allocating warehouse demand to factories – Least cost allocation of warehouse and SKU (stock keeping unit) level demand to factories. 4. Direct dispatch (sales) from factories to distribution centres – Identification of factoryproduct-distribution centre combinations where direct dispatch is feasible and cost effective. Warehouse locations have been considered fixed in this model. The major cost components that have been considered are – Raw material costs, manufacturing costs (this could include labor and energy costs), loading and unloading costs at factories and warehouses, handling cost at warehouses, primary distribution costs and secondary distribution costs. Additionally, taxation costs (such as excise duties) and incentives (such as government subsidies) can be included in the overall conversion costs. Conversion cost represents raw material costs, manufacturing costs, taxation costs and subsidies. Cost benefit analysis for direct dispatches has been limited to warehouse handling costs; primary freight and secondary freight, since direct dispatches are opportunity based and may not necessarily result in reduction of warehousing space. Thus, reduction (if any) in warehouse space due to direct dispatches has not been considered. Distributors have been mapped to the closest warehouse as this approach minimizes the secondary distribution cost. The major assumptions of the model are listed below 1. Average quantity per shipment is assumed to be 5 tons i.e. a full truck load is assumed to be equivalent to 5 tons 2. Direct dispatch for a given factory-distributor combination is assumed to be feasible only when a a minimum of one shipment (of 5 tons) per week could be formed
4
3. Material handling cost (i.e loading / unloading) factories and warehouses have been considered same for all SKUs 4. Primary and secondary freight costs are assumed to be not impacted by type of SKU 5. Total number of annual shifts per line at a factory have been assumed to be 900. This has been calculated as 3 shifts/day * 25 working days/month * 12 months. Notations Used Following notations are being introduced to formulate the problem Factory specific variables 𝐹𝐹𝐶𝑓𝑙 = 𝐹𝑖𝑥𝑒𝑑 𝑐𝑜𝑠𝑡 𝑜𝑓 𝑙𝑖𝑛𝑒 𝑙 𝑎𝑡 𝑓𝑎𝑐𝑡𝑜𝑟𝑦 𝑓 𝐿𝐶𝐹𝑓 = 𝐿𝑜𝑎𝑑𝑖𝑛𝑔 𝐶𝑜𝑠𝑡 𝑝𝑒𝑟 𝑡𝑜𝑛 𝑎𝑡 𝐹𝑎𝑐𝑡𝑜𝑟𝑦 𝑓 𝑅𝑀𝑠𝑓 = 𝑅𝑎𝑤 𝑚𝑎𝑡𝑒𝑟𝑖𝑎𝑙 𝑐𝑜𝑠𝑡 𝑝𝑒𝑟 𝑡𝑜𝑛 𝑜𝑓 𝑆𝐾𝑈 𝑠 𝑎𝑡 𝐹𝑎𝑐𝑡𝑜𝑟𝑦 𝑓 𝐶𝐶𝑠𝑓𝑙 = 𝐶𝑜𝑛𝑣𝑒𝑟𝑠𝑖𝑜𝑛 𝑐𝑜𝑠𝑡 𝑝𝑒𝑟 𝑡𝑜𝑛 𝑜𝑓 𝑆𝐾𝑈 𝑠 𝑜𝑛 𝑙𝑖𝑛𝑒 𝑙 𝑜𝑓 𝐹𝑎𝑐𝑡𝑜𝑟𝑦 𝑓 𝐹𝐿𝑜𝑝𝑒𝑛𝑓𝑙 = 𝐼𝑛𝑡𝑒𝑔𝑒𝑟 𝑣𝑎𝑟𝑖𝑎𝑏𝑙𝑒 𝑓𝑜𝑟 𝑙𝑖𝑛𝑒 𝑙 𝑜𝑓 𝑓𝑎𝑐𝑡𝑜𝑟𝑦 𝑓 𝑆𝑂𝑠𝑓𝑙 = 𝑆ℎ𝑖𝑓𝑡 𝑜𝑢𝑡𝑝𝑢𝑡 𝑖𝑛 𝑡𝑜𝑛𝑠 𝑜𝑓 𝑆𝐾𝑈 𝑠 𝑎𝑡 𝑙𝑖𝑛𝑒 𝑙 𝑜𝑓 𝑓𝑎𝑐𝑡𝑜𝑟𝑦 𝑓 Warehouse specific variables 𝐿𝐶𝑊𝑤 = 𝐿𝑜𝑎𝑑𝑖𝑛𝑔 𝐶𝑜𝑠𝑡 𝑝𝑒𝑟 𝑡𝑜𝑛 𝑎𝑡 𝑊𝑎𝑟𝑒ℎ𝑜𝑢𝑠𝑒 𝑤 𝑈𝐶𝑊𝑤 = 𝑈𝑛𝑙𝑜𝑎𝑑𝑖𝑛𝑔 𝐶𝑜𝑠𝑡 𝑝𝑒𝑟 𝑡𝑜𝑛 𝑎𝑡 𝑊𝑎𝑟𝑒ℎ𝑜𝑢𝑠𝑒 𝑤 𝐻𝐶𝑊𝑤 = 𝐻𝑎𝑛𝑑𝑙𝑖𝑛𝑔 𝐶𝑜𝑠𝑡 𝑝𝑒𝑟 𝑡𝑜𝑛 𝑎𝑡 𝑊𝑎𝑟𝑒ℎ𝑜𝑢𝑠𝑒 𝑤 Distribution specific variables 𝐹𝐹𝑊𝑓𝑤 = 𝑃𝑟𝑖𝑚𝑎𝑟𝑦 𝐹𝑟𝑒𝑖𝑔ℎ𝑡 𝑝𝑒𝑟 𝑡𝑜𝑛 𝑓𝑟𝑜𝑚 𝐹𝑎𝑐𝑡𝑜𝑟𝑦 𝑓 𝑡𝑜 𝑊𝑎𝑟𝑒ℎ𝑜𝑢𝑠𝑒 𝑤 𝐹𝑊𝐷𝑤𝑑 = 𝑆𝑒𝑐𝑜𝑛𝑑𝑎𝑟𝑦 𝐹𝑟𝑒𝑖𝑔ℎ𝑡 𝑝𝑒𝑟 𝑡𝑜𝑛 𝑓𝑟𝑜𝑚 𝑊𝑎𝑟𝑒ℎ𝑜𝑢𝑠𝑒 𝑤 𝑡𝑜 𝐷𝑖𝑠𝑡𝑟𝑖𝑏𝑢𝑡𝑜𝑟 𝑑 𝐹𝐹𝐷𝑓𝑑 = 𝑆𝑒𝑐𝑜𝑛𝑑𝑎𝑟𝑦 𝐹𝑟𝑒𝑖𝑔ℎ𝑡 𝑝𝑒𝑟 𝑡𝑜𝑛 𝑓𝑟𝑜𝑚 𝐹𝑎𝑐𝑡𝑜𝑟𝑦 𝑓 𝑡𝑜 𝐷𝑖𝑠𝑡𝑟𝑖𝑏𝑢𝑡𝑜𝑟 𝑑 Demand and supply specific variables
5
𝑋𝑠𝑓𝑙𝑤𝑑 = 𝑄𝑢𝑎𝑛𝑡𝑖𝑡𝑦 𝑖𝑛 𝑡𝑜𝑛 𝑜𝑓 𝑆𝐾𝑈 𝑠 𝑝𝑟𝑜𝑑𝑢𝑐𝑒𝑑 𝑎𝑡 𝐿𝑖𝑛𝑒 𝑙 𝑜𝑓 𝐹𝑎𝑐𝑡𝑜𝑟𝑦 𝑓 𝑎𝑛𝑑 𝑑𝑖𝑠𝑝𝑎𝑡𝑐ℎ𝑒𝑑 𝑡𝑜 𝑑𝑒𝑎𝑙𝑒𝑟 𝑑 𝑣𝑖𝑎 𝑤𝑎𝑟𝑒ℎ𝑜𝑢𝑠𝑒 𝑤 𝑌𝑠𝑓𝑙𝑑 = 𝑄𝑢𝑎𝑛𝑡𝑖𝑡𝑦 𝑖𝑛 𝑡𝑜𝑛 𝑜𝑓 𝑆𝐾𝑈 𝑠 𝑝𝑟𝑜𝑑𝑢𝑐𝑒𝑑 𝑎𝑡 𝐿𝑖𝑛𝑒 𝑙 𝑜𝑓 𝐹𝑎𝑐𝑡𝑜𝑟𝑦 𝑓 𝑎𝑛𝑑 𝑑𝑖𝑠𝑝𝑎𝑡𝑐ℎ𝑒𝑑 𝑡𝑜 𝑑𝑒𝑎𝑙𝑒𝑟 𝑑 𝑑𝑖𝑟𝑒𝑐𝑡𝑙𝑦 𝑆𝐴𝑠𝑑 = 𝐴𝑛𝑛𝑢𝑎𝑙 𝑠𝑎𝑙𝑒𝑠 𝑖𝑛 𝑡𝑜𝑛𝑠 𝑜𝑓 𝑆𝐾𝑈 𝑠 𝑎𝑡 𝑑𝑖𝑠𝑡𝑟𝑖𝑏𝑢𝑡𝑜𝑟 𝑑 Optimization Function 𝑀𝑖𝑛𝑖𝑚𝑖𝑧𝑒 𝐷𝑖𝑠𝑡𝑟𝑖𝑏𝑢𝑡𝑖𝑜𝑛 𝐶𝑜𝑠𝑡 = ∑ ∑ 𝐹𝐿𝑜𝑝𝑒𝑛𝑓𝑙 ∗ 𝐹𝐹𝐶𝑓𝑙 𝑓
𝑙
+ ∑ ∑ ∑ ∑ ∑(𝑋𝑠𝑓𝑙𝑤𝑑 + 𝑌𝑠𝑓𝑙𝑑 ) ∗ (𝑅𝑀𝑠𝑓 + 𝐶𝐶𝑠𝑓𝑙 + 𝐿𝐶𝐹𝑓 ) 𝑠
𝑓
𝑙
𝑤
𝑑
+ ∑ ∑ ∑ ∑ ∑ 𝑋𝑠𝑓𝑙𝑤𝑑 ∗ (𝐹𝐹𝑊𝑓𝑤 + 𝑈𝐶𝑊𝑤 + 𝐻𝐶𝑊𝑤 + 𝐿𝐶𝑊𝑤 + 𝐹𝑊𝐷𝑤𝑑 ) 𝑠
𝑓
𝑙
𝑤
𝑑
+ ∑ ∑ ∑ ∑ 𝑌𝑠𝑓𝑙𝑑 ∗ 𝐹𝐹𝐷𝑓𝑑 𝑠
𝑓
𝑙
𝑑
Constraints ∑𝑓 ∑𝑙 ∑𝑤 𝑋𝑠𝑓𝑙𝑤𝑑 + ∑𝑓 ∑𝑙 𝑌𝑠𝑓𝑙𝑑 = 𝑆𝐴𝑠𝑑 ∑𝑠 ∑𝑤 ∑𝑑 ∑𝑠 ∑𝑙
𝑋𝑠𝑓𝑙𝑤𝑑
𝑌𝑠𝑓𝑙𝑑 53
𝑆𝑂𝑠𝑓𝑙
≥ 5
(1)
∀𝑠, 𝑑
𝑌
+ ∑𝑠 ∑𝑑 𝑆𝑂𝑠𝑓𝑙𝑑 ≤ 900 ∗ 𝐹𝐿𝑜𝑝𝑒𝑛𝑓𝑙 ∀𝑓, 𝑙 𝑠𝑓𝑙
(2) (3)
∀𝑓, 𝑑
𝑋𝑠𝑓𝑙𝑤𝑑 ≥ 0
(4)
𝑌𝑠𝑓𝑙𝑑 ≥ 0
(5)
Constraint (1) matches demand at SKU and warehouse level with supply from factories. Manufacturing capacity constraint is formulated in (2) with the assumption of maximum of 900 annual shifts per line of a factory. Load formation feasibility at factories has been modeled in (3). It has been assumed that direct dispatch would be feasible from a factory to a particular distributor only when a minimum of 1 weekly shipment (of 5 tons each) could be formed. Thus, a minimum of 53 shipments on a factory distributor lane are required annually. Constraints (4) and (5) are non negativity constraints on quantity flows via warehouses and via direct dispatches respectively.
6
Conclusion In the paper, the authors have examined a supply network model involving manufacturing facilities, warehouses and distributors with the objective of minimizing total landed cost at distributors. The model is useful in finalizing the supply network in mid to long term. Future demand projections along with changing demand patterns (in terms of geography and product line) can be fed to the model to identify the best fit supply network. It is also recommended using direct dispatches from factories to distributors as a part of decongesting warehouse traffic and minimizing secondary distribution costs. A mixed integer program has been formulated to simulate the supply network model. Since we focused on a multi echelon supply chain involving factory lines, warehouses and distributors, the total number of variables and hence the size of problem could get very large. Managerial discretion in aggregating demand centers (represented by distributors), products and factory lines would be helpful in controlling the problem size. The proposed algorithm can also be used to design and optimize supply chain networks of manufacturing companies operating in multiple geograhpies. References Ambrosino, D. and M. G. Scutella, 2005. Distribution network design: new problems and related models. European Journal of Operational Research, 165(3), 610–624. Beamon, B.M., 1998. Supply chain design and analysis: models and methods. International Journal of Production Economics, 55 (3), 281–294. Chopra, S. and Meindl, P., 2004. Supply chain management: strategy, planning and operation. 2nd ed. Englewood Cliffs, NJ: Prentice-Hall. Crainic, T. and Laporte, G., 1997. Planning models for freight transportation. European Journal of Operational Research, 97 (3), 409–438. Ghiani, G., Grandinetti, L., Guerriero, F. and Musmanno, R., 2002, A lagrangean heuristic for the plant location problem with multiple facilities in the same site. Optimization methods and software, 17, 1059-1076. Lee, J., I. Moon and J. Park, 2010. Multi-level supply chain network design with routing. International Journal of Production Research, 48(13), 3957-3976. Manzini, R., Gamberi, M. and Regattieri, A., 2006, Applying mixed integer programming to the design of a distribution logistic network. International Journal of Industrial Engineering: Theory, Applications and Practice, 13, 207-218. Riccardo, M. and Gebennini, E., 2008. Optimization models for the dynamic facility location and allocation problem. International Journal of Production Research, 46 (8), 2061–2086. Tiwari, A., P.C. Chang and M.K. Tiwari, 2012. A highly optimized tolerance-based approach for multi-stage, multiproduct supply chain network design. International Journal of Production Research, 50(19), 5430-5444. Wu, T., Low, C., and Bai, J., 2002. Heuristic solutions to multi-depot location-routing problems. Computers and Operations Research, 29 (10), 1393–1415. Young, K.K. and Hwan, K.K., 2003, Locating a single facility considering uncertain transportation time. International Journal of Industrial Engineering: Theory, Applications and Practice, 10, 467-473.
7