1 Introduction

The operational landscape of global supply chains (SCs) has grown increasingly uncertain and susceptible to severe and frequent disruptions (Nikookar and Yanadori 2022). Disruptions, ranging from minor disturbances to major upheavals, profoundly impact the flow of goods, materials, and services (Scholten et al. 2020). In recent years, the global COVID-19 pandemic, inflationary pressures, and disruptions at key logistical bottlenecks had severe impacts on SC performance (Dierker et al. 2024; Ivanov 2020) and reportedly three-quarters of organizations annually encounter supply chain disruptions (Scholten et al. 2020). The frequency of disruptions and the heightened uncertainty highlight the necessity for enhancing the resilience capabilities of the operations and planning of SCs. Resilience refers to the capacity of SCs to persist, adapt, or transform in the face of disruptions (Wieland and Durach 2021). It can also be understood as the adaptive capability to reduce the probability of disruptions, resist the propagation of disruptions, and deploy response actions for the recovery and enhancement of SC performance (Kamalahmadi and Parast 2016; Ponomarov and Holcomb 2009).

Resilient SCs distinguish themselves by being less vulnerable to disturbances and more adept at coping with unexpected events (Guntuka et al. 2023). These attributes can be found both at the individual firm level and the SC level. Although SC resilience can be approached from multiple perspectives, a core focus of SC resilience lies in maintaining the essential function of supplying goods to end customers (Sá et al. 2019). Thus, resilience can be measured in terms of the ability to maintain supply to meet end customer demand (Behzadi et al. 2020). This ability is particularly vital for SCs dealing with perishable products, where a swift recovery can significantly prevent the spoilage of products (Prakash 2022). Based on this conceptualization, resilience can be operationalized quantitatively as the time required to recover from disruptions and restore operations (Guntuka et al. 2023).

Flexibility (e.g., Li et al. 2023) and visibility (e.g., Brandon-Jones et al. 2014; Jiang et al. 2023) have been identified as key high-level capabilities for building resilience. Flexibility refers to the capability of SCs to adapt production volume and variety through leveraging response actions such as multi-skilled workforces, buffer capacity, modularity in products and processes, and retaining multiple sourcing options (Chowdhury and Quaddus 2017; Pettit et al. 2013; Kazancoglu et al. 2022). Similarly, the visibility capability focuses on increasing the effectiveness and flow of information, coordination, collaboration, encompassing real-time data collection, processing, and accessibility (Ivanov 2022; Statsenko et al. 2023; Yaroson et al. 2021). Roy (2021) emphasizes the role of visibility in integrating supply chain actors and enhancing overall coordination and performance.

Commonly adopted Resource-Based View (RBV) posits that incorporating flexibility and visibility capabilities uniformly enhances the SC resilience (Razak et al. 2023; Essuman et al. 2022; Nandi et al. 2020). RBV has been instrumental in exploring SC resilience, and its application raises important questions regarding the homogeneity of resource impact across different organizational contexts (Barney 2012). It posits that firms can attain a competitive advantage by strategically managing resources that are valuable, rare, and difficult to imitate (Barney 2012). Our study takes the contingent RBV to SC resilience, outlined by Brandon-Jones et al. (2014), and we propose that the effectiveness of flexibility and visibility capabilities is substantially context-dependent. For instance, visibility as a resilience capability can be integrated into supply chain frameworks through activities like product tracking, efficient information flow, and predictive analytics. Internal and external contingencies can moderate the influence of these resources in nuanced ways (Barney et al. 2021), but the boundary conditions of these disruption management resources are not as well known.

Brandon-Jones et al. (2014) found that SC visibility enhances resilience, but the effect can be moderated by SC complexity. We take the contingent view on resilience impact of flexibility and visibility moderated by speed of the operational dynamics. SCs can be generally characterized by the responsive (fast) or efficient (slow) matched by the types of products that they focus on (Fisher 1997). We study how these fast/slow characteristics impact the resilience effect of capabilities in various disruption events. By highlighting these contingencies, the study aims to provide supply chain managers with insights into when and where specific disruption management resources are particularly beneficial in building resilience.

In line with this perspective, our study poses the research question: "Is the effectiveness of resilience capabilities (flexibility and visibility) and associated response actions dependent on the operational dynamics of the SC?" Response actions refer to the actions that are linked to specific resilience capabilities. They are based on our empirical grounding of the simulation analysis and highlight the dynamic and reactive use of these capabilities in practice. To address this question, we employ a system dynamics simulation modeling grounded on the observations of response actions deployed in the Finnish grocery SCs during the COVID-19 pandemic (Saarinen et al. 2020). The observations of the severe disruption impacts and the deployed response actions provide relevant practical motivation for the study of the contingencies of developing SC resilience. We utilize these empirical observations in a stylized system dynamics model, grounded in empirical data (Appendix A), to explore counterfactual scenarios. Our model builds and extends on previous models studying supply chain dynamics and the bullwhip effect (Fransoo and Udenio 2020; Sterman 2000; Udenio et al. 2015).

We make contributions from two perspectives. Firstly, we demonstrate the pivotal role of supply chain speed in influencing the effectiveness of resilience enhancing capabilities and resources. Our findings specifically reveal that flexibility in increasing capacity significantly enhances resilience in the fast supply chains. In contrast, for the slower supply chains, visibility emerges as the most effective capability. The results underscore the importance of aligning resilience strategies with the operational tempo of the supply chain. This contribution is an elaboration of the contingent resource-based view of supply chain resilience. Secondly, we identify the role of visibility, resultant increased information sharing and planning coordination to avoid and mitigate destocking in response to dropping demand and phantom ordering in response to scarcity. These two observed responses to disruption will substantially undermine resilience in specific contexts. With these results we provide insights for supply chain managers in preparing for and responding to future crises.

2 Literature review

2.1 Theoretical framework

The concept of resilience in SC literature has garnered significant attention in recent years, particularly in the aftermath of high-impact global disruptions (Castillo 2022; Hohenstein et al. 2015; Ponomarov and Holcomb 2009). While various definitions of SC resilience exist, the most widely adopted definition is the system's ability to revert to its original state or move to a better state after disturbances or disruptions (Christopher and Peck 2004). An improvement in state can be demonstrated through growth in market share, improved financial results, and exceptional customer service (Hohenstein et al. 2015). Subsequently, resilience has also been defined as the capability to effectively deal with unavoidable risks or disruptions, gain additional resources and thus improve the competitive advantage in the process (Aouag et al. 2020; Sheffi 2005).

SC resilience can be divided into four phases: readiness, responsiveness, recovery, and growth. These phases were initially introduced by Ponomarov and Holcomb (2009) and Proag (2014), with the fourth phase being identified by Hohenstein et al. (2015). The readiness phase encompasses the preparation strategies that are put in place to prevent the occurrence of risks and reduce the impacts of potential disruptions in the event of risk realization. Responsiveness allows for appropriate reactions and adaptation to risk-bearing events, often in the form of swift redesign of SC processes and reallocation of resources. Recovery builds on the response layer and refers to the process of returning to at least the pre-disruption operational parameters. Growth refers to improving capabilities from the pre-disruption state by learning and adaptation to build capabilities (Ali et al. 2017; Hohenstein et al. 2015).

Firms strategically plan for SC resilience by development of specific capabilities tailored to various phases of resilience. These capabilities are instrumental in enabling SCs to effectively respond to diverse disruptions. For instance, flexibility is a resilience capability, which empowers the SCs to swiftly mobilize additional resources or reallocate existing resources to mitigate bottlenecks arising from disruptions (Iftikhar et al. 2021). This overarching capability encompasses a spectrum of capabilities and resources, representing response actions that SCs can adopt and implement for mitigation of disruptions. Examples include maintaining a multi-skilled workforce, flexible capacity, backup suppliers, and diversified sourcing strategies (Ali et al. 2017; Scholten et al. 2014). Similarly, visibility, encompassing collaboration, coordination, and data sharing, stand out as key capability for bolstering SC resilience, referring to enhancing information flow and fostering increased cooperation throughout the SC (Jiang et al. 2023; Jüttner and Maklan 2011; Scholten and Schilder 2015).

Resource-Based View (RBV) theory has been extensively deployed to gain insight into SC resilience (Barney 1991). RBV posits that firms can attain competitive advantage by possessing a set of resources that are valuable, difficult to imitate, and lack substitutability (Barney 2012). This theory has been widely applied within the realm of supply chain management (Chowdhury et al. 2019; Essuman et al. 2022; Nandi et al. 2020). A contingency perspective is often added to the RBV theory, which states that the effectiveness of adopted resources is dependent on the internal and external contingencies (Brandon-Jones et al. 2014). They developed the contingent RBV theory on resilience and highlighted how the supply chain structure and design of production systems can impact the value of resources and capabilities for resilience building.

To address the environmental dynamism the dynamic capabilities view on SC resilience has been identified as useful in recent literature (Barney et al. 2021). Dynamic capabilities, in this context, refer to the integration and capacity for the reconfiguration of both internal and external resources to effectively respond to rapidly changing environments. Dynamic capabilities linked to enhanced resilience are, for example, adaptable planning and information processing capabilities (El Baz and Ruel 2021) or ad-hoc structuring and orchestration of new supply chains (Müller et al. 2023). Moreover, the possession of these dynamic, unique, and heterogonous set of capabilities aid SCs in combating threats and maintaining firm performance while facing disruptions (Sabahi and Parast 2020). These extended perspectives have gained widespread acceptance in SC resilience and management studies (Huang et al. 2023; Modgil et al. 2021; Müller et al. 2023; Yu et al. 2019).

We draw on the contingent RBV and dynamic capabilities theories to investigate our research questions. We contribute to the literature by studying the effectiveness of resilience capabilities adopted by the SC managers in response to the disruptions that unfolded during the global pandemic shock in grocery supply chain. We focus on the varying effectiveness of these capabilities based on different operational dynamics of the SCs, namely how responsive and efficient SCs differ in priority of developing flexibility and visibility capabilities.

2.2 COVID-19 and grocery supply chains

The COVID-19 pandemic was a unique disruption event with global reach and multiple interlinked disruptions with cascading effects over time. Prior to the pandemic, SC research had predominantly focused on localized disruption scenarios. Studies addressing potential epidemics and pandemics mainly focused on medical SCs, while in other SC contexts, the discussion has particularly focused on selecting resilient suppliers (Golan et al. 2020). Recent research has increasingly focused on the preparedness, response, and recovery aspects of SC resilience (Chowdhury et al. 2021).

Post-pandemic resilience literature is now focusing on the resilience capabilities that were effective during the disruptions and added value (Jiang et al. 2023; Li et al. 2023; Müller et al. 2023). For instance, Nikookar and Yanadori (2022) discovered that collaboration between SC managers strengthened their resilience against disruptions. Similarly, a case study by Spieske et al. (2022) found that the key capabilities for maintaining performance were agility, collaboration, and flexibility. Similarly, grocery SCs are responsible for delivering food products to consumers while ensuring quality, safety, and preservation. The perishability of products adds further complexity to these SCs as it is challenging to add buffers against demand fluctuations and transportation alternatives (Zhong et al. 2017). Perishability factors into the classification of grocery SCs into two types: fast and slow grocery SCs. Fast SCs deliver food products that have higher perishability and require shorter lead times and more frequent replenishments compared to slow SCs. Durable food products, such as many consumer-packaged goods, allow for slow and efficient SCs that are characterized by longer lead times, more geographical dispersion, and inflexible capacity.

The pandemic disruption posed significant challenges to the grocery SCs as food availability was threatened. As a result, grocery SCs adopted a wide range of resilience approaches, including ramping up production capacities, enhancing visibility across the SC, identifying potential disruptions, geographically diversifying suppliers, building and increasing collaboration with stakeholders, prioritizing and focusing on essential activities, increasing standardization of activities, and increasing package sizes (Saarinen et al. 2020; Ivanov and Dolgui 2021; Leite et al. 2020; Lozano-Diez et al. 2020; Paul and Chowdhury 2020).

To examine SC resilience capabilities in grocery SCs, we characterize two contrasting types of SCs for our simulation study. The fast and slow SC represent the typical differing operational dynamics characterized by the replenishment frequency and speed of the material flow, resulting in a varying set of challenges and opportunities for each type of SC. This characterization is analogous to the seminal categorization of SCs to responsive and efficient by Fisher (1997). We use these stylized and contrasting SCs to study the effectiveness of varying observed response actions, linked to the two general capability categories, under diverse disruption scenarios with system dynamics simulations.

2.3 System dynamics modelling

System dynamics (SD) modeling is a tool for studying complex SC systems with dynamic feedback mechanisms, time delays, and time-dependent behavior (Sterman 2000; Sterman et al. 2015). With its ability to study dynamic, structure dependent and complex systems, SD modeling is a suitable method for studying disruptions to SCs (Petropoulos et al. 2023). Feng (2012) developed a SC system dynamics model to demonstrate the benefits of information sharing among SC echelons. Udenio et al. (2015) examined SC dynamics of the Dutch manufacturing industry following the worldwide financial crisis of 2008. Wang et al. (2020) and Heidary (2022) used SD modeling to evaluate the impacts of COVID-19 on Chinese and Iranian SCs, respectively. Similarly, Udenio et al. (2015) extended their system dynamics model to analyze the SCs disruptions caused by the COVID-19 crisis (Fransoo and Udenio 2020). Georgiadis et al. (2005) constructed a multi-echelon food SCs model to analyze the key challenges of strategic SCM in long-term capacity planning. Özbayrak et al. (2007) developed a four-echelon SC model, which was designed around a medium-sized manufacturing company, to compare the outcomes of different real-life operational scenarios.

We utilize a simulation study to assess the variability in the effectiveness of response actions, which is attributable to the distinct operational dynamics of the disrupted SCs. SD simulation enables addressing two interesting aspects linked to our research problem using general and previously validated SC models for disruptions. First, what were the resilience enhancing impacts of the response actions that were observed to be deployed in practice. Are they all effective and how does the effectiveness vary with different disruption events? Second, how do the structure and different characteristics of supply chains impact the effectiveness on enhancing resilience. Resilience can be seen as a complex and abstract construct that is difficult to measure, especially in the context of rare but high-impact disruption events like the COVID-19 pandemic. As a result, testing resilience strategies empirically under real disruption conditions poses a significant challenge. Simulation provides a valuable methodology with a controlled environment wherein the effectiveness of resilience strategies can be isolated and assessed.

Operationalizing resilience within this study, we concentrate on metrics such as Time to Recovery (TTR) and the capacity to meet customer demand post-disruption. TTR, as a quantitative resilience metric, aligns with the conceptual definition of resilience as the capability to quickly and effectively recover from disruption (Behzadi et al. 2020). Empirical studies that estimate SC recovery time as a dependent variable are scarce with lacking evidence of actual measures of TTR (Guntuka et al. 2023). Building upon existing literature, we adopt an extended approach to evaluate resilience that combines initial loss in system performance, time until recovery, and maximum acceptable recovery time (Zobel et al. 2021; Bruneau et al. 2003; Pettit et al. 2013; Proag 2014; Zobel et al. 2021). This operationalization serves as the foundation for our investigation into various response strategies to enhance SC resilience, particularly in contexts characterized by high disruption susceptibility due to SC complexity and lean focus (Nikookar and Yanadori 2022).

We augment SC resilience literature and practice in three ways. First, we examine the effectiveness and importance of concrete disruption response actions drawing on real-world evidence on disruption event studies. This aids identification and prioritization of response actions. Second, we investigate how the effectiveness of these response actions varies across different product types and supply chain configurations. Our analysis reveals novel contingency of response action effectiveness, underscoring the importance of context in SC resilience strategies. Third, with these results we elaborate and extend the current understanding of contingent RBV and Dynamic Capability literature on SC resilience.

3 Data collection and research methodology

We employ SD modeling as the research method, building on the established tradition and published models for multiechelon SCs from previous studies (Fransoo and Udenio 2020; Sterman 2000; Udenio et al. 2015). SD models use feedback loop structures built on differential equations to analyze system behavior particularly focusing on delays and feedback, and how the dynamic behavior is dependent on the structure of the system and deployed control policies. This allows for the study of complex systems with interdependent actors. The model used in this study is based on the well-established and widely recognized SC model of Sterman (2000), that was extended by Udenio et al. (2015) and Fransoo and Udenio (2020) for analysis of the impact of disruptions in supply chains. The disruption scenarios and the response actions were gathered from qualitative and observational research studies (Saarinen et al. 2020; Aday and Aday 2020; Chenarides et al. 2021). These studies described the disruptions and response actions but did not quantify the impacts or the potential impacts of non-action. Furthermore, we accessed demand index data from Finnish grocery retailers that correspond to 83% of the sales volume of the total market size (Appendix A). This data was used to inform the demand disruption scenarios.

3.1 Disruption scenarios

SC risks can be classified as operational and disruptive risks. The former addresses the risks related to the day-to-day operations, while the latter deals with the low-frequency, high-impact events. Disruption risks have highly significant and abrupt impacts throughout the supply chains and can potentially halt operations for considerable time periods. COVID-19 was a unique disruptive event due to its magnitude, rapid propagation, uncertain period, and long-lasting effects (Ivanov 2020). Next, we describe the most significant disruption impacts that will be studied through simulation.

  • The purchase patterns changed because of changed routines due to lockdowns and work-from-home policies. During the initial stages of the pandemic, grocery stores observed a significant rise in demand for items with long shelf lives as daily food cooking became more convenient. In contrast, the demand for pre-made food products decreased (Aday and Aday 2020). The demand for fresh bread and vegetables in the European nations, during the week when the pandemic was declared, rose by 76% and 52%, respectively (Barman et al. 2021).

  • Panic buying and hoarding was observed worldwide, despite the assurances given by governments as well as private sectors regarding the ability to meet the demand. The rapid and unanticipated increase in demand resulted in certain short-term stock-outs in grocery SCs.

  • The demand for food service meals reduced steeply owing to the closure of restaurants, bars, and other dining facilities (Chenarides et al. 2021).

  • During the pandemic, e-commerce became popular as consumers adopted online shopping due to the implementation of lockdowns and consumers’ preference to avoid crowded locations (Thilmany et al. 2021). This unexpected change also created specific logistics bottlenecks as distributors’ capacity was insufficient to fulfil the rising demand (Saarinen et al. 2020).

  • Labor shortages were observed across the globe as mobility restrictions and isolation policies were implemented, and workers reported illnesses. The situation was further exacerbated due to the employment of seasonal laborer’s across the industry, especially in developing countries (Burgos and Ivanov 2021; Aday and Aday 2020). The situation is evident from the fact that India, the world’s biggest rice exporter, had to suspend its exports in 2020, and Vietnam had to reduce its exports by 40%, mainly due to a lack of workforce.

  • The social distancing rules, as well as other similar policies, significantly reduced the capacity of the manufacturers and other players to operate at total capacity, resulting in disruptions (Burgos and Ivanov 2021; Thilmany et al. 2021).

  • Phantom ordering and inflated orders were observed across the SC from retail to manufacturers (Saarinen et al. 2020), with actors trying to order significantly more than their demand as a measure of redundancy in increasing inventories and as a hedge against scarcity of supply. Phantom ordering had significant potential adverse impact as it increased strain on suppliers and manufacturers and intensified the bullwhip effect.

3.2 Resilience enhancing response actions

Firms across the grocery SCs deployed rapid response actions that provided an opportunity to study and learn on resilience enhancing practices. A sample of the observed response actions from the literature in the context of grocery SCs are discussed below:

  • Sharp increase in demand for grocery products led to logistics bottlenecks constraining the replenishment of retail stores. The response was to increase the SC throughput by prioritizing and protecting the bottlenecks, first of logistics capacity through dynamic resource reallocation and resource increase. For instance, employees were moved to work in bottleneck operations such as retail store replenishment, distribution warehouses, and e-commerce services. In addition, retailers collaborated with suppliers to leverage their transportation capacity and to deliver directly to the stores, bypassing logistics hubs (Saarinen et al. 2020).

  • Firms invested in information processing technology that enables data sharing and automated decision making. The crisis spurred further gathering, sharing, and utilization of POS data in the supply chains (Saarinen et al. 2020). Firms reported activating daily sharing of information and increasing resources to data sharing together with new collaborative planning (Saarinen et al. 2020). The retailers, logistics managers, suppliers and other key players increased information exchange on demand, operation plans, and capacity through daily meetings and conference calls. Monitoring measures were also enhanced to ensure visibility across supply chains.

  • The pandemic also created bottlenecks at the production and supply echelons the of SCs as the pipeline inventories were depleted. In coordination with the retailers, the manufacturers identified the products where demand was rising and prioritized their production. The coordinated planning changed the SC control to bottleneck and push control mode from the replenishment order-based control mode. This enabled the subordination of other resources to the SC’s bottleneck resource to maximize output and product availability for consumers. Manufacturers and retailers also restricted the portfolio being produced to maximize effective capacity. Finally nominal production capacity was increased through employee reallocation and the activation of secondary supply sources.

  • During the initial stages, the SCs realized the importance of transparency and information sharing downstream, towards the customer end. The lack of this transparency might lead to inconsiderate ordering by the retailers, creating further disruptions and bottlenecks due to bullwhip effects. However, as the retailers could observe suppliers' inventory levels, production schedules, shipping status, and their ability to respond to SC disruptions, they could order more prudently across the pool of their suppliers.

These response actions correspond to two major SC resilience capabilities; flexibility and visibility. Flexibility refers to the ability to engage additional resources to upscale or modify the production and supply of products. Similarly, visibility refers to the increased information sharing and transparency related response actions. Thus, modeling of the mentioned response actions enables the analysis of flexibility and visibility as SC resilience capabilities.

3.3 Supply chain system dynamics model

Our model is derived from the SC model of Sterman (2000), and it consists of three echelons: supplier, manufacturer, and retailer. The three echelons have been denoted as E3, E2, and E1, respectively. The information flows from E1 to E3, and the goods flow from E3 to E1. Each echelon further comprises three decision areas: forecasting and ordering, manufacturing, and delivery (See Fig. 1).

Fig. 1
figure 1

Overview of a single echelon of the supply chain model

Full size image

3.3.1 Forecasting

Forecasting and ordering deals with creating forecasts, ordering, and receiving demands, and ordering and receiving goods. The demand is forecasted using first-order exponential smoothing of the actual order rate. The difference between the incoming order rate (On−1) and the previous forecast (Fn) is accumulated and updated accordingly. If the incoming order rate exceeds/undercuts the previous forecast, it is updated upwards/downwards. Forecast is smoothed by updating with the difference divided by time to update forecast. Using exponential smoothing in forecasting is common within firms and provides a useful model of the actual process of forecast updating.

$$Forecasted\ Order\ Rate=\int \left( Change\ in\ Sales\ Forecast,{Forecasted\ Order\ Rate}_{t0}\right)$$
(1)
$$Change\ in\ sales\ forecast=\frac{Incoming\ order\ rate- Forecasted\ order\ rate}{Time\ to\ average\ forecast}=\frac{O_{n-1}-{F}_n}{\tau_n(F)}$$
(2)

3.3.2 Production

The production part of each echelon maintains the inventory and regulates production. The Incoming material (In) equals the delivery rate from the immediate upper echelon (Dn + 1), smoothened over the incoming delivery lead time (LTn).

$${I}_n=\mathit{\operatorname{Max}}\Big(0, Smooth\left({D}_{n+1,}{LT}_n\right)$$
(3)

Incoming material is classified as the Work in Progress (WIPn), before it is used in production.

$${WIP}_n={I}_n-{P}_n$$
(4)

The supply line (SLn) keeps track of the orders, that have been placed but their receipt is pending, while on-hand inventory (Sn) is the difference of the delivery rate (Dn) and production rate (Pn).

$$\frac{d}{dt}\ {SL}_n={O}_n-{I}_n$$
(5)
$$\frac{d}{dt}\ {S}_n={P}_n-{D}_n$$
(6)

The production rate (Pn) is modelled using a first order delay, where the production is delayed by the production lead time (PTn), and the production is limited by the available production capacity at the echelon.

$${P}_n=\mathit{\operatorname{Min}}\Big({c}_n, Delay\left({I}_n,{PT}_n\right)$$
(7)

The desired supply line \(\left({SL}_n^{\hat{\mkern6mu}}\right)\), desired on hand inventory level \(\left({S}_n^{\hat{\mkern6mu}}\right)\), and desired Work in Process \(\left({WIP}_n^{\hat{\mkern6mu}}\right)\) are obtained as exhibited in Eqs. 8, 9 and 10.

$${SL}_n^{\hat{\mkern6mu} }={F}_n\ast {LT}_n$$
(8)
$${S}_n^{\hat{\mkern6mu} }={C}_n^{\hat{\mkern6mu}}\ast {F}_n$$
(9)
$${WIP}_n^{\hat{\mkern6mu} }={PT}_n\ast {F}_n$$
(10)

Where LTnis incoming delivery lead time, Fnis sales forecast, \({C}_n^{\hat{\mkern6mu} }\) is on-hand inventory coverage, and PTnis production lead time.

3.3.3 Ordering

The supply line, on-hand stock, and WIP (work in progress) are regulated through a negative feedback mechanism. The supply line adjustment time, stock adjustment time, and WIP adjustment time signify the time allowed for those values to reach desired levels.

Supply line adjustment of orders, Stock adjustment of orders, WIP adjustment of orders, and outgoing orders is exhibited in Eqs. 11, 12, 13 and 14.

$${O}_n(SL)=\frac{SL_n^{\hat{\mkern6mu} }-{SL}_n}{\uptau_n(S)}$$
(11)
$${O}_n(S)=\frac{S_n^{\hat{\mkern6mu} }-{S}_n}{\tau_n(S)}$$
(12)
$${O}_n(WIP)=\frac{WIP_n^{\hat{\mkern6mu} }-{WIP}_n}{\tau_n(WIP)}$$
(13)
$${O}_n=\mathit{\operatorname{Max}}\left(0,{F}_n+{O}_n(SL)+{O}_n(S)+{O}_n(WIP)\right)$$
(14)

3.3.4 Delivery

The delivery area monitors incoming orders and backlogs. Backlogs (Bn) keeps track of the outstanding orders and is calculated as the difference between the incoming order rate and the delivery rate.

$${B}_n={O}_{n-1}-{D}_n$$
(15)

The desired delivery rate is dependent upon the current backlog and the expected delivery delay.

$${D}_n^{\hat{\mkern6mu} }=\frac{B_n}{\tau_n(L)}$$
(16)

The maximum delivery rate (Max(D)n) is dependent on the current inventory level (Sn) and the minimum time to fill orders (τn(I)).

$$\mathit{\operatorname{Max}}{(D)}_n=\frac{S_n}{\tau_n(I)}$$
(17)

The delivery ratio (DRn) is the number of outstanding orders that can be delivered from stock.

$${DR}_n=\mathit{\operatorname{Min}}\left(1,\frac{\max \left({D}_n\right)}{D_n^{\hat{\mkern6mu} }}\right)$$
(18)

The order delivery rate (Dn) is the actual rate at which orders are shipped out to meet incoming customer orders. The delivery rate does, under normal circumstances, equal the desired delivery rate, except when inventory levels are adequate, resulting in a reduced delivery ratio.

$${D}_n={D}_n^{\hat{\mkern6mu}}\ast {DR}_n$$
(19)

3.4 Incorporation of disruptions and response actions in the SD model

The flow of information, in the form of orders, is upstream, and the flow of material, in the form of deliveries, is downstream. In our study, the exogenous data are the demand scenarios that are informed by the data collected from the Finnish grocery retailers during the initial phases of COVID-19 crisis (Appendix A). The demand and orders for other echelons were calculated from the upstream flow of these orders.

  • Destocking is one of the common responses of firms that are struggling with supply or demand disruptions. In the wake of crisis, firms often lower their inventory targets to reduce inventory risk and increase financial liquidity. Additionally, when demand for certain products drops, the retailers usually engage in destocking. This phenomenon has been incorporated into the SD model, by reducing the on-hand inventory coverage levels by 50% at E1 and E2.

  • Phantom ordering has been modelled as inflated order when faced with a sharp demand increase that depletes echelon inventory. In case retailer echelon faces surge in demand, the outgoing orders are inflated by further by 30%.

$${O}_1=\mathit{\operatorname{Max}}\left(0,{PO}_1\ast \left({F}_1+{O}_1(SL)+{O}_1(S)+{O}_1(WIP)\right)\right)$$
(20)

Inflated ordering is limited to the period of demand exceeding the inventory levels of the retailers. With the decline of orders, the outgoings are readjusted by setting the multiplying factor to 1.

  • The logistics bottleneck is modelled as a logistic constraint (LC1) at E1 during the first week of demand increase.

$$\mathit{\operatorname{Max}}{(D)}_n=\mathit{\operatorname{Min}}\ \left(\frac{S_n}{\uptau_n(I)},{LC}_1\right)$$
(23)

This limit enables the firms in the simulation to increase their logistic capacity by 23% higher than the standard logistic capacity. Additionally, most retailers responded by increasing their logistic capacity through the reallocation of resources, collaboration with suppliers, and other means. This resilience enhancing response reaction is modelled by shortening the bottleneck constraint to one day.

  • The production capacity disruptions due to the lockdowns and social distancing are modelled in the E3 for the duration of the four weeks, when the available production capacity (c3) is set to zero.

  • The increased information sharing is modelled in E1 and E2. This is done through adjusting the time to average the forecast to one day. As a result, the echelons receive data in an accurate and timely manner. The simulation model only incorporates faster sharing of demand information, we did not model the observed planning coordination actions or informal sharing of the inventory level data, expert opinion, and other ways of temporary increases of SC visibility.

  • The flexibility to increase the capacity was modelled by increasing the production capacity in the E2. For these scenarios, the production capacity is increased by 30% if needed. The increase in flexibility runs parallel to the increased information sharing phenomenon. After the receipt of information of a sharp rise in demand, the increase in capacity is implemented, with the manufacturer a one-week capacity ramp-up period.

  • To avoid further complicating the simulation model by modeling complex structure for increased information sharing and visibility of capacity and inventory status across the echelons, we model this response action through its hypothesized effect. Visibility leads to cascading benefits in improved supply orders from the lower echelons; therefore, the lack of phantom ordering to upper echelons was used as a manifestation of the increased transparency.

3.5 Analysis of counterfactual via simulation experiments

We run the simulation experiments for two stylized SCs: Fast moving grocery SCs where products have a shorter shelf life and a slow SCs for products with a relatively long shelf life. The two kinds of SCs were selected as contrasting examples to inform of the observed disruptions and associated response actions in the SCs. A total of three market demand scenarios collected from the literature review are being analyzed: Demand Pulse, Demand Shock with Step Change, and Demand Drop with Recovery (Fig. 2). The period for analysis was six months for the first two scenarios, and the third scenario was analyzed for a period of twelve months.

Fig. 2
figure 2

Market demand scenarios

Full size image

The parameters for each SC are presented in Table 1. The parameterization is based on the SC models of (Fransoo and Udenio 2020; Udenio et al. 2015), the literature review and validation through background discussions with grocery SC practitioners.

Table 1 Operational parameters for each echelon (in days)
Full size table

As shown in the Table 1, the fast SC has shorter lead times, lower inventory coverage, and more frequent forecasting updates relative to the slow SC. Moreover, the following simplifications were also made.

  • The uppermost echelon has unlimited inventory, i.e., the supplier does not run into supply shortages.

  • The lead time at each echelon is determined as the time between order placement and order delivery.

  • The lead times are assumed to be constant, and the orders are placed immediately following a continuous-review policy.

  • The production time of one unit of product is set to 1.

3.5.1 Demand pulse scenarios

In the demand pulse scenarios, market demand remains stable for first ten weeks. It increases by 50% in the eleventh week and remains constant for the next five weeks. During week 16 and 17, there is a linear decrease in demand; from week 18 onwards, the demand is back at normal. The simulation is run for a period of six months. Furthermore, for the demand pulse scenario, six sub-scenarios were analyzed for both kinds of SCs. In the first two sub-scenarios, the disruptions were analyzed, while the last four ones dealt with response actions. The sub-scenarios are described in Table 2.

Table 2 Overview of all sub scenarios
Full size table

3.5.2 Demand shock with step change scenarios

In the demand shock scenarios, the market demand is stable for the first ten weeks, followed by an increase of 100% in the twelfth week. During week 13 and 14, the demand reduces linearly by 35% of the updated values and becomes stable for the remaining time. A total of five sub-scenarios were analyzed for both kinds of SCs. The sub-scenarios are described in Table 2.

3.5.3 Demand drop and recovery scenarios

In the demand drop scenarios, the market demand is stable during the first nine weeks, and during week ten, the demand reduces to zero. The demand is maintained at zero for two weeks and then ramps up towards initial level from week 16. From the week 20 onwards, the demand becomes stable at the original level. The simulation is run over a period of twelve months. A total of five sub-scenarios were analyzed for both kinds of SCs, as presented in Table 2.

3.5.4 Resilience measurement in simulations

We use the formula in Eq. 24 for measuring SC resilience in each scenario, as defined by (Zobel et al. 2021).

$$RI\left(\overline{X},T\right)=\frac{X^{\ast }{T}^{\ast }-\overline{X}T}{X^{\ast }{T}^{\ast }}\kern0.24em \overline{X}\in \left[0,{X}^{\ast}\right],T\in \left[0,{T}^{\ast}\right]$$
(24)

Where RI is the resilience index, \(\overline{X}\) is the average loss over the duration of the disruption, X* is the upper bound for the loss, T is the time until recovery, and T* is the maximum acceptable recovery time.

To make results over different simulation scenarios comparable, we set the X* and T* to the maximum of \(\overline{\textrm{X}}\ \textrm{and}\ \textrm{T}\) observed across all simulations for a given scenario. X* T* thus represents the worst-case scenarios and serves as a baseline for comparing simulations. For example, in a particular scenario, all simulations are considered, and then the X* and T* are defined as the maximum unfulfilled demand and the maximum number of days where demand remained unfulfilled. The loss of resilience has been quantified as the unfulfilled demand of the end customer during a time of disruption.

4 Results

4.1 Demand pulse scenarios

The results of the demand pulse scenarios are presented in Table 3, with sub scenarios for Fast and Slow SCs and RI for the resulting resilience indexes.

Table 3 Results of demand pulse scenarios. (Sub scenarios 1.x/2.x refer to Fast/Slow Supply Chain)
Full size table

As evident from Table 3, in the context of fast SCs, flexibility to increase capacity is the most effective resilience enhancing strategy, followed by increased information sharing. Resilience is enhanced at most when both these strategies are adopted. In slow SCs, increased information sharing is the most beneficial response action, followed by increased transparency and focus on increase in lead times of logistics bottlenecks.

Panels of Fig. 3 show the ability of Fast and Slow SCs to meet the end customer demand rate. The graphs show how the ability to fill the demand differs between the disruption scenario without response actions (E.g., Scenario 1.1. from Table 3) compared to the best-case scenario with all recommended response actions activated (E.g., 1.6 in Table 3). The fill rate figures show how the fast SC can increase capacity faster and decrease the backlog inflated phantom ordering can be combated through transparency and investment in flexible capacity enables meeting the increased need. For the slow SCs in the disruption the increased information sharing the transparency through collaborative planning across the echelons are very valuable for the increase in resilience.

Fig. 3
figure 3

Retail supply rate - demand pulse scenarios

Full size image

The backlog at the retail echelon also validates the importance of increased capacity and information sharing as resilience enhancing strategies. Figure 4 shows, how a fast SC is not able to recover all the backlog that has been accumulated due to inflated ordering. Slow SC performance is impacted less due to high inventory buffers and slow reaction to the changing forecast. The magnitude and duration of the backlog are reduced with surge capacity and information sharing.

Fig. 4
figure 4

Backlog in demand pulse scenarios

Full size image

In addition, the RI mainly considers the ability to meet consumer demand at the retail level and does not account for the bullwhip effect. In the case of phantom ordering, the demand is amplified severely at the upper echelons and the bullwhip effect is likely to affect the upstream echelons. Transparent availability and allocation of scarce capacity is a countermeasure for phantom ordering, which inhibits phantom ordering, and minimizes the bullwhip effect (Fig. 5).

Fig. 5
figure 5

Phantom ordering in demand pulse scenarios

Full size image

4.2 Demand shock with step change scenarios

The results of the demand shock with a later level change of demand scenarios are presented in Table 4.

Table 4 Results of demand shock with step change scenarios
Full size table

In the case of demand shock and then demand levelling out to a higher level, the flexibility to increase capacity has an obvious positive impact in the case of both fast and slow SCs. The demand shock increase cannot be covered with the limited capacity increase that we have simulated as an input scenario, this demand must be satisfied from pipeline inventories and partially with capacity increase. Due to the observed effects in Fig. 6, the SCs in practice moved to increase throughput and limit product portfolios to trade variety to volume. Increased information sharing is beneficial, and comparatively has bigger impact for the slow SCs. The phenomenon is also indicated by the ability of SCs to meet end customer demand, as described in Fig. 6.

Fig. 6
figure 6

Demand shock and step change scenarios

Full size image

Phantom ordering leads to extraordinary demand swings, especially in upper echelons. However, transparency is an effective countermeasure, which limits the bullwhip effect (Fig. 7).

Fig. 7
figure 7

Phantom ordering in demand shock and step change scenarios

Full size image

4.3 Demand drop and recovery scenarios

The results of the demand drop, and recovery scenarios are presented in Table 5.

Table 5 Results of demand drop and recovery scenarios
Full size table

In the case of demand drop scenarios, destocking has the most adverse impact on both fast and slow SCs, and intuitively the effect is more adverse for the slower chains. Moreover, information sharing, and transparency are the most beneficial response actions. The flexibility to increase production is of no consequence for the demand drop scenarios.

Figures 8 and 9 depict the ability of the SCs to meet customer demand during and after the demand drop instance. In both cases, destocking compromises the ability to meet end customer demand after the recovery. However, if destocking is avoided through countermeasures of transparency and increased information sharing, SCs do not struggle to meet customer demand in the aftermath of demand drop and recovery. The backlog rate for these scenarios further validates this.

Fig. 8
figure 8

Retail supply rate - demand drop and recovery

Full size image
Fig. 9
figure 9

Backlog in demand drop and recovery scenario

Full size image

In summary, both the type of SCs and the demand change scenarios affect the efficacy of resilience enhancing strategies. The lead times are of major consequence, as they shift the balance between the flexibility to increase capacity and increased information sharing measures as favorable resilience strategies. With the increase in lead times, the flexibility to increase capacity is replaced by increased information sharing as a relatively prioritized response choice. The impact of phantom ordering is mostly consistent across SC types, and it should be countered through transparent availability and allocation of supply. In addition, destocking in demand drop scenarios has adverse effects on resilience and is more severe in the case of slow supply chains. Slow SCs have longer lead times and forecast periods that lead to destocking having long-lasting effects in case the demand recovers.

5 Discussions and practical implications

The results of the simulation analysis can be utilized in the context of disruptions in SCs to enhance their resilience. The pandemic disruption threatened stability and performance of the grocery SCs through disruptions in supply capacity, logistic network, drastic demand changes, hoarding, shortages, and delayed bullwhip effect. We tested with simulation experiments how the response actions linked to information sharing, transparency, planning coordination, capacity flexibility, and careful management of inventory levels would impact the resilience, measured as ability to fulfill end-user demand. It is noteworthy that the responses by the disrupted firms (e.g., Saarinen et al. 2020), were in line with the seminal research literature on resilience and mitigation of bullwhip effect (e.g., Ambulkar et al. 2015; Lee et al. 1997); research has bridged the gap to the practitioners in the OM field. We contribute to previous simulation and modeling research, such as Burgos and Ivanov (2021) on the disruption management, with the staggered simulation study of the empirically grounded response strategies.

Our resilience framework is grounded in the Contingent Resource-Based View and the Dynamic Capabilities theories. While these theories emphasize the creation of a non-imitable, non-substitutable, and valuable set of resources that align with the turbulent nature of the environment, we introduce an additional perspective focusing on the internal operational dynamics of SCs. As commonly understood, RBV and DC theories would equally recommend incorporating flexibility and visibility into the resilience framework, but we elaborate these theories to particularly focusing on the contingent RBV and suggests priorities in inclusion of these capabilities to the SC resilience framework depending on the type of the SC on fast to slow (responsive to efficient) categorization. We extend the contingent SC resilience theory, by identifying that SC speed in addition to the scale driven complexity (Brandon-Jones et al. 2014) impact the effect of visibility on SC resilience. As our modeling demonstrates, factors like perishability, lead times, inventory buffer coverage, replenishment cycles and forecast-adjustment time, can shift the balance between flexibility and visibility as the most effective capabilities. This highlights the importance of tailoring resilience strategies to the specific operational characteristics of supply chains, recognizing that a one-size-fits-all approach may not be suitable in all circumstances (Table 6).

Table 6 Synthesis of the evaluation of response strategies in context of disruption scenario and supply chain type
Full size table

The findings revealed that in the context of fast SCs, effectively addressing the sudden and substantial surge in demand, requires flexibility. This is due to the high perishability of the products, which require daily replenishment cycles and have low pipeline inventories. Therefore, to increase resilience the fast SCs should primarily focus on capacity flexibility and responsiveness. After flexibility, the next priority is the investment in visibility capability, the adoption of timely and effective information sharing mechanisms. The importance and utility of flexibility as a resilience capability has been stressed in several empirical studies, for example, in studies of COVID-19's impact on Iranian and Chinese food markets (Heidary 2022; Wang et al. 2020). Our study aligns with this perspective, highlighting that the effectiveness of flexibility capability is enhanced by the inclusion of visibility capability in the SC resilience framework. We posit, along with other similar studies, that the impact of COVID-19 on SCs was largely contained, as the grocery SCs were able to leverage both flexibility and visibility capabilities (Lozano-Diez et al. 2020; Veselovská 2020). Providing a counterpoint to this success, disastrous bullwhip effects have been observed in other industries characterized by longer delays and lower flexibility, such as the garment industry (e.g., Meyersohn 2022; News 2022) or the semiconductor industry (e.g., The Economist 2022).

SCs can invest in flexibility capability by adopting strategies such as eliminating non-essential activities and redirecting personnel resources for production can be implemented (Leite et al. 2020). Additionally, product portfolios can be temporarily rationalized, package sizes can be adjusted according to bottleneck operation, and the standard size of products can be reduced (Butt 2021). Flexibility can focus on firm’s internal operations and resources and extend to production system design but SC flexibility also encompasses ability to adapt and reconfigure supply networks for recovery of capacity or building new production and distribution capacity (Müller et al. 2023).

The analysis of slow SCs, where the perishability of products is low, inventory buffers higher, and lead times are relatively longer, indicates the visibility capability should be prioritized. Disruptions in slow SCs are marked by short-term and fluctuating variations in supply and demand parameters. Where the value of visibility capability and resultant information sharing pertaining to operational status and predictions regarding future demand becomes particularly crucial. Importance of information sharing is heightened when the lead times surpass the durations of disruptions. Thus, the timely and accurate availability of operational insights and predictive analytics become integral for effective decision-making.

Identification of the global bottleneck of the SC during the disruption, and the ability to engage in coordinated planning to maximize the output from the bottleneck is an important planning and coordination capability (Saarinen et al. 2020). Furthermore, data sharing, and data analytics play a crucial role in ensuring optimal production capacity to meet end customer demand and maintain profitability (Mehrotra et al. 2020). Similarly, SCs can plan and integrate visibility capability in their resilience framework by increasing the access and flow of information through early warning communications, real time monitoring, resource location and status tracking, performance monitoring, vertical and horizontal collaboration, joint decision making, and predictive analytics (Ali et al. 2017; Hohenstein et al. 2015; Johnson et al. 2013). However, an effective resilience framework for slow SCs should also include complementary flexibility mechanisms such as inventory buffers, and data analytics for efficient planning and information sharing, and securing logistics capacity (Paul and Chowdhury 2020).

In addition to inclusion of flexibility and visibility in resilience framework, we highlight importance of avoiding adverse response actions. Phantom ordering and inventory destocking result in negative impact on SC resilience. Both of these response actions will lead to bullwhip propagation (Lee et al. 1997). The destocking initially works well for the focal firm, however as the disruptions pass and demand recovers, the lack of inventory could result in lost sales. Thus, destocking response is counterproductive to performance and profitability if the demand-loss is temporary (Fransoo and Udenio 2020; Udenio et al. 2015). Therefore, it should not be a go-to measure in demand drop cases, and the situation should be analyzed carefully. SCs should increase transparency and information sharing as unbiased demand information will yield best results with careful management of production volumes and inventories.

Our analysis also demonstrated the impact of phantom ordering on the bullwhip effect observed from retailer to manufacturing echelons. In scenarios of demand fluctuations, phantom ordering is one of the key reasons for the bullwhip effect, which is initially adopted as a countermeasure at the retailer echelons (Geary et al. 2003; Lee et al. 1997). Phantom ordering results in longer delivery times, which further increases the scarcity of resources, pressure to inflate orders further, and hence the lowering of resilience (Sterman and Dogan 2015). To prevent this, increased transparency across the SCs regarding the availability and allocation of resources and utilizing different transparent ways to share the scarcity, for example, according to historical volumes or previous orders, should be implemented.

In summary, to increase the resilience of SCs and better prepare for future disruptions, SCs should design their resilience framework, in consideration of the operational dynamics of their products and SCs. Responsive SCs should focus on flexibility capability, that will enable effective adaptation. Efficient SCs should prioritize visibility capability that enables aligning correctly the structurally slower response through the chain. In general, bolstering and developing information sharing and coordination mechanisms with the ability to adapt operational planning to exceptional circumstances across the firm boundaries, i.e., visibility capability, is valuable for resilience. Agile planning and coordination across the chain will help fast sensing the direction and extent of disruptions and help adopting correct responses. The information should be shared across the supply chains so that all stakeholders can make informed judgments.

6 Conclusion

SC resilience has become a key concern for SC practitioners, owing to the increasing uncertainty and turbulence in the operating environments. SC resilience is enhanced through the adoption of capabilities that enable dynamic and effective response strategies when disruptions happen. We contribute to the SC resilience literature by highlighting the impact of internal contingencies of SCs on the effectiveness of resilience capabilities.

We particularly focused on the case of Finnish grocery SCs during the COVID-19 pandemic and investigated the effectiveness of the resilience capabilities in accordance with the speed of the SC and nature of the products in terms of perishability. We leveraged the contingent RBV and DC theories to study the impact of operational dynamics on effectiveness of resilience related resources. The results indicate that flexibility to increase capacity is the highest impact resilience-enhancing capability for fast SCs, while visibility is the most effective capability for slow SCs. The study also found that destocking should not be the sole response to demand drops, and agile information processing and planning should be used to adapt operations and plans to the changing demands of the disruptions. In addition, the study highlights the importance of preventing phantom ordering through transparency across the supply chains.

Our study has some limitations. For example, it does not consider the impact of increased online shopping on demand, and the element of profitability is not incorporated in the analysis. We did not consider costs for the strategies in their evaluation, which will be highly context dependent. Future research could also compare these results to counterfactual analysis and validate the results with empirical data, as our empirical grounding is fully on qualitative data. Similarly, future studies could also analyze the effectiveness of resilience strategies while considering additional resilience metrics in addition to the time taken to recover operations. Further studies could also add additional contemporary SC aims like sustainability, efficiency, etc. to explore the synergy between resilience capabilities while focusing on the internal contingencies. Despite the current limitations, the findings of this study provide valuable insights for designing resilient supply chains in the wake of disruptions.