Managing Uncertainty in Production Planning: A Real Option Approach
Risk and uncertainty are inherently part of the business world that affects all parts of production planning. Furthermore, such uncertainty is by no means a simple phenomenon. There are complex issues inherent in risk and uncertainty, which affect the models most usefully applied in production planning. Further complicating the issue is also that all uncertainty and risk factors are unique for each industry. There is therefore no global risk assessment or uncertainty model that can be universally applied with uniform effectiveness. Businesses can however benefit from generalizations, such as the fact that a lack of attention to uncertainty in planning increases the risk of losing in terms of time, money, or quality. Hence, an examination of how risks can best be mitigated via uncertainty models and production planning is usefully applied, even if this needs to be modified for the specific industry in question.
Mula, Poler and Garcia-Sabeter (2007: 783) for example emphasize that general uncertainty factors within the industrial decision environment include market demand, capacity data and cost information. The authors have gone further by qualifying the various manifestations of uncertainty in terms of three types, namely randomness, fuzziness, and lack of knowledge (also known as epistemic uncertainty.
Yang (2009: 22) in turn focuses on the automotive manufacturing industry to demonstrate the critical necessity of including uncertainty in the production planning process. This industry lends itself well to the investigation of uncertainty contingencies, as it is a capital intensive and slow to change. These factors make constructive and effective planning essential. At the same time, the vast scale and scope of the industry bring unique challenges to the planning process. Specifically, uncertainty in this industry relates to variation in terms of product design and its related market demand. The problem is that manufacturing planning and design decisions are needed years before the start of actual production, and before market demand can be known. Market uncertainty remains and evolves even after production. Hence, the author maintains that uncertainty must necessarily be part of the planning and design processes within this industry. This needs to relate to all elements within the production sequence, including supply chain management.
Supply chain management lies at the basis of production planning, and should therefore also be at the heart of contingency planning and uncertainty risk. Snyder, Daskin and Teo (2004: 2) for example state that design decisions made within the supply chain network are necessarily costly and difficult to reverse. Part of the inherent difficulty here is the fact that costs, demands, distances and lead times can undergo significant changes once design decisions are in effect. This could lead to the risk of economic losses as a result of not properly planning for contingencies related to such uncertainties within the supply chain.
In order to plan effectively for such eventualities, the authors note that the traditional approach has been to either focus on the strategic aspects of supply chain design such as facility location or its tactical aspects such as inventory management. This approach does not however consider both of these factors simultaneously. In strategic models, parameters then tend to be treated as deterministic, while the tendency within tactical models have been to assume that strategic decisions are already made.
The shortcomings of these singular approaches have however been mitigated with the rise of a new model that includes both the strategic and tactical approaches within a single modeling platform, the location model with risk pooling (LMRP), developed by Shen, Coullard and Daskin (cited by Snyder, Daskin and Teo 2). The shortcoming of this model is however that it does not account for the changing nature of the supply chain environment, and can therefore not adequately take into account the full implications of the risks and uncertainties involved.
According to Gupta and Maranas (2003: 1219), this is a problem in the current climate of increasing competitive pressures on a global scale. For this reason, supply chain planning as the basis of production planning has become the highlight of business practice for most industries. In order to demonstrate this point, the authors focus on the chemical industry.
What makes uncertainty modeling particularly complex within this industry is the fact of often conflicting objectives within the various business divisions in terms of marketing, distribution, planning, manufacturing and purchasing. In order to mitigate the complexities and conflicts that contribute to the uncertainty factor, the authors suggest that all these functionalities should integrate within supply chain planning. Focusing on this element will also coordinate and integrate the key business activities of the most complex business environment. Appropriate tactical models then need to be integrated with the planning process in order to mitigate eventualities and uncertainties relating to the various functions within the supply chain.
Rosenblit, Ben-Tal and Galany (2010: 2) in turn focus their attention upon a Robust Optimization (RO) methodology to solve problems relating to demand uncertainty and the manufacturing process. In this model, production decisions are led by the trends within previous production periods, where surplus or shortfall are used to calculate probabilities for future market demands. Costs are then incurred in terms of holding or shortage unit costs as a result of surplus or shortage in response to actual market demands. The benefit of this model is that it provides a quantification of uncertainty that can be used as a guideline. This can then minimize the probability of extreme surplus or shortfall problems, unless unforeseen eventualities occur to drive such deviations.
Kazaz, Dada and Moskowitz (2005: 1101) provide an explanation of how the RO model tends to be used by global electronics manufacturers specifically, and multinational companies in general. For electronic manufacturers, the component manufacturing division for example receives a computer-generated projection of future customer demands for specific components. This is the first stage of the RO methodology. The projections are then used to plan for the maximization of expected profits by manufacturing an optimal amount of each component for each plant. The production plan is then accompanied by budgets and transfer prices for sales in the various countries where they operate. This then includes the uncertainty factor of fluctuating foreign exchange rates. While these fluctuations are generally not major on a , uncertain market conditions could cause sudden spikes or falls.
According to the authors, the usual approach to this problem is to use historical values to indicate the likely exchange rate values for the future. Uncertainty parameters are then replaced by these projected values, like the case is with any other domestic uncertainty factor. The problem is that such a uniform approach does not account for the fact that some exchange-rate realizations are known before allocation decisions are made. Hence, the adverse effect of such fluctuations are not managed adequately. A specific model is then needed for specific global factors such as the exchange rate problem.
. (2003: 97) mention that the uncertainty factor is one among four key aspects of supply chain management challenges, in which the main uncertainties are product demand and price, raw material supply cost, and production cost. They cite various authors that have made modeling suggestions to handle these contingencies. Most importantly, such planning must occur in terms of multiple integrated elements in order to address all the contingencies and complexities involved. Because businesses today are complex, their supply chains must be managed with attention to their nature as multi-level phenomena. Hence, uncertainty management must include a variety of factors from multiple perspectives.
In the light of the above, the research problem can then be said to revolve around the best possible general model that can be developed for managing uncertainty in production planning. Clearly, no author or business person would deny that uncertainty planning is important, and indeed vital to production planning. Without taking uncertainty into account, production problems in terms of quality, cost, and quantity will most likely occur to some degree. It is therefore impossible to simply state the necessity of accounting for uncertainty within production planning as a research problem.
However, a significant problem does relate to the complexity of the issues involved. In production planning, as seen above, uncertainty has many different aspects. Uncertainty in the supply chain for example relates to supply, raw materials, and supplier reliability, while production uncertainty itself relates to market demands, manufacturing costs. Post-production uncertainty relates to the possible surplus or shortage of products, the exchange rates, and other related factors. In addition to the variety of global industries, the challenge is then to create a single model that would be generic enough to include many contingencies and as many possible industries as possible.
Developing such a model should therefore benefit the most manufacturers and industries possible, who should be able to use the model by simply making relevant modifications in order to fit their industry precisely.
In short, the research problem is then to find a suitable model to optimally mitigate the uncertainty in the production process of multiple industries, where such a model might apply to any number of significantly-sized global businesses.
Research Objectives and Scope
The main objective of the research then relates closely to the research problem. It is to research the problem of uncertainty as it manifests in the global business environment. Specific issues to be investigated include supply chain management and its related uncertainties, the production process itself and uncertainties related to it, as well as the post-production phase and market uncertainties that are related to it.
Time is also an important factor. Some industries require a long-term time frame in their planning process, which exacerbate uncertainties. The time factor should also be an important consideration in terms of creating a model that can effectively help businesses achieve their manufacturing and revenue goals.
To achieve these aims, the main objective of the research will then be to research industries and companies that operate on a global scale. They will be investigated for the models they have implemented to mitigate risk and uncertainty factors in their supply chain, production, and post-production processes. These models will then be used to create a model that could possibly serve as a generic framework for global businesses to mitigate their risks optimally.
Finally, the research will be consolidated into a report that explicates existing models and how the new model was constructed from elements of these. Existing knowledge serves as a vital springboard for the creation of new and innovative designs; indeed, this is also true of the manufacturing process. Hence, business models, experts, and literature will be consulted for ways in which global business can be optimized in order to minimize the effects of risk and uncertainty.
Indicating the complexity of the issue, authors focusing on production under uncertainty tend to do so from a variety of perspectives and in terms of several respective industries. This complexity is the challenge that this research must face in order to optimize its proposed model for managing uncertainty within production planning.
Mula, Polder and Garcia-Sabeter (2007: 784) address this uncertainty from the platform of fuzzy data. The authors hold that foreseen demand, based as it is upon concrete historical data such as sales, supplies or competition, can be no more than fuzzy, as the future necessarily includes uncertainty. Furthermore, unforeseen events might include manufacturing breakdowns, faulty production or preparation delays. Planning costs are also necessarily fuzzy, as they tend to be valuated by human perception; in itself not an exact or crisp science. Other contingencies, such as overtime, loss of clients and backlogs can only be calculated based on past human experience. This is necessarily subject to human perception and therefore fuzzy.
The prevalence of the fuzzy element in so much of production planning, according to the authors, can be usefully mitigated by means of using fuzzy set theory concomitantly with linear programming. Particularly, the authors suggest a model that uses fuzzy programming through a possibilitistic approach as applied to (MRP) problems. Through this model, the very nature of uncertainty is used in order to create a model that better applies to the necessities of uncertainty mitigation.
Wu and Ierapetritou (2007: 1129) take a more general approach to the issue by examining existing literature on process operations optimization. Because of the importance of the time horizon in optimal planning processes, a large body of literature suggests that planning and scheduling should occur simultaneously in order to minimize uncertainty. In such models, several time-frame and multipurpose approaches are used. One such model for example determined both the optimal duration of the operating cycle and the schedule to be followed during each cycle.
A second major approach in the literature on the issue is a hierarchical decomposition resulting in various planning and scheduling levels. In such models, the time horizon is divided into various time periods, each with its own demands and contingencies to be managed. In complex environments, this serves to simplify and streamline the production process somewhat.
It has already been mentioned that model predictive control is often used to mitigate the uncertainty associated with the production process (Wu and Ierapetritou 1130). This provides projected targets for sales, revenues, manufacturing costs and concomitant market demands.
Snyder, Daskin and Teo (2004: 2-3) handle uncertainty by allowing the parameters in their model to be described by discrete scenarios, each with a specifically assigned probability value. Issues such as distribution center locations, assigning retailers, and setting inventory levees can then occur optimally as a result of such projections. Ultimately, the goal is to minimize the total expected system wide cost. The authors named this model the stochastic location model with risk pooling (SLMRP).
In this model, strategic decisions and tactical decisions are not made simultaneously. Instead, only strategic decisions — such as facility location — are made immediately, before uncertainties are known. Tactical decisions — such as the assignment of retailers and setting inventory levels — are only made once the uncertainty is resolved. This in itself mitigates uncertainty by fracturing the decision process to match the requirements of uncertainty.
The main advantage of this model is its potential use with various complex supply chain demand problems. Indeed, it appears that this model can be applied across a variety of industries within the global arena.
The main problem under discussion, according to Gupta and Maranas (2003: 1222), is the fact that many decisions must be made regarding optimal operations in the future, but with the availability only of current information. This is where the advantage of the above-mentioned SLMRP model becomes apparent. By dividing the decision process in time before and after uncertainty, some of the uncertainty is mitigated.
According to Gupta and Maranas (2003: 1222), there are two main approaches to mitigate the problem of uncertainty: the scenario-based approach and the distribution-based approach. In the first, scenarios are created that project possible futures for the uncertainties in question. Probability levels are assigned to these scenario levels. These probability levels are then used as the basis of decisions. The main limitation is that all possible futures must be projected. As seen above, this is based upon human perception and experience, which may detract from its potential accuracy and optimal use. In fact, the human element tends to add uncertainty rather than mitigate it.
The distribution-based approach is used when there is not possibility of identifying discrete scenarios. Only a continuous range of potential futures is possible. Each of these then receive a probability distribution. There is no need to forecast possible scenarios in this approach.
Another interesting approach is suggested by Eppler, Platts and Kazncioglu (2006: 5). The author suggest that visualization is a possible way to simplify the manufacturing process and mitigate its uncertainties and problems. This is particularly so in the complex environment of strategizing. Among the many models suggested for minimizing risk, the possibility of visualization provides a platform for better understanding the problems and uncertainties at hand. Where as fuzzy set theory accepts the limitations and non-specific nature of production risks and uncertainties, visualization gives these a concrete form. By facing the challenge of uncertainty in this way, the model provides a concrete form to a non-concrete risk factor. The manager can then better mitigate the risk by graphically manipulating the model and projecting scenarios where the decision-making process can be applied optimally.
Visualization techniques can then be said to be particularly useful in concretizing the uncertainty situation in complex environments such as the global business arena. Although this approach will doubtlessly also have its limitations, such as the possible lack of accuracy in its predictions and mitigation measures, it might in fact be used in conjunction with other models in order to better address the issue.
Many models are available to address the complex issue of production and uncertainty risks. Like various businesses themselves, these models vary in their complexity and scope. By using certain elements from these models in conjunction, a specific business can create a unique platform from which to optimally mitigate its risks and uncertainties. In this regard it is important to recognize that there is probably no single model that can either handle all uncertainties at an optimal level at all times. Indeed, the very nature of uncertainty suggests that this would be a futile endeavor. Instead, the uncertainty model should be required to simply meet the needs of it user.
Fuzzy set theory can for example be used in conjunction with a visualization model in order to better understand the nature of the specific uncertainties involved in a model, business strategy or production effort. Although it has been said that the issue is a complex one, it is therefore not impossible to consolidate relevant literature and knowledge in order to find the best scenario to generally apply to global business.
Eppler, Platts, Kazancioglu 5
Alonso-Ayuso, A., Escudero, L.F., Garin, A., Ortuno, M.T. And Perez, G. An Approach for Strategic Supply Chain Planning under Uncertainty based on Stochastic 0-1 Programming. Journal of Global Optimization, No. 26, 2003. Retrieved from http://chentserver.uwaterloo.ca/aelkamel/che720/che720-methods-of-optimization-pse/stochastic_optimization/05100412180122714.pdf
Eppler, Martin J., Platts, Ken, and Kazancioglu, Emre. Visual Strategizing: the Systematic Use of Visualization in the Strategy Process. Institute for Corporate Communication, Dec. 2006. Retrieved from http://doc.rero.ch/lm.php?url=1000,42,6,20061207091432-KB/wpca0607.pdf
Gupta, Anshuman and Maranas, Costas D. Managing demand uncertainty in supply chain planning. Computers and Chemical Engineering, No. 27, 7 Feb. 2003. Retrieved from www.sciencedirect.com.
Kazaz, Burak, Dada, Maqbool and Moskowitz, Herbert. Global Production Planning Under Exchange-Rate Uncertainty. Management Science, Vol. 51, No. 7. July 2005. Retrieved from http://myweb.whitman.syr..pdf
Mula, J., Poler, R. And Garcia-Sabater, J.P. Material Requirement Planning with fuzzy constraints and fuzzy coefficients. Fuzzy Sets and Systems, No. 158, 2007. Retrieved from http://etc.-jxzzjc.hrbeu.edu.cn/%E4%B8%89%E6%96%B0%E6%96%87%E7%8C%AE%E8%B5%84%E6%BA%90%E5%BA%93/%E6%9C%BA%E6%A2%B0%E5%8A%A0%E5%B7%A5/%E5%85%88%E8%BF%9B%E5%88%B6%E9%80%A0%E6%A8%A1%E5%BC%8F/%E7%89%A9%E6%96%99%E9%9C%80%E6%B1%82%E8%AE%A1%E5%88%92/Material%20Requirement%20Planning%20with%20fuzzy%20constraints%20and%20fuzzy%20coefficients.pdf
Rosenblit, Michal, Ben-Tal, Aharon, and Golany, Boaz. Robust Optimization of Multi-Period Production Planning Under Demand Uncertainty. Israel Institute of Technology. Retrieved from http://msom.technion.ac.il/conf_program/papers/TB/2/137.pdf
Snyder, Lawrence V., Daskin, Mark S., Teo, Chung-Piaw. The Stochastic Location Model with Risk Pooling. European Journal of Operational Research, 18 Nov. 2004.
Wu, Dan and Ierapetritou, Marianthi. Hierarchical approach for production planning and scheduling under uncertainty. Chemical Engineering and Processing, No. 46, 2007.
Yang, Yingxia. A Screening Model to Explore Planning Decisions in Automotive Manufacturing Systems under Demand Uncertainty. Massachusetts Institute of Technology, 2009. Retrieved from http://esd.mit.edu/staging/people/dissertations/yang.pdf
Zanjani, Masoumeh Kazemi, Nourelfath, Mustapha, and Ait-Kaid, Daoud. A Stochastic Programming Approach for Production Planning with Uncertainty in the Quality of Raw Materials: A case in Sawmills. CIRRELT, Feb. 2009. Retrieved from https://www.cirrelt.ca/DocumentsTravail/CIRRELT-2009-08.pdf