Excerpt
Table of Contents
Abstract
Figure Index
Table Index
List of Abbreviations
1 Introduction
2 Background and Review of Related Literature
2.1 The Importance of Inventories
2.2 Overview of Supply chain Disruptions
2.3 Assessing the Impact of Supply chain Disruptions
2.4 Managing and Mitigating Supply chain Disruptions
3 Case study: An Aseptic Production process of a Pharma-ceutical company
4 Methodology
4.1 Description of the Base Model
4.2 Simulation Methodology and Production Disruption measures
5 Analysis and Results
5.1 Illustration of the Risk Exposure
5.2 Sensitivity Analysis
5.3 Risk Management Strategies based on different Scenarios
6 Conclusion and Outlook
6.1 Conclusion and Discussion of the Results
6.2 Limitations and Outlook for further Research
Appendix
References
Abstract
Driven by the case of a pharmaceutical company, this thesis examines the effect of severe production process disruptions over the market cycle of a high-margin product. Using Monte Carlo simulation, the financial, as well as operational risks and performance implications, are quantified under various scenarios. Thereby, the model includes the possibility of repeated disruptions based on stochastic occurrences and the resulting short and long-term consequences due to the availability of substitutes. Besides, the study evaluates appropriate risk management strategies based on safety stock policies and different risk preferences of the decision-maker. The results indicate rather moderate implications on expected profits due to production disruptions but a substantial impact on the downside risk for a firm. The optimal inventory policy depends significantly on the disruptions parameters, characteristics of the firm and risk appetite of the decision-maker and cannot be generalized. In case of a highly profitable product, conservative inventory policies are advantageous, as they do not only mitigate risks but also lead in most scenarios to a higher expected profit than lean policies.
Figure Index
Figure 1: Log-normal distribution of lead time
Figure 2: Poisson distribution of production disruption duration
Figure 3: Cumulative profits (CPL) frequency distribution in the 1st scenario
Figure 4: Cumulative profits (CPL) frequency distribution in the 2nd scenario
Figure 5: Mean cumulative profits (CPL) in relation to safety stock
Figure 6: Value at Risk (0.95) cumulative profits (CPL) in relation to safety stock
Table Index
Table 1: Overview of the analysed scenarios
Table 2: Descriptive statistics for the 1st scenario
Table 3: Descriptive statistics for the 2nd scenario
Table 4: Sensitivity analysis for probability of production disruption
Table 5: Sensitivity analysis for production disruption duration (A)
Table 6: Sensitivity analysis for production disruption duration (B)
Table 7: Overview of the analyzed risk scenarios
List of Abbreviations
Abbildung in dieser Leseprobe nicht enthalten
1 Introduction
In 2011, the nuclear crisis in Japan, caused by the earthquake and tsunami, leading to a production drop at Toyota by 40,000 vehicles, and costing the company $72 million in profits per day (Pettit, Croxton, & Fiksel, 2013). Just recently, in September 2019, the United Auto Workers’ 40-day strike caused in General Motor’s plants in Mexico and Canada temporary shutdowns and consequently, a loss of about 300,000 units in production. The strike is expected to cost the firm up to $4 billion in 2019, and also its suppliers will be significantly affected (Wayland, 2019). These are just two examples illustrating which tremendous impact supply chain disruptions can have on a firm’s operational and financial performance. Empirical research by Hendrick and Singhal (2003; 2005a; 2005b) has demonstrated that even small disruptions can have a negative and often long-term effect on sales growth, stock price performance, and shareholder wealth. Consequently, supply chain disruptions constitute a substantial risk for corporations.
However, history has demonstrated that firms can effectively manage supply chain risks and mitigate the impact of disruptions enormously. A famous example is the case of the telecommunication equipment manufacturers Ericsson and Nokia during the beginning of the 21st century. In March 2000, a fire destroyed parts of the production line at a semiconductor plant in Albuquerque, New Mexico (USA), of Philips Electronics, a supplier of significant importance for both firms. While Nokia was able to mitigate the consequences by temporally switching production to alternative plants and suppliers, Ericsson incurred lost sales of $ 400 M (Latour, 2001).
Motivated by these events and especially by the case of a pharmaceutical company with a highly profitable product, this thesis analyses the impact of severe production disruptions on the operational and financial risk exposure as well as performance using Monte Carlo simulation (MCS). Short and long-term consequences, the possibility of repeated disruptions over the market cycle of the product based on stochastic occurrences, and scenario analysis with changes in disruption parameters are included in the study. In particular, the possibilities of mitigating the risk and increment of expected profits are examined based on different safety stock policies.
The remainder of this thesis is structured as follows. The next chapter provides background information on inventories and an overview of related literature concerning supply chain disruptions. Also, it is outlined how this study relates and contributes to the current state of research. Chapter 3 presents the case study on which the motivation and data for this study rely on. In chapter 4, the parameters and assumptions of the periodic review inventory model, as well as the simulation methodology, are explained. Moreover, the disruption event and its implications are described. Chapter 5 shows the results of the simulation and presents sensitivity analysis based on different risk scenarios. Besides, appropriate inventory-based mitigation strategies are derived based on different scenarios. Finally, chapter 6 concludes the study and limitations, as well as possibilities for extensions and further research, are discussed.
2 Background and Review of Related Literature
2.1 The Importance of Inventories
Even though considered as a waste of resources serving no practical purpose, according to the Just-in-time manufacturing (JIT) theory (Walters, 2003), there are important reasons for obtaining and holding inventory. Especially for process uncertainties, fluctuations in demand, unreliability in supply, price protections, and lower reordering costs (Muller, 2011). From a functional perspective, the literature has identified several types of inventory. For instance, Slack, Chambers and Johnston (2010, p. 343) classify them as “buffer inventory, cycle inventory, de-coupling inventory, anticipation inventory, and pipeline inventory”. Buffer inventory, which is also called safety inventory, ensures that there is always a certain number of products in stock and hence functions as a reserve for emergencies (Walters, 2003). These emergencies or operational risks include mostly uncertainties in the supply process (e.g., the unreliability of suppliers, machine breakdowns) and unexpected fluctuations in demand. By protecting against these risks, an organization avoids stock-outs, which may incorporate costs such as less profitable emergency orders and attendant costs of expediting and rescheduling (Silver, Pyke, & Peterson, 1998). Moreover, it evades costs by failing to supply to customers. Some customers are probably not willing to wait and take their business elsewhere (Slack, et al., 2010), which might negatively impact demand in the future.
However, stocks are expensive. There are costs for the storage of the products, such as renting, heating, or lighting the warehouse, as well as insurance costs (Slack, et al., 2010). Additionally, an organization encounters costs of the tied-up capital due to the time lag between paying the suppliers and receiving payments from the customers. The associated costs include the funding costs and opportunity costs of not investing the capital elsewhere (Slack, et al., 2010; Walters, 2003).
Because of this inventory holding related costs, the trend to leaner inventory methods such as JIT, Total quality management, and lean thinking has increased over time. As Japanese manufacturers effectively started to develop this approach in the 1970s, many European manufacturers followed, and in the meantime, almost all major organizations use some elements of JIT (Walters, 2003). Research has provided empirical evidence that firms which employed these techniques decreased their inventory levels and thus improved their financial performance (e.g., gross profit, operating profit) (Cannon, 2008; Capkun, Hameri, & Weiss, 2009; Zbigniew & Bieniasz, 2016).
2.2 Overview of Supply chain Disruptions
“Supply chain disruptions are unplanned and unanticipated events that disrupt the normal flow of goods and materials within a supply chain” (Craighead, Blackhurst, Rungtusanatham, & Handfield, 2007, p. 131). Consequently, they can cause tremendous losses in shareholder value, production, sales, and reputation and might harm relationships with various stakeholders of the company (e.g., customers, suppliers) (Bode, 2008). In the 21st century, the research of academics in the field of supply chain disruptions has significantly increased, and academic journals have published various articles (Snyder, et al., 2016; Ho, Zheng, Yildiz, & Talluri, 2015). According to Snyder, et al. (2016) who summarize the recent literature, four major reasons have caused this effect. First, the described trend to leaner supply chain practices like JIT has increased the vulnerability of a firm to disruptions, as it works perfectly under normal conditions but leaves almost no more space for errors in the system. Secondly, significant disasters such as the terrorist attacks on the World Trade Centre on September 11, 2001, in New York (Sheffi, 2001) and Hurricane Katrina, which devasted the south coast of the USA in 2005 (Katz, 2005), caused tremendous losses and increased public attention. Moreover, corporations act increasingly on a global scale, and hence their supply chains are distributed all over the world. Some parts are also in politically and economically unstable countries. Finally, the momentum effect, since increasing public attention encourages more scholars to study the topic.
Supply chain disruptions can vary in different factors, such as their cause, magnitude, or impact. Scholars have therefore derived different categories, “usually labelled as supply chain risk sources, in terms of being a known source from which supply chain disruptions emerge with a certain probability” (Bode, 2008, p. 10). The applied method and number of sources varies between different articles. For instance, Olson and Wu (2010) propose two categories (internal and external), while Manuj and Mentzer (2008a) use eight (operational, demand, supply, security, macro, policy, competitive, resource). Ho, et al. (2015) apply a very comprehensive method. First, they distinguish between macro and micro sources. Macro factors include external legal issues, political and economic stability, sovereign risk, natural disaster, and terrorist attacks. The category "micro" further splits into the subcategories demand (e.g., inaccurate forecasts), manufacturing (e.g., labour dispute), supply (e.g., delivery failure), and different types of flows (information, transportation, financial).
2.3 Assessing the Impact of Supply chain Disruptions
After the identification of supply chain risk sources, the analysis of these potential risks plays a crucial role. In the literature, various studies focus on quantifying the impact of supply chain disruptions under different conditions and supply chain configurations. As the analytical evaluation of the impact of disruptions in complex supply chains is rather challenging, simulations are a suitable approach to gain useful insights (Snyder, et al., 2016). Various scholars use discrete-event simulation and MCS to quantify the risks related to different disruptions.
Deleris, Elkins, and Paté-Cornell (2004) analyse the losses from a fire hazard on manufacturing supply network to provide support in decision making for the managers. They integrate the hazard and operations model into a Semi-Markov process model and use MCS to get the probability distribution of the total losses for the firm over time. Thereby, the authors define supply chain loss as financial loss per operating quarter, which includes direct (e.g., property damage) and indirect losses (reduction in supply chain output). Deleris and Erhun (2005) use a similar approach to simulate the supply chain risk exposure of a high-tech company. In addition to Deleris, et al. (2004), their approach incorporates multiple external events, such as strikes, political instability, natural disasters that disrupt the economic activity and includes dependencies between products and facilities in the supply network. However, the model is still relatively static; for instance, it does not generate lead time estimations.
Schmitt and Singh (2009) incorporate a more flexible approach by using MCS and discrete-event simulation to model downtime and the impact on customer service due to disruptions at a large consumer products company. They quantify the potential disruptions in the supply chain by creating risk profiles and include flexibility by testing the impact of different parameters, such as disruption probabilities and customer behaviour. Thereby, they demonstrate an entire dependence of customer service on inventory levels at the beginning of a disruption and hence changing risk levels over time. Moreover, Schmitt and Singh (2011) show in a discrete-event simulation for a multi-echelon supply chain that disruptions may be local while they occur but their impact can be global, and the cost of loyalty and backordering are critical since they lead to complicated dynamics while recovering from a disruption.
2.4 Managing and Mitigating Supply chain Disruptions
Assessing the impact of potential disruptions in a supply chain is crucial for a corporation. However, it is even more essential to manage these risks and to find effective mitigation strategies. Scholars have identified various tactics of disruption management. Tomlin (2006) clusters them into four main categories: passive acceptance, operational contingency, financial mitigation and operational mitigation. Passive acceptance means merely accepting the risk and hence, leaving the company open for the potential of severe losses. But as mitigation strategies are not free of charge, passive acceptance can be appropriate in some circumstances. Operations contingency is rather reactive, meaning the organization acts only as soon as the disruption occurs . Mitigation strategies, on the other hand, are proactive, organizations take actions in advance of potential disruptions. Financial mitigation refers to the use of insurance to protect against business interruptions. In contrast, operation mitigation includes sourcing and especially inventory strategies which is, according to Snyder, et al. (2016), one of the most straightforward ways to protect the supply chain against disruptions.
Several scholars focused on finding the optimal inventory policy for different scenarios and decision variables (e.g. costs, reorder point, continuous or periodic review) in the presence of disruptions. To the best of my knowledge, Parlar and Berkin (1991) have been the first to consider supply chain uncertainty in a continuous-review inventory model. They assume one supplier whose availability changes between ON (available) and OFF (unavailable) states. Moreover, they set the reorder point to zero, so once the inventory level is zero and the status is ON, an order can be placed. By using the renewal reward theorem, they find the optimal order quantity. Bar-Lev, Parlar and Perry (1993) extend the study by adding stochastic demand as a new factor to the Idea of OFF and ON states. Gürler and Parlar (1997) and Heimann and Waage (2007) add the possibility of multiple unreliable suppliers whose availability may be ON or OFF at different times for random durations. Furthermore, they analyse the case of the reorder point as a non-negative decision variable. Heimann and Waage (2007) prove that this results in cost savings compared to the zero-inventory option.
Besides, periodic-review models have been used by various authors to determine the optimal parameters in the presence of supply chain disruptions. Parlar, Wang and Gerchak (1995) consider a periodic-review setting with a finite planning horizon, random demand and an unreliable supplier where the ON and OFF states are geometrically distributed. Moreover, they include two setup costs, one whenever an order is placed and one when an order is filled. With this set-up, the authors prove that the optimal solution to this problem is an (s, S) policy where s depends on the availability state of the supplier in the previous period while S is independent of the availability state. Özekici and Parlar (1999) provide a rather general model that attempts to find the optimal (s, S) solution for changes in the Markov environment which influences demand, supply and the cost parameters. Their research shows that the environment-dependent (s, S) policy is optimal when the order cost is linear in the order quantity. Several researchers have extended the studies by Parlar, et al. (1995) and Özekici and Parlar (1999). For instance, Li, Xu and Hayya (2004) also consider the case of lost sales, besides, to demand backlogging, Atan and Rousseau (2015) determine the optimal base-stock level for perishable products and Saithong and Luong (2019) model the length of the supply chain disruption as a continuous random variable in contrast to former studies using a discrete random variable.
System dynamics has also been used by various authors to study supply chain disruptions and to provide mitigation strategies. Just recently, Strohhecker and Größler (2019) present a study of a pharmaceutical company in which they investigate the impact of severe production breakdowns on operational and financial performance to find the optimal inventory policy. The authors show that the one best inventory policy does not exist but is somewhat dependent in a highly non-linear way on economic product characteristics and shows threshold behaviours. To the best of my knowledge, this is the first study to include customer adaption to service level changes which leads not only to unfilled demand when a disruption occurs but also reduces future demand due to the availability of a substitute. Thereby, they also examine variations of the availability of a substitute product.
This thesis extends the study of Strohecker and Größler (2019) by including the possibility of repeated disruptions over the market cycle of the product based on stochastic occurrences. As a base model, a periodic-review inventory model with variable lead time is used. Demand depends on the service level provided by the firm to the customers and can hence lead to long-term damages in case of stock-outs caused by a production disruption. Compared to previous studies, which rather focused on optimizing the periodic-review model under disruption scenarios (e.g. Parlar, et al., 1995; Li, et al., 2004; Saithong & Luong, 2019), this study uses a MCS to provide the decision-maker with robust results showing the financial risks under different what-if scenarios and sensitivities. The financial risks are illustrated as a distribution of cumulated profits over the market cycle (500 weeks) of the product. Furthermore, the study examines inventory-based mitigation strategies depending on different risk scenarios and thereby considering a risk-neutral decision maker, focusing on optimizing expected value, as well as a risk-averse decision-maker.
Since the first breakthrough by Daniel Bernoulli in 1738 (Originally published in Latin and later translated in English in Bernoulli (1954)) showing that people may value the same lottery differently due to differences in their psychology, the concept of risk aversion has widely been studied in economics, finance and psychology. However, in the context of supply chain disruptions, relatively few studies have considered models with a risk-aversion (Snyder, et al., 2016). This study incorporates the concept by providing inventory-based strategies for a risk-averse decision-maker who prefers mitigation down-side risks due to disruptions, rather than maximizing expected profit.
3 Case study: An Aseptic Production process of a Pharma-ceutical company
The case is based on a multinational company’s plant in Germany of a pharmaceutical company described in the study of Strohhecker and Größler (2019). Apart from other products, the company recently started to produce and to market a new pharmaceutical aerosol. The aerosol must be prescribed by a doctor and should be inhaled by patients regularly to be effective. As a patent protects the medicament, there is no comparable alternative available on the market. The substitutes sold by competitors have a lower therapeutic effect and cause temporarily adverse effects when customers switch to these drugs. The market cycle of the patent lasts ten years (approximately 500 weeks); afterwards, it generics will be introduced by the competition, leading to decreasing margins and sales.
The drug is produced in an aseptic production process, a manufacturing technique wherein a sterile product is packaged in a previously sterilized container to exclude infectious organisms. To ensure the eminently sterilized conditions, the case company uses a highly automated production process, well-trained employees and a modern aseptic cleanroom.
The production process consists of four stages. In the first one, all aerosol’s active ingredients are produced in chemical processes. These ingredients are then used as inputs to produce the first solution in the second stage which is then stored in sterile stainless-steel containers and filled in cartridges using a modern aseptic clean room in the third production step. The latter must be carried out by the workforce in the cleanroom and thus contains the highest risk of contamination in the production process because an average person can easily release millions of particles per day. To mitigate the risk of contamination, the employees are highly qualified, provided with methods and regular training, must wear sterile gowns and are required to follow specific clothing procedures. Additionally, unidirectional flow air currents in the cleanroom provide a barrier against free-flowing microbes and blow the potential contamination away from the exposed drugs. In the fourth stage, the packaging process, the cartridges are labelled and inserted together with the inhalation device into a folding box. The folding boxes are then closed, sealed and put in a cardboard box that is sent to the finished products inventory (FPI).
Due to the prescribed contamination risk, the regulation requires the pharmaceutical company to examine the sterility of the production process in so-called “media-fill” tests. These tests must be carried out twice per year; thus, 20 media-fill tests over the remaining market cycle (500 weeks) of the pharmaceutical aerosol. The probability of a positive test result lies between 2 % and 3 %, according to approximations (1-2 % probability that the production process has been contaminated; about 1 % probability that the media-fill itself contaminates the production process). Hence, over the complete market cycle, the probability of a positive result lies between 33.2 % (= 1–0.9820; for a 1 % chance of a contaminated production process) and 45.6 % (= 1–0.9720; for a 2 % chance of a contaminated production process).
4 Methodology
4.1 Description of the Base Model
Different models have been developed by researchers to provide managers with support in determining the optimal inventory policy, which provides a balance between customer service and costs. In this study, a periodic-review inventory model is used as a base model for the subsequent simulation with production disruptions. In the periodic-review model orders are placed at fixed and regular time intervals and of varying size to reach a determined level (Slack, et al., 2010). The determined level is called the Order-up-to level or Target stock level (S) which is the sum of the expected demand over the lead time (LT) and the order interval (T) plus the safety stock (ss) (Walters, 2003). As the firm place orders weekly in its manufacturing lines, it is, in this case, a more appropriate method than a continuous-review model which requires to check the stock level continuously. In the following, the parameters of the model are explained. Some are adopted by the study of Strohhecker and Größler (2019). A detailed list in the Appendix which also provides values for each parameter provides information about the origin.
Following Strohhecker and Größler (2019) and Gonçalves, Hines and Sterman (2005), customer demand (d) is treated as endogenous, as a function of constant demand (dic) adjusted by a linear function which incorporates the deviation of perceived service levels (SLp) from desired ones (SLd):
For instance, in the case of stock-outs demand decreases because the perceived level falls below the desired one. As widely used in the literature (Lant, 1992; Oliva & Sterman, 2001; Strohhecker & Größler, 2019), the perceived service level is adjusted by an asymmetric first-order exponential smoothing of the actual service level:
Thereby, the adaption to higher service levels is slower than to lower ones (τsld < τslu). The service level measure (sl) is defined as the fill rate, which is the percentage of the products delivered (dl) from the desired deliveries (dd) (Beamon, 1999):
Hence, demand is dependent on the service level provided by the firm to its customers. Furthermore, as the case study firm is selling a pharmaceutical product, unfilled demand is not backlogged but directly lost, as customers cannot wait to take their medicine and switch over to the competition. However, as competitive drugs are inferior compared to the case company, customers return progressively as soon as the drug becomes available.
As the four-stage production and delivery are prone to variability, the lead time (LT) included in the model is set to be variable, following a log-normal distribution. The lower bound of the lead time for the case company is eight weeks, the parameters µ and σ of the log-normal function are initially set to µLT = 12 and LT = 3, as illustrated in Figure 1.
Abbildung in dieser Leseprobe nicht enthalten
Figure 1: Log-normal distribution of lead time
( µLT = 12; LT = 3; MinLT = 8)
To measure performance, not only costs are considered, but rather cumulative profit/loss (CPL) is defined as the main financial key performance indicator (KPI) in the model, in comparison to many other studies. CPL is equal to the sum of the weekly profits/losses (pl), calculated based on the weekly values of for contribution margin (cm), inventory holding costs (ci), stock-out costs (cso) and fixed capacity costs (ccf), over the complete market cycle:
The contribution margin (cm) is equal to the contribution margin per unit (cmu), which is the difference of selling price (s) and variable costs per unit (c), times the deliveries (dl) to customers per week. Variable costs per unit are composed of the variable capacity costs per unit (ccvu) and other variable per-unit costs (co), which include mainly direct labour and direct material costs. The variable capacity costs per unit (ccvu) capture in a semi-linear way the additional costs for maintenance and depreciation, which result due to increasing utilization of the production equipment. They are calculated by multiplying the variable capacity costs per unit at the reference level of capacity utilization (ccvu#) with the fracture of actual (cu) and reference utilization (cu#) adjusted by an exponent (ζ):
Finally, the stock-out costs per week (cso) are the product of units not delivered (so) and the stock-out costs per unit (p) and the weekly inventory holding costs (ci) are the number of finished products inventory (FPI) times the inventory holding costs per unit (hl):
As the demand, lead time and cost parameters are defined, the service level, safety stock and Order-up-to level of the periodic-review model can be computed. According to Cachon and Terwiesch (2013), the optimal service level attempts to minimize the backorder costs (b) as well as the costs of holding inventory (h). Choosing a too high service level leads to high holding costs while choosing a too low level dissatisfies customers and increases back-order costs. Following Cachon and Terwiesch (2013), the critical ratio is used to determine the optimal service level (measured as the in-stock probability), which minimizes costs. In the periodic review model, the in-stock probability is:
Hence, The backorder costs in this study are calculated as the lost contribution margin (cm = s – co – ccvu#) plus the penalty stock-out costs (p) and holdings costs as the weekly holdings costs per unit (hl):
A probability of 99.86 % under normal distribution corresponds to Z = 2,98.
Based on the service level, the appropriate safety stock (ss) and Order-up to level (S) can be defined as:
The lead-time demand is approximated with a normal distribution (σLTD) (Walters, 2003) because the resulting error has only a small influence on the optimal levels (Tyworth & O'Neill, 1998). As demand is affected only by changes in the service level and otherwise constant (dic), the standard deviation (SD) of the demand rate (σD) is assumed to be zero.
[...]