A system dynamics model for sustainable corporate strategic planning

Paper aims: This paper presents a manufacturing process model for assessing the effects on economic, social, and environmental targets, given variations on corporate strategies of production, innovation, marketing, and demand for final goods. Originality: The model integrates economic, social, and environmental dimensions that are validated through three main scenarios: Business as Usual (no strategic application), Business as Investment (strategic application), and Business as Vision (changes in demand). Research method: The model estimates the social, environmental, and economic performance through time based on the System Dynamics methodology. Main findings: The results demonstrate the model’s suitability as a decision-support tool for sustainability planning in a corporate environment. Implications for theory and practice: The model facilitates the analysis of the effects of resource allocation on corporate strategy.

Dimensions such as the technological, which is beyond those in the triple bottom line (economic, environmental and social) in the SD models, are discussed in the research of Kim et al. (2014); Joung et al. (2013); and Musango et al. (2012). Furthermore, the political dimension is discussed in the research of Bautista et al. (2019) and quality and environmental management accounting (EMA) in Petry et al. (2020).
To develop SD models applied to industry, Zhang (2019) considered that manufacturing processes are holistic as there are sustainability problems with characteristics such as behaviour, interconnectedness, boundaries, delays, perspectives, uncertainty, and resilience. Identification of these characteristics can then aid the derivation of 9 relationships from relationships between economic, social, environmental, and quality factors (Felicetti et al., 2022). The integration of these relationships serves for an analysis of the risk and uncertainty in these relationships. To map the system relationships, an SD model is used to incorporate the system's behaviours and the assessment of its sustainability. This understanding consequently facilitates decision-making (Zhang et al., 2021). Abdelkafi & Täuscher (2016) showed from the perspective of value creation, integrate four partial models: the firm, the environment, decision making, and the client and their relationships with value creation. Furthermore, decision makers' beliefs and standards regarding sustainability are included, wich are capable of eliciting behavioral changes in both the business model and the way it feeds on the environment and customers. Harik et al. (2015) studied a sustainability index composed of environmental, social, economic, and manufacturing variables is developed from a holistic perspective; the latter includes direct and indirect manufacturing management. Orji & Wei (2015) simulated supplier behavior in a diffuse environment, taking into account two sustainability criteria: green design and disclosure of information in order to determine the best possible sustainable supplier. In that regard, four suppliers' investment budget is taken into account.
For manufacturing SMEs, Zhang et al. (2021) considered that decision-making in sustainable production is improved when there is an understanding of the interconnections between technical, environmental, social, and economic performance metrics. Marcelino-Sádaba et al. (2015) added aspects such as operation, plant levels, products and internal processes and Nicoletti Junior et al. (2021) also included marketing and finance.
Concerning energy production industries, Bautista et al. (2019) included environmental aspects such as land use, water demand, energy ratio, GHG emission savings, and emissions affecting air quality. Furthermore, Musango et al. (2012) added the community's perception of social aspects and the amount of glycerol accumulated as a result of biodiesel production within environmental aspects. Jin et al. (2019) considered the use of cellulosic fuels, including N 2 O emissions from nitrogen as a fertilizer and CO 2 emissions from cellulosic ethanol emissions. Ansari & Seifi (2012) analysed the effects of subsidies on energy prices with low, moderate, and high energy efficiency scenarios on the iron and steel industry. Kim et al. (2014) developed the DS model is geared to GHG mitigation through technology by comparing BAU and TECH scenarios (use of technology), concluding that the introduction of technologies reduces CO 2 emissions.
In order to reduce the carbon footprint, Thirupathi et al. (2019), used DS to determine sustainable areas in the automotive industry. The elements taken into account were learning and growth, technological growth, market and customer growth, and financial growth.
Similarly, there has been academic research to lead measurement and evaluation tools of sustainability practices (Reverte et al., 2016), most of them focusing on relating social and financial or environmental and financial performances. Waddock & Graves (1997) identified positive, negative, and neutral relationships between social and financial performance. Dobre et al. (2015) analyze the financial impact of environmental and social disclousure indicators in exchange-listed corporations. Reverte et al. (2016) analyzed CSR practices by considering financial and non-financial indicators and the importance of innovation as link between social and financial performance. Rintala et al. (2022) considered it essential to develop ambidexterity in logistics to improve the relationship between environmental and financial performance, promoting the simultaneous exploitation of existing competencies and explorimg new opportunities to increase organizational performance. Throughout Environmental, Social and Governance research, looking for new alternatives to mitigate environmental and social risks such as climate change and human rights and governance to obtain long-term investment returns in the long term. It operates in the financial domains, focusing on risk and financial return, while CSR operates in the corporate fields (MacNeil & Esser, 2022;Siew, 2015).
Acknowledging the multidimensional character of sustainability, develop a framework for measuring environmental performance that considers the production processes, the qualitative nature of the indicators, and the complexities in developing a synthetic indicator.  proposed a model for environmental performance evaluation based on Markov chains, which simulate the operational aspects of an industrial process. This model was later extended to include all sustainability dimensions within a supply chain (Shokravi & Kurnia, 2014). These latter two studies are noteworthy as they consider the dynamical character of an organization in measuring CSR performance.
CSR practices are also influenced by marketing and innovation strategies (Chaudhri, 2016;Padilla-Lozano & Collazzo, 2021;Revuelto-Taboada et al., 2021). For example, Pergelova & Angulo-Ruiz (2013) considered that green marketing and socially responsible consumption have a weak correlation with corporate performance. Singh (2016) studied the effect that an alliance between companies with similar ethical identities have consumer perceptions about social responsibility. Brower & Mahajan (2013) identified 35 variables related to social performance based on a study of the factors that affect demand from stakeholder groups, including marketing intensity, brand strategy, investment in research and development, among others. Broadstock et al. (2020) determined that innovation activities positively affect social performance. Veronica et al. (2020) considered that an organization's orientation towards sustainable innovation depends on both its tangible and intangible stakeholder capabilities. However, none of these studies examined marketing and innovation strategies considering the dynamic nature of an organization.
System Dynamics modeling has been employed to analyze the sustainability of a project. For example, Ozcan-Deniz & Zhu (2016) analyzed the effect that changing conditions affected the sustainability of a construction project. Duran-Encalada & Paucar-Caceres (2009) used a model to simulate the sustainability dimensions in urban development projects. Fang et al. (2017) modeled the economic, population, waste, and energy components. Jin et al. (2009) incorporated the concept of an ecological footprint into an SD model. Finally, Xu & Coors (2012) proposed a model for the analysis of residential development using sustainability indicators. However, none of these models analyzed the impacts on CSR performance over time of a corporate strategy. In the next section, a model is presented aiming to breach this gap.

Methods
System Dynamics methodology has been widely used for prospective models, this methodology includes analysis of the model components, development of the dynamic hypothesis in the causality analysis, computational modeling in the flow and level diagram, model validation, scenario runs and analysis of results. Several applications are developed for making decisions under controlled simulation scenarios in supply chain management, energy, environment, society, finance, and public policies. This paper presents a novel model of the manufacturing supply chain, as references for application in the corporate resources planning process considering combined sustainability elements.

Model design, causal loop and stock-flow diagrams
The structure of the model includes the main sectors in a typical company, such as transformation and support processes proposed by Sterman (2000), through problem articulation (see section 1), formulation of dynamic hypothesis (see Figure 2), formulation of simulation model (see Figure 3), testing (see section 3.2) and, policy desing and evaluation (see section 4). This can be adapted to other manufacturing processes, industrial, commercial, and services sectors. Figure 1 illustrates the main elements of the proposed model, which is based on a manufacturing system.
These are: (a) The direct and indirect production materials. The former refers to those raw materials necessary for manufacturing the finished product, and of greater volume or cost; while the latter refers to complementary elements such as packaging. Indirect materials are subclassified into purchased and recycled indirect materials. (b) The manufacturing process produces the final products to meet the demands of the end consumer, which is stimulated through innovation and marketing strategies. It comprises the materials, workers, and machines. (c) The Complex Performance Indicator (CPI) measures the system's performance system from social, environmental, and economic perspectives. Social performance is calculated from the worker sector. Economic performance is calculated from the results of the manufacturing process, the machine sector, and the demand of the final consumer. Finally, environmental performance is calculated from the result of the manufacturing process and the production materials used. Figure 2 shows the Causal Loop Diagram (CLD) developed using SD methodology to answer the question posed in Section 1. It has three main negative feedback loops.
The first two are the direct and indirect materials loops that control the essential and auxiliary elements for manufacturing the products. An increase in the desired production results in increased orders of either or both groups of materials, subject to revision of inventory levels and safety stock. After delays due to delivery, the inventory of either or both groups of materials increases. These loops affect the behavior of the environmental indicator. The third loop is the production loop that controls increases in the desired production. An increase in demand increases the desired production, subject to inventory levels and safety stock, which in turn triggers a request to modify the number of workers hired within a time frame. The production volume is related to each worker's productivity, and it is measured as the number of units produced per unit of time. Production volume increases the stocks of the final product, which decreases as orders are dispatched. This loop affects the social indicator through the workers' behavior and the economic and environmental indicators through the dynamics of production.  The model is divided into interconnected physical and information flow sectors as illustrated in Figure 1, whose individual structure is illustrated in the Stock and Flow Diagram (SFD) in Figure 3. The physical flow sectors are those that simulate the actions within the process that result in the manufacture of finished products. These are: • The Worker Sector (W) simulates the changes in the production staff according to production needs, through hiring and dismissal. This sector also simulates the effects of occupational diseases, beaches of ethical conduct, and changes in wages; • The Innovation Sector (I) simulates the company's innovation strategy. It aims to model the effect that innovation incentives have on product improvements affecting health and safety; • The Marketing Sector (M) simulates the company's marketing strategy. It aims to model the effect that marketing investments have on the demand for the finished product; • The Machinery Sector simulates the acquisition and disposition of machinery, depending on the needs of the manufacturing process; • The Production Sector (P) simulates the manufacturing process. Its inputs are the direct and indirect materials, and its output is the stocks of the finished products, as required by market demand; • The information flow sectors calculate the Key Performance Indicators (KPI) derived from physical flows. The KPIs are based on those proposed by Pavláková Dočekalová & Kocmanová (2016), which are: • Environmental KPIs (EnvI) measures material and energy consumption, waste production, CO 2 emissions and other environmental costs, based on the equations from Table 1; • Social KPIs (SocI) measures the proportion of workers in collective agreements, the frequency of professional illnesses, products that impact safety and health, identification of client needs, salaries, and failures in the code of ethics, based on the equations from Table 1; • Economic KPIs (EcoI) measures cash flows and return on investment, based on the equations from Table 1; • Sector CPI calculates the consolidated indicator derived from the three above, based on the equations from Table 2; Economic performance indicator EcoI = (0.708 × ecoKPI1) + (0.292 × ecoKPI2) Complex Performance Indicator Each of these consolidated indicators consider weights defined as a result of the research by Pavláková Dočekalová & Kocmanová (2016) and detailed in Table 2. Corporate KPIs are calculated through the benchmark studies in Pavláková Dočekalová & Kocmanová (2016), but are not considered in the model developed in this paper. Due to the case study scope and the availability of the information.
The model was constructed using the software iThink. A detailed list of equations of the model is provided in the Appendix A, whereas the parameters used are provided in Appendix B. These parameters were collected from the public databases of Colombian governmental entities, such as the National Administrative Department of Statistics (DANE), the National Planning Department (DNP), the Superintendence of Industry and Commerce (SIC), the National Association of Entrepreneurs of Colombia (ANDI), and the Federation of Colombian Insurers (Fasecolda).

Model validation through scenario and sensitivity analyses
The model's consistency, confidence, and robustness were validated following a multipronged approach. Initially, the structure and behavior of the model were checked for dimensional consistency and sensitivity as defined by Barlas (1996) and Qudrat-Ullah & Seong (2010). For the former, using the "Check Units" tool included in iThink, the equations' units of measurement are verified for consistency. For the latter, the upper and lower bounds for each variable are defined. This upper and lower bounds represent significant changes in the behavior of the model, added to the adjustment capacity in the planning of corporate strategies, which shows, on the one hand, the robustness of the model's behavior, and on the other hand, the viability in the implementation of the actions to be executed. Then, the model is tested using these extreme conditions. The model presented is a proposal for the sustainability analysis in the manufacturing planning processes, considering the main data collection of the Colombian manufacturing sector, for this reason, the validation is not focused on the historical data of a particular company.
Once these tests were successful, the model was checked for robustness and sensitivity using scenario analysis. A scenario is defined as a broad set of initial conditions and changes applied during a simulation to observe the responses of the model, a corporate strategy is defined as a percentage change made to the initial values of one or more variables of the model, and an option is defined as the individual settings of each variable, which are used to observe the behavior over-time of the system KPIs. The analysis focuses on the effects that a corporate strategy implemented by management has on the behavior of the Environmental Indicator (EnvI), Social Indicator (SocI), Economic Indicator (EcoI) and Complex Performance Indicator (CPI), during a five-year planning horizon (in manufacturing planning, a five-year time is considered a long-term period for the decision-making process), assuming initial conditions corresponding to the Colombian manufacturing sector, as mentioned in Section 3.1. The results are reported as the accumulated average of each KPI, aiming to observe convergence at the end of the planning horizon. Moreover, the results are also indicators of the viability of the strategy under those conditions. Table 3. shows three global scenarios and its strategies, which include both increases in investment levels and optimistic and pessimistic changes in demand.
Through the system dynamics methodology, scenarios are analyzed, ranging from the base scenario or the current situation, scenarios that evaluate investment possibilities, to scenarios with changes in the future behavior of the environment (Becerra-Fernandez et al., 2020). The scenarios included in this research are:  (Hickel et al., 2021;Hickel & Kallis, 2020;Latouche, 2012;Lehmann et al., 2022;Petschow et al., 2020) demand of the finished product, with one, both or none of the PM and IMM strategies implemented. Percentual changes in demand were based on the annual growth rate of the industrial value-added of the OECD national accounts historical data. Figure 5 illustrates the variables in the system adjusted in this scenario.
• Business as Usual (BAU) is the base-case scenario, where average growth in demand is experienced but none of the proposed strategies are implemented; • Business as Investment (BAI) considers increased levels of investment through production management (PM) and innovation and marketing management (IMM) strategies. Figure 4 illustrates the variables in the system modified in this scenario. PM strategies affect both the Production (P) and Workers (W)    Once these scenarios were completed, those options that resulted in the highest change in the cumulative average CPI are taken as reference for further analysis of the model results.

Outcomes of the sensitivity analysis
Sensitivity analysis is performed for the relevant variables of the model, the detail is not presented but the result is expanded through the following section. Sensitivity analysis allows to observe the behavior of the model Figure 7 shows the results for the best options for the production management strategies in the BAI scenario. At the end of the planning horizon, EnvI improves 5.1% (from 5.5 to 5.79) when Sector W variables are modified; SocI improves 138% (from -4.3 to 1.65) when Sector W variables are modified or with the combined PM strategies. EcoI improves 2.6% (from 73.1 to 77.21) with the combined PM strategies. Finally, CPI improves 30.6% (from 11.3 to 15.7) with the combined PM strategies. Worker-related strategies have the highest effect on CPI, with minor improvements when combined with production-oriented strategies. Figure 8 shows the results for the best options for the innovation and marketing management strategies in the BAI scenario. EnvI improves 6.9% (to 5.89) when Sector I variables are modified. SocI improves 12.4% (to -3.80) when Sector I variables are modified or with the combined IMM strategies. EcoI improves 0.3% (to 73.25) when Sector W variables are modified. Finally, CPI improves 3.4% (to 11.7) when Sector I variables are modified, meaning that implementing innovation strategies alone has the necessary effect, and a combination with marketing strategies have a minor additional effect. Figure 9 shows the results for the best options for the BAV scenario, considering an optimistic growth in demand. EnvI improves 11.1% (to 6.12) when IMM strategies are implemented. SocI improves 150.7% (to sectors in response to changes in the defined variables. The resources of the processes are finite and changes in their allocation represent economic efforts for the companies. In the results section, the amount of variation that significantly impacts the sustainability indicators, allows prioritizing the allocation of resources. In the following section, the best results from the sensitivity analysis are analyzed during the sixty-month planning horizon, where for each KPI is presented as a percentage increase from the BAU conditions. 4.2. Analysis of the best options during the planning horizon Figure 6 illustrates the results of the BAU scenario, where no strategies have been implemented and demand growth is based on historical performance. As such, BAU is used to compare the results of all other strategies. At the end of the planning horizon, EnvI reaches a value of 5.5, SocI reaches a value of -4.3, EcoI reaches a value of 73.1, and CPI reaches a value of 11.3.  -2.2), EcoI improves 6.4% (to 77.72) and CPI improves 37% (to 15.49) when AS strategies are implemented, meaning that significant effects on CPI are obtained with production management strategies only. Contrary to expectations, innovation and marketing strategies did not significantly improve CPI. Figure 10 shows the results for the best options for the BAV scenario considering a pessimistic growth in demand. EnvI improves 6.4% (to 5.86), SocI improves 150.7% (to 2.2) and CPI improves 33.7% (to 15.12) when AS strategies are implemented. Meanwhile, EcoI improves 4.2% (to 76.12) when PM strategies are implemented. As observed in the optimistic demand scenario, CPI improves mainly due to production management strategies. Table 4. shows the consolidated results by KPI according to the changes by sector and the implementation of strategies. In boldface are the results with the highest values, which are obtained for the BAV scenario by combining strategies and optimistic demand growth, and for the BAI scenario by applying the PM strategies. Figure 11 show the best results for each KPI depending on the implemented strategy. Figure 11a corresponds to EnvI, whose best result of 6.12 is obtained with the IMM strategies with an optimistic growth in demand, representing a 3.9% increase compared to modifying the Sector I variables in the BAI scenario (a value of 5.89), and an 11.1% increase compared to the BAU scenario. These results indicate that implementing innovation initiatives always improves EnvI; however, marketing actions are required in conditions of growing demand, and production initiatives are required in conditions of falling demand. Figure 11b corresponds to SocI, whose best result of 2.2 is obtained by implementing all strategies in both optimistic and pessimistic demand scenarios. This represents a 33.3% increase compared to modifying the Sector W variables or implementing the PM strategies in the BAI scenario (a value of 1.65 in both cases), and an increase of 150.7% compared to the BAU scenario. These results indicate that SocI remains stable with changes in demand. Improvements can be obtained by combinations of production, innovation, and marketing initiatives, and to a lesser extent with worker-related initiatives.  Figure 11c corresponds to EcoI, whose best result of 77.72 is obtained by implementing all strategies with an optimistic growth in demand, representing an increase of 0.6% compared to implementing the PM strategies in a BAI scenario (a value of 77.27), and an increase of 6.4% compared to the BAU scenario. These results indicate that EcoI improves with production initiatives, even when demand for finished products drops. Increases in demand require the combination of these initiatives with innovation and marketing ones to obtain better performance.

Consolidated results by indicator
Finally, Figure 11 d) corresponds to CPI, whose best result of 15.49 is obtained by implementing all strategies with an optimistic growth in demand, representing an increase of 2.8% compared to implementing the PM strategies in a BAI scenario (a value of 15.07), and an increase of 37% compared to the BAU scenario. These results indicate that CPI improves with higher levels of investment in production strategies in conditions of growing demand. Moreover, production initiatives have a positive impact on all indicators, including CPI.

Discussion
The contribution of this research focuses on a model with feedback that interrelates the dimensions of sustainability with strategy planning supports decision making in prospective, multidimensional, and sometimes counter-intuitive ways. For example, contrary to expectations, both innovation and marketing strategies did not significantly improve CPI in an optimistic demand growth trend. These observations coincide with similar SD models applied in supply chain (Sterman, 2000), sustainability (Bockermann et al., 2005), organizational performance management (Bianchi & Rua, 2017), applications in the mitigation of emissions in the energy supply (Cardenas et al., 2016), among others. As such, decision-makers can establish investment priorities or halt proposed changes, if modeling shows undesirable and unexpected results.

Conclusions
This paper presented a System Dynamics based model for assessing strategies of production, (with 39 variables), innovation (with 7 variables), and marketing (with 5 variables), accounting for the demand for final goods, by measuring their impact on the sustainability of a manufacturing process, measured through environmental, social, and economic performance indicators. The model was validated through several scenarios, considering changes in demand and strategy. The results demonstrate the suitability of the model as a decision-support tool for sustainability planning in a corporate environment. The model proposes a support tool for decision-making in manufacturing companies with similar characteristics; the proposed structure provides a reference for its application in service companies. That is, it allows the designers to observe the behavior of the performance indicators over time, and decide investment priorities, changes to be scaled back, or contingency plans to be implemented, given the available resources, the business vision of the company, and optimistic and pessimistic changes in the market. Through the model, given some initial conditions, planning horizon, and average demand growth, it was observed that environmental performance can be improved through the implementation of innovation strategies. If demand is growing at an above-average rate, environmental performance can be improved through the combination of innovation and marketing strategies. On the other hand, social performance can be substantially improved under diverse conditions by implementing strategies that increase worker welfare and equity in hiring both men and women. Production strategies, such as lowering costs and increasing the usage of recycled materials improved the economical performance under average demand growth conditions. However, with above-average growth, the implementation of production, marketing, and innovation strategies did not provide a substantial increase in economic performance. Finally, analyzing the sustainability performance in aggregate, the improvement of working conditions, and reduction of waste had the most impact throughout the planning horizon. However, with above-average growth, the implementation of production, marketing, and innovation strategies had the largest effect on aggregate performance. Nevertheless, the model has some considerations. Firstly, the weights used for each indicator, as presented in Table 2., are the same as those proposed by Pavláková Dočekalová & Kocmanová (2016). These weights were based on the Czech manufacturing sector and may require adjustments for new conditions. Secondly, the model does not implement the corporate governance indicator proposed by Pavláková Dočekalová & Kocmanová (2016), as some of the data required was not available on the GRI reports or the systems were hard to model. Thirdly, the model places low emphasis on energy consumption and emission levels, which are deemed to have a significant impact on the environmental performance of a corporation. These important limitations will be tackled in further work, along with further analysis of the variables on the model that could be exposed to the decision-makers' control. Moreover, multi-criteria optimization methods could be used to automatically determine the best strategy, given the existing conditions. Besides, the model can be adapted to corporations with other characteristics, such as service providers. Additionally, concepts such as regenerative capitalism can be included to broaden the perspective of the model (Elkington, 2020).