- 1. Robert R. Inman and Dennis E. Blumenfeld General Motors Research and Development, Warren, MI 48090-9055, USA; Department of Industrial and Operations Engineering, University of Michigan, Ann Arbor, MI 48109-2117, USA (Received 20 February 2013; final version received 13 March 2013) Publisher: Taylor & Francis Lecturer: Dr. Soleimani Summarizer: Mohammad Beikzadeh
- 2. •Fisher (1997) advocates that firms match their supply chain to their products’ demand characteristics. Slone, Dittmann, and Mentzer (2010) present strategies for achieving supply chain excellence, selecting appropriate technology, and managing change in the supply chain. •Paulonis and Norton (2008) report survey results indicating that the factor with the most influence over supply chain strategy is ‘increasing complexity of products/services What firm does not want a lean supply chain? What firm wants to carry unnecessary inventory? Yet a firm should be wary of adopting another firm’s supply chain model, even a renowned world-class supply chain model, if its product differs on key criteria. •Vachon and Klassen (2002) observe a correlation between the complicatedness of the product/process and poor delivery performance.
- 3. •Modeling remanufacturing, Ferrer and Ketzenberg (2004) observe that the value of short lead time increases with the number of parts. •Burkett (2008) observes that ever longer value chains and shorter product life cycles increase the sensitivity of operations to product complexity. there appears to be a gap in the supply chain research literature regarding the impact of a product’s complexity. This paper helps to address this important deficiency in the literature. For a motivating example, compare a department store with a manufacturer of a complex product – say an automobile. • If a department store is missing a single stock- keeping unit, it forfeits the margin on a potential sale. • If an automotive manufacturer is missing a single part, it can shutdown the entire assembly line.
- 4. DEFINITIONS OF PRODUCT COMPLEXITY •Clark and Fujimoto (1989) suggest that the number of total parts in a model is related to product complexity. •Novak and Eppinger (2001) identify the following three elements product complexity: number of components, component interactions product novelty •Lucchetta, Bariani, and Knight (2005) argue that while product complexity is due both to the number of parts and the difficulty of generating those parts, the number of parts is a practical measure of complexity because the number of production steps is strongly related to the number of component parts.
- 5. • The more unique parts there are in a product, the greater the risk that one of the parts is not available on schedule for production. Although a reasonable measure of product complexity, the number of parts is only a proxy for supply chain complexity. An alternative measure of supply chain complexity related to the number of parts is the number of unique supply chains. •a single supplier location supplies multiple parts that differ only in their colour or configuration (such as parts for the left versus right side of a product), then the number of supply chains would probably be more accurate measure of complexity. •a single supplier location will supply dramatically different parts. Even if the parts have the same next tier suppliers further up the supply chain, if they have different processing requirements, it would be more accurate to count parts, not supply chains. Hence, there does not appear to be any perfect and yet straightforward measure of complexity. •This paper’s analysis approach could be applied using the number of unique supply chains instead of parts. we use it because it is simple, unambiguous, known prior to supply chain design, and has precedent in the literature.
- 6. Model for impact of missing parts on supply chain risk To produce or customize a product, all the parts or components must be available on schedule for efficient production. . If any part is missing, the product cannot be manufactured as planned. Countermeasures may be available to mitigate the impact – such as : substituting a different part, installing a feature at a later date, deleting a feature altogether • but countermeasures carry at least some extra cost or negative side-effect. • Imagine a firm assembling two versions of a product: a standard version and a full-featured version
- 7. •In the second option we should notice several points: The full-featured version includes all of the standard version’s parts and augments with many additional parts unique to the full-featured version. it has a greater risk of missing a part than the standard version because that requires all the standard version’s parts – and then some more. The full-featured version carries all the standard version’s risk plus the additional risk of missing one of the additional parts. •A part can be missing when desired for many reasons, such as the following: logistics delays (caused by traffic, weather, customs delays, or political unrest) scheduling errors poor quality causing a shipment of parts to be rejected parts misplaced in the plant inventory shrinkage inaccurate parts inventory records at the plant mislabelled parts supplier behind schedule insufficient supplier capacity
- 8. • Manning (2008) reports data from a survey of 138 companies indicating that, in 2012: 99% suffered a supply chain disruption 58% suffering financial losses as a result 39% of the respondents reported delayed, damaged, or misdirected shipments in the past year. The following model quantifies the relationship between : the total number of component parts in a product, the probability of missing any individual part, the impact on the risk of an incomplete kit of parts for production. Notice: In this model assumes delivery times of different components are independent, and estimates the probability that all components for the product arrive on time. our model assumes for simplicity that shortages of different parts are independent. Such independence may not always be the case, such as, for example, if some parts are sourced from the same supplier. N = number of parts per product = probability that component part i is available (i = 1, 2, ..., N) = 1 – pi = probability that component part i is missing M = probability that at least one component part for the product is missing
- 9. • Assuming shortages of different parts to be independent, the probability that all N parts for the product are available is simply: • Hence, the probability that at least one part is missing, M, is: • As missing one or more parts disrupts production, Equation (1) gives the probability of disruption. • If the probabilities are the same for all parts (i.e., pi = p, qi = q, for all i), then the probability of missing least one part, M, (i.e., probability of disruption) becomes M •Equation (2) shows how the disruption probability M depends on the number of different parts N the product requires and the probability q that an individual part is missing. •Figures 1 and 2 display the disruption probability, M, for different combinations of each part’s shortage risk q, and the number of parts in the product, N.
- 10. •If N= 2 and q= %1 as a result M= %2 •If N= 20 M is supposed to be %2 as a result q should %0.1 •If N= 2000 and q= %0.1 as a result M= %20 Equation (3) can be used to estimate how low q must be to ensure an acceptably low supply chain risk, as measured by the disruption probability M. •If M is supposed to keep %2 and N= 2000 as a result q should be %0.001 Table 1 shows the values of q from Equation (3) for given disruption probabilities M and given numbers of different parts per product, N.
- 11. •Table 1. Relationship of missing part probability to overall supply chain risk.
- 12. •If M is supposed to be %2 as a result for N= 20 q should be % 0.1010 , N= 200 q should be % 0.0101 or N= 2000 q should be %0.0010 •Figures 1 and 2 and Table 1 highlight product complexity’s dramatic impact on risk. The more parts that must be simultaneously available, the greater the risk of disruption. Firms that assemble complex products face higher supply chain risk than assemblers of simple products and much higher risk than retailers. • For example, Holweg and Pil (2001) report: 1. largest personal computer-makers manages, N= 15–20 components per computer, according to fig If q= %0.1, as a result M= %2. 2. automotive manufacturers, N= 2000 components per vehicle, according to figure 1:If q= %0.1, as a result M= %90. • As a result the automaker is much more sensitive to supply chain disruptions because its product is much more complex.
- 13. Supply chain design also affects the risk of disruption. Chopra and Meindl (2004) identify the following decisions as part of the supply chain design process: •the number and location of production facilities • the amount of capacity at each facility •the assignment of each market region to one or more locations •supplier selection for sub-assemblies •components and materials Meixell and Gargeya (2005) extend this definition to include the decision of selecting international facilities, which involves additional globalization considerations. Besides, we describe four supply chain design decisions that affect the likelihood of disruption.
- 14. These four decisions loosely follow the first four stages of Govil and Proth’s (2002) list of the major supply chain activities: Buy Make Move Store Sell (1) Make or buy (in-source or out-source) (2) ) Supplier location (3) Shipping mode and route (4) Consolidation and deconsolidation centers • A point : Consolidation and deconsolidation centers often reduce average shipping cost but increase lead time and the chance of missing a shipment. They increase the lead time for three reasons: First, the consolidation or deconsolidation centre is rarely on the shortest route from the supplier. Second, every shipment is double handled with extra unload, store and load steps. Third, consolidation centers can delay shipments International Journal of Production Research 5 Downloaded by waiting to fill a complete load.
- 15. • In general (assuming suppliers have identical internal reliability) : • the greater the lead time • number of stakeholders •number of steps in the supply chain • the more risk there is of disruption. Of course, often these distant or complex supply chains yield lower piece prices or nominal shipping costs that may offset the increased risk • These factors offer the firm a choice between higher nominal cost with low risk or a lower nominal cost with higher risk and longer lead times. • Explicitly considering the product’s complexity will lead to better supply chain designs – supply chains appropriate for the product they support. • To integrate thoroughly product complexity with supply chain design, ideally we would quantify the impact of each element of supply chain design on part availability risk. Then we would assess the risk of production disruption for each supply chain design.
- 16. •Figure 3 suggests that risk increases with shipping lead time whether The lead time is due to distance, mode Mode consolidation/deconsolidation centers •Analysing data from a global personal computer manufacturer that purchased components globally: Levy (1995) observed that supply chain disruptions increased with distance. Bozarth et al. (2009) surveyed 209 plants and found that long supplier lead times have a significant negative impact on schedule attainment, unit manufacturing cost, customer satisfaction and competitive performance
- 17. •considering two hypothetical suppliers that are identical in every respect, such as facilities, machinery, personnel, processes and systems, except their location: •One is located next door The supplier next door moves material to our plant via truck with a lead time of 5 min. •other is 5000 miles away The supplier 5000 miles away moves material with a combination of truck, rail and ocean moves with a lead time of 5 weeks – including the same 5 min truck trip as the supplier next door. If the supplier lead time is 5 min, the new part can be used 5 min after it is produced. If the lead time is 5 weeks, we either need to wait 5 weeks or, more likely, pay air freight charges for 5 weeks of production as the normal mode pipeline fills with the new part. •Also, instead of lead time, we could use: distance the number of mode changes the number of border crossings •While many factors influence the supply risk of a particular part, a simple aggregate measure that comprehends many of these factors is shipping lead time.
- 18. • Other factors which increase a part’s risk of delay: The more border crossings consolidation and/or deconsolidation centers mode changes, the longer the lead time. While distance would be an alternative metric, the transportation mode and supply chain design also affect the risk, for example: an inter-continental air shipment may have less risk than a domestic truck shipment via a cross dock. Hence, for the purposes of a high-level model of supply chain risk, shipping lead time seems preferable to distance. •Of course lead time is not a perfect metric. For example: although short lead times are generally to be preferred, there is a benefit to long lead times in cases of production disruption. A long lead time provides buffer inventory for the final assembly plant, so that parts still arrive if a production stoppage occurs upstream in the supply chain. This allows time for the assembly plant to respond and adjust to the changed conditions, thus mitigating production risk. This benefit could also be achieved by holding safety stock at the assembly plant, regardless of the lead time. Still, there is a potential trade-off between short and long lead times, when both production and supply chain risks are considered.
- 19. •With the above justification and caveats in mind, we now model each part’s availability probability as a function of its lead time. We assume that the probability qi (= 1-pi) that part i is missing increases linearly with lead time Li, i.e., •Where: Li is the lead time from part manufacturer to assembly plant for part I ai is a constant for part i LC is a constant. •Then, Equation (1) for the probability that at least one part is missing, M, becomes •As before, if the probability parameters are the same for all parts (i.e., ai = a, Li = L for all i), then the probability of missing at least one part, M (i.e., probability of disruption), becomes:
- 20. •Figure 4 illustrates that the probability of disruption M depends on lead time L (in days) and number of parts N, based on Equation (5), for an example where a= %0.01 , LC = 10 days and N= 2 or even 20 lead time L has relatively little impact on the disruption probability. However, for a large number of parts, the disruption probability increases substantially with lead time. Figure 4 shows that, for all lead times, the probability of disruption increases with product complexity as measured by the number of parts.
- 21. •In many situations, parts are shipped from several different suppliers at different locations. Lead time L may therefore be different for different parts. We illustrate this with an example of two supplier locations, A and B, where shipments of parts to the assembly plant have a long lead time from A, and a short lead time from B. N= number of parts per product (as before) f = fraction of number of parts shipped from supplier A LA = lead time for shipping from supplier A to assembly plant LB = lead time for shipping from supplier B to assembly plant •From Equation (4), assuming as before that parameters a and LC are the same for all parts, the probability of missing at least one part, M, is:
- 22. •Using Equation (6), Figure 5 plots the probability of disruption M against number of parts N and the fraction f of parts sourced from the distant supplier. for an example where LA = 30 days and LB = 1 day, and where (as before) the constant a = 0.01% and the constant LC = 10 days. The disruption probability increases substantially as the fraction f of parts shipped from the distant supplier increases. This increase becomes more dramatic as the number of parts increases. •Figures 4 and 5 display the compounding effect of product complexity and lead time on the risk of disruption. According to this simplified model, if a firm assembles complex products, it should pay particular attention to supplier lead times.
- 23. Figure 6 displays this conceptual framework as a two-dimensional schematic where the area inside the rectangle is the total cost of disruptions. Disruption prevention Disruption mitigation •The following two subsections discuss some strategies for preventing disruptions from occurring and for mitigating the impact of disruptions that slip through. These strategies are neither exhaustive nor mutually exclusive, but present practical steps for managing potential disruptions.
- 24. •Disruption prevention Below are selected strategies for reducing the likelihood of disruptions. Product simplification Applying postponement to delay differentiation Sourcing to reduce the risk of shortages Incorporate disruption risk in planning the logistics Carry inventory to reduce the risk of disruption Improve demand forecasting Adjust prices to shape demand in accordance with parts in supply Componentization •Disruption mitigation While the preceding strategies help to prevent disruptions, despite a firm’s best efforts some disruptions inevitably occur. When they do, the following tactics reduce the disruption’s impact. Figure 7. Managing risk reduces the probability of plant disruption.
- 25. Apply business continuity planning to the supply chain Expedited freight Material tracking systems/supply chain visibility Design the product with substitutable parts Local sourcing Dual sourcing •The tragic 2011 earthquake in Japan did far worse than disrupt logistics networks; besides the terrible human cost, it destroyed or damaged many production facilities. The ramifications of this catastrophe do not pertain to logistics delays, but do contain useful lessons for understanding the impact of product complexity on supply chain resilience. Automobile production, with its very complex product, was particularly disrupted by the earthquake.
- 26. •While the Japan earthquake impaired critical component production, which in turn disrupted production of the plants using those components, many supply chain disruptions occur within the logistics network. One example was the 10-day US West Coast port strike in 2002. Krause (2002) reports that ‘automobile manufacturers were hit especially hard’ ,causing two major automobile manufactures to close US plants temporarily, laying idle thousands of workers each. •Another example was the 2010 volcanic eruption in Iceland, which disrupted air freight. Automotive Logistics (2011) reports that the eruption forced a major German automaker to halt production at three plants in Germany and slow down output at a US plant, and cost a major Japanese automaker production of 2000 vehicles because it was unable to import its air pressure sensors from Ireland. •Facing heightened supply chain disruption risk, manufacturers of complex products learn to manage this risk. Figure 7 displays a conceptual schematic of how managing risk reduces the likelihood of disrupting the plant. The more complex the product •To compensate for the increased risk, more effort must be spent on: disruption prevention (such as parts inventory or sourcing from local or reliable suppliers) disruption mitigation (such as investing in logistics and material tracking systems and employing premium freight when needed)
- 27. •As the unmanaged risk increases with the number of parts, so does the amount of effort and investment needed to manage this risk. Recognizing that the risk management burden increases with product complexity, firms should consider their product complexity when designing their supply chains. •Another research opportunity would be to recognize that supply chain risks are probably not independent; correlated risks would be more accurate. Many components may come from the same supplier plant and hence be subject to similar risks. As most production facilities receive components from multiple sources, each with its own distance and probability of disruption, incorporating a range of supplier lead times and risks could be a rewarding analysis. •Finally, it could be worthwhile to investigate more sophisticated models of product complexity, such as incorporating the relationship between product variants and constituent parts, recognizing that while some parts are required on every product variant, some parts are optional and may be correlated or dependent on the usage of other parts.