Food retail supply chain resilience and the COVID-19 pandemic: A digital twin-based impact analysis and improvement directions (2024)

2.1. Case study and data

We used anyLogistix SC simulation and optimization software to study a multi-stage SC for a retail company in Germany. This software has been frequently and successfully used for SC resilience analysis (Ivanov, 2017, Ivanov, 2018, Aldrighetti et al., 2020, Dolgui et al., 2020a, Ivanov, 2021c). Without loss of generality, we restricted ourselves to the consideration of ten product categories and 28 supermarket locations in five different countries (Germany, Austria, the Czech Republic, Italy, and Hungary). We selected product categories that experienced significant changes due to new consumer trends during the COVID-19 pandemic – Fresh Fruits, Fresh Vegetables, Fresh Meat, Fish & Sea Food, Rice, Pasta, Convenience Food, Frozen Meals (ready-to-eat), Wheat Flour, and Confectionary Food. Product demand was calculated by multiplying the annual per capita consumption for each product category by the retail company’s market share. The selling prices for each product were taken from Statista data regarding the average price for each product category in 2020.

Next, we created a sample of three suppliers per product category (30 suppliers in total) by analysing supermarket data and manually identifying supplier locations. Three distribution centre (DC) locations were selected, one in east, one in west, and one in south Germany. Fig. 1shows the SC design.

Open in a separate window

Fig. 1

Supply chain design (interface from anyLogistix Studio Edition).

The suppliers produce and ship their products via trucks to the DCs. We assumed that the costs of shipping the products are already included in the final price agreed with the retailer. Inventory spending corresponds to the expenses for replenishing the inventory. This is the sum of the initial inventory purchase costs and the replenishment costs.This cost is calculated by anyLogistix automatically for suppliers by multiplying initial stock units by the costs for the corresponding product.

2.2. Digital twin and simulation model

Digital SC twins are defined as ‘computerized models that represent the network state for any given moment in time’ (Ivanov and Dolgui, 2021c). The main difference between digital twins and simulation models is comprised of three aspects, i.e., system complexity, real-time connectivity, and decision-making integration. First, the system complexity in digital twins is higher as in simulation models. The digital SC twins are comprised of multiple layers such as network structure, flows, process control algorithms, and operational parameters. Second, data in the digital twins is updated through connectivity with external systems and databases. Third, integration of digital twins into decision-making is much higher as compared to models. Simulation and optimization models are important parts of digital twins whereby digital twins can offer more functions for decision-making support as classical offline modelling, e.g., performance analysis of suppliers, updating SC data in ERP systems, and comparison of the existing processes with the optimal ones. Digital twins allow for both computations of some numerical outcomes and an extended analysis of bottlenecks, comparison of different options, and analysis of resilience investments. A digital SC twin is a part of SC management toolbox, control towers, and decision-making support at strategic, tactical, and operational levels. Besides, data-driven modelling in digital twins allows using artificial intelligence in order to support decision-making by so called dispatch advisors or digital companions. For example, a decision-maker can ask the digital twin to find all current shortages in the SC and suggest possible ways to resolve these problems. In the next step, the shortage data can be used for optimization and simulation to develop most appropriate action plans.

Digital twins can be used for building end-to-end SC visibility, enhancing resilience, and assessing contingency plans. Cavalcante et al. (2019) point out that digital SC twins display the ‘physical SC based on actual transportation, inventory, demand, and capacity data’. Therefore, they can be utilized by decision-makers for planning, monitoring, and supervising. Hence, digital SC twins can improve SC visibility and, as a result, SC resilience. Ivanov and Dolgui (2020) and Hosseini et al. (2020) state that there is an urgent need to visualize SC networks because of the increasing number of SC disruptions.

Digital SC twins enable real-time transparency about important logistics data such as financial key performance indicators (KPIs), inventory level, stock level, service level, capacity, and transportation data. They are powerful data-driven tools and firms’ control towers. Performance-based simulation models help create efficient contingency plans to prevent or recover from disruptions by simulating and creating what-if scenarios that predict the future impact. Digital twins not only visualize SCs and associated risks but also offer supplier performance and risk analysis along with forecasts of SC interruptions and risks. In addition, they can allow to establish and examine detailed backup routes, including estimated time of arrival (ETA) calculations. During disruptions, digital twins utilize real-time data to calculate the impact of the disruption, build alternative SC networks and perform KPI analysis to get real-time data about inventory levels, service level, financial parameters and demand (Ivanov, 2020a). With the help of anyLogistix SC simulation and optimization software, a digital SC can be designed (Fig. 2).

Open in a separate window

Fig. 2

Digital supply chain design for disruption analysis using anyLogistix (based on Ivanov and Dolgui, 2021b).

Fig. 2 shows the structure of a digital SC twin created in this study for disruption analysis. The digital SC twin encompasses three major perspectives– the network, the flows, and the parameters. The SC network can be designed using different location objects, such as customers, DCs, factories, and suppliers. The flows in the network can be flexibly arranged to represent the specifics of different SCs. The flows are associated with some design (i.e., maximum) capacities in production, warehouses, and transportation and controlled by the associated production, inventory, sourcing, and shipment policies. These policies can be flexibly adapted to the specifics of the SC and its management rules. Finally, different operational parameters such as demand, lead time, and control policies’ thresholds (e.g., re-order point, target inventory, and minimum vehicle load) can be defined. With that functionality, a digital model of a physical SC (i.e., a digital SC) can be created and used for optimization and simulations to analyse SC operations and performance dynamics under disruptions.

The simulation model structure developed in the anyLogistix digital twin is shown in Fig. 3.

Open in a separate window

Fig. 3

Simulation model design.

See Also

Effect of COVID-19 on food security, hunger, and food crisisGroceries and Covid-19 | safefoodFood service industry in the era of COVID-19: trends and research implications

The simulation model is based on production-inventory control with five main components in the control loop (Fig. 3): demand, lead time, continuous inventory control with a re-order point and a target stock setting, production control, and transportation control. It is assumed in this simulation that the DCs follow (s,S) inventory control policy. For this simulation, the re-order point(s) per product per DC equals the daily demand per product at each DC since the lead-time is less than 1day. Similarly, the target inventory (S) corresponds to twice the value of the re-order point. The computation of target inventory (also known as the baseline stock) is a complex task subject to specific requirements (e.g., a dependence on some service level thresholds) and analysis objectives (Chen and Disney, 2007, Minner and Transchel, 2010, Svoboda et al., 2021)). In this study, we do not pretend to analytically determine optimal re-order points and target inventory for our system. The role of inventory control in our model is to enable dynamic simulation of the SCs. With that in mind, we rely on studies (Ivanov, 2017, Aldrighetti et al., 2020) that used as a target inventory the twice value of the re-order point. Inventory carrying costs for DCs is $0.22. It is assumed that suppliers have limited inventory. Therefore, an (s,S) inventory policy with the same replenishment logic for the DCs was applied.

The Expected Lead Time (ELT) for each order is 1day. In other words, if the order is delivered within one day, it is considered on-time delivery. Otherwise, the order is counted as a delayed delivery. The delayed deliveries have a negative impact on the ELT service level (i.e., on-time delivery), which is defined as a ratio of on-time delivered orders to the total number of orders.The facilities have some processing costs, which can be classified as inbound and outbound costs. These costs correspond to the expenses incurred in receiving shipments from a supplying site and from sending shipments to a receiving site, respectively.

The pandemic is modelled by setting some disruption and recovery events for supply and capacities along with surges in demand. The results of the simulations are evaluated through the following KPIs (following studies by Ivanov, 2017, Ivanov, 2018, Singh et al., 2021, Dolgui et al., 2020a):

• ▪

  Financial indicators:Statistics related to this group provide detailed information on generated revenue and incurred expenses during the simulation experiment for the specified scenario. They include profit, revenue, and costs (inventory carrying costs, transportation costs, inventory spending, inbound processing costs, outbound processing costs, and other costs).

• ▪

  Expected lead time:It displays statistics on the delivery time of every ordered product item. It is updated each time a shipment containing the order is delivered (all delivered orders are considered whether they are on time or delayed).

• ▪

  Average daily available inventory:It shows statistics on the daily integral mean of the available volume of products in stock.

• ▪

  Demand (product backlog):It illustrates the quantity of processed products for incomplete orders (orders that currently lack the required number of products).

• ▪

  Fulfillment (late orders):It shows statistics on the number of orders that failed to arrive within the specified ELT (e.g., the orders that are still being delivered after the ELT has expired).

• ▪

  ELT service level by orders: It shows the service level based on the ratio of on-time orders to the overall number of outgoing orders.

We validated the model in several ways. First, we tested the model on the ideal (i.e., business-as-usual) scenario (see Fig. 5). Second, we compared the results of our simulations with statistical data for German retail in 2020, which confirmed the trends identified in our experiments (e.g., growing demand and increasing transportation costs). Third, we visually checked the dynamics of material flows through the simulations jointly with two retail SC managers, who confirmed the correlation between our simulation and performance dynamics and real-life settings. Fourth, we performed a set of variation experiments with different parameters (e.g., re-order point and demand). They confirmed the model’s sensitivity. The sensitivity analysis with individual parameters did not reveal any interesting or novel managerial implications, so we have omitted the presentation of these test computations and focussed on the managerial implications of our experimental results. However, another set of sensitivity analyses with consideration of time-to-recovery provided interesting insights which are discussed in Section 5. Finally, we used the output data analysis in the log files and replication tests for the validity proof. We selected the timing of disruptions to avoid the ‘noise’ at the start of the simulation experiment.

Open in a separate window

Fig. 5

Experiment description: Bottlenecks in transport.

3.1. Scenario 0. Disruption-free (baseline) scenario

This simulation analyses the performance of the food retail SC in a scenario free of COVID-19 disruptions. We consider it as a baseline scenario. The results serve as a basis for comparing and analysing the food retail SC performance during COVID-19 pandemic-induced disruption events.

3.2. Scenario 1. Increase in demand

Four fixed periods that are distinct in demand are simulated (Table 1).

The demand disruptions belong to the most important vulnerabilities in SCs (Shen and Li, 2017). The logic of this experiment is as follows. Base demand (i.e., 100%) corresponds to a business-as-usual scenario. This period corresponds to the beginning of 2020, when COVID-19 was not yet widespread in Germany. The base demand increase by 75% corresponds to the first lockdown period. It starts when the first restrictions on public life were imposed and ends when social distancing restrictions were relaxed and a slow return to public life began (following the timeline described above). During this period, panic buying increased the demand enormously in different product categories, which is reflected in our scenarios (Paul and Chowdhury, 2021). Next, a period of 10% increase in the base demand is introduced. This period corresponds to the time when coronavirus control measures were relaxed. Finally, the third period of base demand increase (of 35%) corresponds to Germany’s second lockdown period, which impacted the demand to a lesser extent than the first lockdown. The objective of this scenario is to simulate the impact of increasing demand over the whole year. The results are analysed using the KPIs explained in Section 4.1.

3.3. Scenario 2. Shutdown at suppliers’ factories

The pandemic has caused temporary shutdowns of factories, which in turn has resulted in a sharp decline in production. Limits on people’s mobility have reduced seasonal workers’ availability for planting and harvesting in many countries. In addition, food processing industries have been affected by social distancing rules and other measures aiming to contain the virus’s spread, which have reduced operations’ efficiency (OECD, 2020). Thus, we simulate a shutdown at three factories (i.e., suppliers) to analyse the impact of their closure on food retailers. Table 2shows the experimental setup for the event simulation and the period when the shutdown occurs.

The shutdown period for each site is consistent with the German government’s period of restrictions to contain the coronavirus spread (explained in the timeline above). The objective of this scenario is to simulate the impact of closing the factories of two major suppliers during a fixed period of 10days in March and one supplier’s factory during a 15-day fixed period in April.

3.4. Scenario 3. Bottlenecks in transport

Bottlenecks due to increased border controls are simulated in this experiment according to real-life scenarios encountered in spring 2020 (BMI, 2020). Although no border closures for the transportation of goods were imposed, traffic jams at the border and long queues occurred because of the temporary border controls during the first lockdown in Germany. This limited the normal flow of transports into the country. Bottlenecks in transport due to border control delays are simulated for six different periods. This scenario simulates an interruption of material flows over a period of 2days in each experiment. Fig. 5shows the interrupted paths with their respective durations.

The disruption period for each route is consistent with the period during which the German government’s restrictions to stop the coronavirus spread were in effect, as explained in the timeline above. The purpose of this scenario is to simulate the impact of closing paths from suppliers to DCs and from DCs to customers during short periods of 2days in March and April.

3.5. Scenario 4. COVID-19: Increase in demand, shutdown of suppliers’ factories, and transport bottlenecks

This scenario simulates the impact of COVID-19 on the food retail industry SC by combining scenarios 1, 2, and 3. This simulation aims to reflect the overall food SC performance during two pandemic waves in 2020.

4.1. Impact of COVID-19 in Germany’s food retail SC performance

This section aims to respond to RQ1: How and why has the COVID-19 pandemic outbreak impacted food retail SC performance? This section presents the simulation results and analysis of the simulation experiments. First, the performance of a disruption-free scenario will be described. Subsequently, the disruption scenarios will be evaluated using four simulation experiments. Additionally, a cross-comparison analysis of the scenarios and generalizations of the COVID-19 impact in the food SC will be provided.

4.1.1. Scenario 0. Disruption-free (baseline) scenario

This section describes the SC performance of the food retail industry in Germany in a pandemic-free (i.e., baseline) scenario. Fig. 6shows the simulation results.

Open in a separate window

Fig. 6

Experiment results: Disruption-free (baseline) scenario.

Regarding Lead Time, Fig. 6 shows a histogram of the daily lead time, representing the delivery time for each product item ordered. The x-axis shows the lead-time in days, and the y-axis indicates the number of occurrences of orders with a particular lead-time. Products are delivered within a time interval of 0 and 12h, but most of the products are delivered within a 2-hour timeframe. Thus, the customer’s ELT of 1day is met for all product categories. This case suggests that lead times in the food retail industry are usually short because of recurring orders and high inventory rotation in DCs.

With respect to the Average Daily Available Inventory, the simulation outcome shows the inventory levels in DCs for all product categories. The order-up-to-level policy (s,S) allows ordering quantities up to level S when reaching the re-order point s. This policy prevents excessive inventory levels and shortages and considers some situational demand fluctuations. From the results, it is evident that there is enough stock availability from day 1 in all scenarios, which allows for high flexibility to match demand. Small fluctuations with low amplitude can also be seen for all products. Moreover, inventory levels do not fall to the 0 level during the year, which means that the inventory policy allows the retailer to satisfy demand across the SC.

Demand satisfaction can be measured in terms of Product Backlog, which indicates the number of processed products for orders that lack the required number of products. In Fig. 6, a backlog of 0 can be observed, meaning that all orders are completely delivered.

Fulfilment (Late Orders) shows the number of orders that fail to arrive within the specified ELT. As the ELT is one day, every order delivered within a frame greater than 1 is considered a late delivery. In Fig. 6, it can be seen that all products are delivered on time because the number of late orders is 0. Therefore, the number of on-time orders equals the total number of orders.

Finally, the ELT Service Level analyses the ratio of on-time orders to the overall number of outgoing orders. Hence, late deliveries have a negative impact on ELT service. Results show a service level of 1, which means that customers receive 100% of their orders without delays and within the expected lead time.

4.1.2. Scenario 1. Increase in demand

Scenario 1 shows the impact on the food SC performance after an increase in demand resulting from the COVID-19 pandemic outbreak. Fig. 7illustrates the results for each KPI.

Open in a separate window

Fig. 7

Experiment results: Increase in demand.

This scenario reveals that an increase in demand positively influences the revenues because a higher volume of products is sold, which increases the profits. Additionally, inventory levels decrease because of demand growth. Consequently, lead time increases, as well as the number of late orders, reducing the ELT service level. An increase in the product backlog can also be seen because of a rise in the number of incomplete orders.

Lead Time for most orders in scenario 1 is between 0 and 10days. Under normal conditions, lead time is a maximum of half a day. Also, with a lower frequency, some orders can take up to 50days to be delivered to the customer. It can be inferred from these results that as the demand increases, procuring the quantities required to meet it takes longer because the available inventory is insufficient to satisfy the growing demand. Delays in delivery occur until the stock is replenished. The results also show that theAverage Daily Available Inventory dramatically decreases because of the increased demand. Fluctuations throughout the year can be seen in the ‘Average Daily Available Inventory’ graph in Fig. 7. When demand increases by 75% (corresponding to Germany’s first lockdown), there is a sharp drop in inventory. Inventory levels do not recover to a level similar to that in the initial scenario throughout the year. On the other hand, during the second lockdown, a slight decrease in inventory can be seen for some product categories, but this decrease is not as abrupt as the decrease during the first lockdown. An increase in the demand for all products simultaneously over a relatively long period causes delays and unattended orders.

In addition, concerning the Product Backlog, the results show that orders are not completely delivered, and thus demand is not 100% satisfied. The ‘Demand (Product Backlog)’ graph in Fig. 6 shows a sharp increase in the product backlog during the first lockdown period, when demand for all products increases by 75% compared to their initial level (in the disruption-free scenario). Then, beginning on day 150 of the year, there is a decrease in the backlog as demand decreases compared to the first lockdown period but remains 10% higher than in a disruption-free scenario. An increase in the backlog appears again on day 300 – which corresponds to Germany’s second lockdown period – representing an increase in demand of 35% compared to its initial level.According to the simulation results, the accumulated backlog throughout the year is 32.1 million kg, while the total demand for all products is 1.157 trillion kg. Therefore, 2.7% of the total demand is not satisfied after an increase in demand due to the COVID-19 pandemic.

The results also indicate that when there is an increase in lead time, the Fulfillment (Late Orders) indicator is affected. In the ‘Fulfillment (Late Orders)’ graph in Fig. 7, a growing line over time indicates the number of orders that are not delivered on time for all products in all customer locations. According to the results, 3,949 orders are delivered late throughout the year, representing 3.7% of the total number of orders placed by the customers.

Similarly, Fig. 7 shows that the ELT Service Level for all products is 100% from the start of the year until the beginning of the first lockdown period in mid-March. Then, the average service level falls to 87.2% and remains at this level throughout the rest of the year. The increasing number of late orders is the reason for these ELT service level dynamics.

4.1.3. Scenario 2. Shutdown in suppliers’ factories

Scenario 2 shows the impact on the food SC performance after a shutdown in suppliers’ factories due to the COVID-19 pandemic outbreak. We simulated a shutdown for three different product categories over periods of 10 and 15days. This yielded the results depicted in Fig. 8. According to these results, a temporary production shutdown at three suppliers’ factories slightly reduces profit as total costs increase. Moreover, a reduction in inventory levels at DCs occurs as suppliers stop delivering the products during the disruption period. Additionally, there is a high increase in inventory when the factories are reopened. This effect can be explained by disruption tails – that is, a destabilization of production-ordering dynamics in the post-disruption period due to a lack of adaptability in the production-inventory control policies when transitioning from the disruption to the recovery period (Ivanov, 2019, Ivanov and Rozhkov, 2020). Consequently, lead time (as well as the number of late orders) increases, which decreases the ELT service level over an extended period. Furthermore, the number of incomplete orders also grows, generating product backlog in some product categories.

Open in a separate window

Fig. 8

Experiment results: Suppliers’ factories shutdown.

After an outage of products, incoming quantities accumulate in DCs. Thus, stock levels grow, causing an increase in inventory carrying costs. Also, during a supply outage, incomplete orders are transported from DCs to customers, decreasing the vehicle capacity utilization. Then, when products become available after the suppliers’ factories re-open, the number of trips escalates to deliver the missing quantities. Therefore, transportation costs increase. Inbound and outbound costs remain the same. Other administrative costs do not vary, as they are fixed costs. The ELT service level falls to an average of 88.2% as the minimum value and improves throughout the year, reaching a level close to 100% at the end of the year.

4.1.4. Scenario 3. Bottlenecks in transport

Scenario 3 shows the impact on the food SC performance of multiple transport bottlenecks due to the COVID-19 pandemic outbreak. Six bottleneck events are simulated for 2-day periods in different paths from suppliers to DCs and from DCs to customers. The results are depicted in Fig. 9.

Open in a separate window

Fig. 9

Experiment results: Bottlenecks in transport.

According to the results of scenario 3, transport bottlenecks have a small impact on profit performance. There is a minor reduction in inventory levels at DCs during the disruption as suppliers stop delivering the products within short time periods. Therefore, the number of late orders increases, although in minimal quantities compared to the total number of orders. Moreover, the number of incomplete orders rises but only slightly, resulting in a very low backlog. As a result, the service level remains close to 100% within the entire period, indicating demand satisfaction.

In terms of Fulfillment (Late Orders), a growing line that stabilizes over time can be observed. 30 orders are delivered late throughout the year, representing approximately 0.03% of the total number of orders placed (102,280). Finally, the ELT service level is approximately 100% during the whole period. This number can be explained by the small number of late deliveries as most orders were delivered on time. Thus, the demand during the year is satisfied overall despite the transport disruptions.

With regard to scenario 3, we can conclude that the magnitude of the disruption impact is significantly lower as compared to scenarios 1 and 2. The short-term bottleneck in transport lasts for two periods, with a lead time of one period and can be considered rather an operational risk event and not as a disruption. Fig. 8 confirms that there is no significant SC performance impact.

4.1.5. Scenario 4. Full impact of COVID-19 pandemic through two waves in 2020 – A combination of all scenarios

Scenario 4 shows the impact on the food SC performance after experiencing a combination of the three disruptive scenarios explained above: an increase in demand, a shutdown in suppliers’ factories, and transport bottlenecks. Fig. 10shows the results of the experiment through the selected KPIs.

Open in a separate window

Fig. 10

Experiment results: COVID-19 scenario.

This simulation illustrates the SC performance in the food retail industry during the COVID-19 pandemic outbreak in Germany. In this scenario, synergetic effects of adding different negative events can be observed. Interestingly, the aggregation of events results in a positive impact on SC financial performance. Increased revenues and decreased total costs have a positive impact on profit. Nevertheless, delays occur, and a considerable percentage of products are not delivered on time or are incompletely delivered (reflected in the fulfilment (late orders) and demand (products backlog) graphs, respectively). These outcomes lead to out-of-stock products, especially during the first lockdown period. Moreover, a decrease in inventory followed by a considerable increase after the government relaxes the lockdown measures is apparent in the average daily available inventory graph.

The second wave of COVID-19 impacts the industry to a lower extent than the first wave as inventory levels do not drop significantly and a smaller backlog is accumulated. Furthermore, the late orders curve is flattened, which implies a smaller number of delayed orders. The accumulated disruptions result in a drop in the service level to 80%, demonstrating that a significant percentage of the demand is still satisfied despite the pandemic outbreak. Therefore, in general terms, the simulation suggests that the food retail industry can benefit from the pandemic in terms of financial performance and growth opportunities.

Transportation costs increase in the COVID-19 scenario because a higher quantity is shipped in response to higher demand during the pandemic. However, there are inefficiencies in transport as incomplete orders are shipped from the warehouse to customers when suppliers’ factories are shut down, leading to lower vehicle capacity utilization. In addition, transportation costs increase because there is an increase in the average number of vehicles used to transport the requested amounts. Inbound and outbound costs also increase because increasing demand leads to more goods needing to be processed at warehouses. Other costs do not vary as they are fixed administrative costs.

The Lead Time graph shows that a considerable number of orders are delivered to the customer within a range of 0 to 10days, but under normal conditions, this takes a maximum of half a day. Less frequently, orders may take up to 70days to be delivered. Thus, the disruptions resulting from the coronavirus outbreak affect the lead time, so providing customers with the required quantities of goods within the expected time becomes a challenge for food retailers.

Moreover, Average Daily Available Inventory levels are affected by the COVID-19 disruption. Initially, a sharp drop in inventory occurs because of the increased demand along with factory shutdowns and bottlenecks in transport during Germany’s first lockdown. During the second lockdown, a slight decrease in inventory can be observed; however, it is not as abrupt as that in the first lockdown. Also, alterations in stock at the end of the period can be observed inducing the ripple effect (Ivanov et al., 2014, Dolgui et al., 2018, Li et al., 2021) of the first lockdown. For some product categories, inventory levels recover to a level similar to that in a disruption-free scenario throughout the remaining part of the year.

Regarding Product Backlog, the results show that some orders are not entirely delivered because of a lack of the required number of products, and thus demand is not 100% satisfied. The graph starts with a sharp increase in the product backlog, which results from the 75% increase in demand, disruptions in transport, and shutdowns in the suppliers’ factories during the first lockdown. Then, beginning on day 150, a decrease in the backlog can be observed. This result occurs as coronavirus regulations are softened, and the country begins a gradual return to normality.As a result, demand decreases compared to the first lockdown period, suppliers’ factories reopen, and transport operates under normal conditions. At the end of the year, a slight increase in product backlog can be explained by the increased demand during Germany’s second lockdown period.

The pandemic outbreak negatively impacted the Fulfillment (Late Orders) indicator, which shows a growing line over the whole period beginning on day 80, which coincides with the first lockdown. This line indicates the number of orders that fail to be delivered on time in all customer locations. According to the results, 9,067 orders are delivered late throughout the year, representing approximately 8% of the total number of orders (102,290). Finally, this scenario shows that the ELT service level is 100% from the start of the year until the first lockdown period in mid-March, when the service level falls to 80% and remains at this level throughout the rest of the year.

4.2. Cross-comparison analysis

After evaluating each simulation scenario separately, we used a cross-comparison analysis of all scenarios to evaluate the results and create generalizations. A summary of the computational results is presented in Table 3.

The overview in Table 3 allows for analysing the effects of the COVID-19 outbreak in the food SC for the selected KPIs. It serves as a guide for addressing the main SC issues in Germany’s food retail industry resulting from the pandemic.

Regarding financial performance in the scenarios, Table 3 displays detailed information on generated revenue and incurred expenses during the initial scenario simulation experiment. Revenue includes the income generated from selling products to customers. Total costs include inventory carrying costs, transportation costs, processing costs (inbound and outbound), inventory spending, and other costs. Profits are calculated by subtracting total costs from revenues. The baseline scenario shows profitability, leading to outstanding performance. Total costs add up to USD 2.750bn and represent 90% of the revenue (USD 3.057bn). Given that companies in the retail industry do not usually produce, but rather purchase, inventory, inventory spending (i.e., 84.6%) typically represents the highest costs in their SC. Inventory carrying costs represent 12.8% of the total costs. As many of the products are perishable, they rotate quickly – typically on a daily basis. Thus, DCs do not accumulate excessive stock, leading to relatively low inventory costs.

For scenario 1, the results suggest that an increase in demand in different periods throughout the year shows a positive impact on sales as profit increases because of increased revenues. The total costs add up to USD 2.723bn and represent around 71% of the revenue (USD 3.825bn). Transportation costs increase and represent 6.2% of the total costs (see Table 3). This increase results from a growth in the number of shipping vehicles as larger quantities are transported from the DCs to customers. Inbound and outbound costs also increase because more goods must be processed to meet demand.

In scenario 2, financial performance is not strongly affected throughout the year after experiencing a stoppage in supply for a certain period. Profit decreases by approximately 1% compared to a disruption-free scenario. Although revenues are not impacted, total costs are affected by increased inventory and transportation costs. Thus, total costs add up to USD 2.753bn and represent 90% of the revenue (USD 3.057bn). Inventory levels are readjusted to satisfy the demand because the min-max policy allows variable order quantities, thus preventing inventory shortages.

In scenario 3, the financial performance shows only a minimum change as compared to the baseline scenario. The COVID-19 scenario (no. 4) has a positive impact on revenues and profits. The increasing demand throughout the year is the principal reason for revenue generation. Total costs add up to USD 2.707bn and represent approximately 71% of the revenue (USD 3.816bn). Regarding inventory costs, the min-max policy allows adjusting the inventory levels to satisfy demand, preventing inventory shortages. Thus, inventory levels change to cope with the multiple disruptions faced. Nevertheless, available inventory is rapidly consumed, leading to lower stock levels at DCs. As a result, inventory carrying costs decrease by 17% in the COVID-19 scenario compared to the disruption-free scenario.

Analysis of the results from the different scenarios yields several important observations. As the overall demand increases (scenario 1) at different levels during the selected period, inventory levels decrease. Also, the number of total shipping vehicles rises as more orders need to be dispatched. At the same time, longer lead times can be observed, leading to an increase in the late-orders ratio, which in turn reduces the service level. Backlogs due to incomplete orders also occur, but they are low when compared to the total demand. From a financial point of view, an increase in demand positively impacts revenues, leading to growing profits.

A shutdown in three suppliers’ factories (scenario 2) for 10 or 15days generates increased inventory levels and a higher number of shipping vehicles used when inventory accumulates at DCs after the factories are reopened. This growth in the number of shipping vehicles used leads to decreased transport efficiency because some vehicles are not fully utilized during the shutdown. Longer lead times can be observed, leading to an increased late-orders ratio and a reduced service level. Backlogs due to incomplete orders are evident.

Multiple short-time bottlenecks in transport (scenario 3) have a small impact on the KPIs selected. The inventory levels and the number of shipping vehicles used slightly increase. The mean lead time remains low, within the expected lead time of 1day, and the number of late orders rises, although in a small quantity compared to the total demand. This leads to a high service level, close to 100%. Backlogs due to incomplete orders occur in small quantities such that the approximate value of backlogs is 0. Profits slightly decrease because of increased total costs and declined revenues. A combination of the disruptions mentioned above is simulated in the COVID-19 scenario, providing insights regarding the synergetic effects. Since disruptions coincide along the simulation period, a more substantial impact on SC performance and operations can be observed. Inventory levels decrease because of a rapid demand increase at the beginning of the first coronavirus wave, which is intensified when products stop being received because of production stoppages and transport barriers. In turn, the inventory decrease leads to higher lead times and unattended orders, generating an increase in the mean lead time and late orders rate, and entailing an overall reduced ELT service level.

TTR=tfailureSL-trecoveredSL,

(1)

• Aldrighetti R., Zennaro I., Finco S., Battini D. Healthcare Supply Chain Simulation with Disruption Considerations: A Case Study from Northern Italy. Glob. J. Flexible Syst. Manage. 2020;20(1):81–102. [Google Scholar]
• Aldrighetti R., Battini D., Ivanov D., Zennaro I. Costs of resilience and disruptions in supply chain network design models: a review and future research directions. International Journal of Production Economics. 2021;235 [Google Scholar]
• Azzi A., Battini D., Faccio M., Persona A., Sgarbossa F. Innovative travel time model for dual-shuttle automated storage/retrieval systems. Comput. Ind. Eng. 2011;61(3):600–607. [Google Scholar]
• Battini D., Calzavara M., Persona A., Sgarbossa F. Sustainable packaging development for fresh food supply chains. Packag. Technol. Sci. 2016;29(1):25–43. [Google Scholar]
• BMI, 2020. Vorübergehende Grenzkontrollen an den Binnengrenzen zu Österreich, der Schweiz, Frankreich, Luxemburg und Dänemark. Bundesministerium des Innern, für Bau und Heimat. Retrieved 4 February 2021, from https://www.bmi.bund.de/SharedDocs/pressemitteilungen/DE/2020/03/grenzschliessung-corona.html.
• Brintrup A., Pak J., Ratiney D., Pearce T., Wichmann P., Woodall P., McFarlane D. Supply chain data analytics for predicting supplier disruptions: a case study in complex asset manufacturing. Int. J. Prod. Res. 2020;58(11):3330–3341. [Google Scholar]
• Carvalho H., Barroso A.P., Machado V.H., Azevedo A., Cruz-Mahado V. Supply chain redesign for resilience using simulation. Comput. Ind. Eng. 2012;62:329–341. [Google Scholar]
• Cavalcante I.M., Frazzon E.M., Forcellinia F.A., Ivanov D. A supervised machine learning approach to data-driven simulation of resilient supplier selection in digital manufacturing. Int. J. Inf. Manage. 2019;49:86–97. [Google Scholar]
• Chen Y.F., Disney S.M. The myopic order-up-to policy with a proportional feedback controller. Int. J. Prod. Res. 2007;45(2):351–368. [Google Scholar]
• Chinn, D., Sjatil, P., Stern, S., Tesfu, S., Windhagen, E., 2020. Navigating the post-COVID-19 era: A strategic framework for European recovery. McKinsey. Retrieved 4 February 2021, from https://www.mckinsey.com/industries/public-and-social-sector/our-insights/navigating-the-post-covid-19-era-a-strategic-framework-for-european-recovery.
• Choi T.-M. Risk analysis in logistics systems: A research agenda during and after the COVID-19 pandemic. Transp. Res. Part E: Logist. Transp. 2020;140 [PMC free article] [PubMed] [Google Scholar]
• Choi T.-M. Fighting Against COVID-19: What Operations Research Can Help and the Sense-and-Respond Framework. Ann. Oper. Res. 2021 doi:10.1007/s10479-021-03973-w. [PMC free article] [PubMed] [CrossRef] [Google Scholar]
• Chowdhury M.T., Sarkar A., Paul S.K., Moktadir M.A. A case study on strategies to deal with the impacts of COVID-19 pandemic in the food and beverage industry. Oper. Manage. Res. 2020 doi:10.1007/s12063-020-00166-9. [CrossRef] [Google Scholar]
• Chowdhury P., Paul S.K., Kaisar S., Moktadir M.D. COVID-19 pandemic related supply chain studies: A systematic review. Transp. Res. Part E: Logist. Transp. Rev. 2021;148 [PMC free article] [PubMed] [Google Scholar]
• Currie C.S.M., Fowler J.W., Kotiadis K., Monks T., Onggo B.S., Robertson D.A., Tako A.A. How simulation modelling can help reduce the impact of COVID-19. J. Simul. 2020;1–15.14(2):83–97. [Google Scholar]
• Deutschland.de, 2020. The timeline regarding coronavirus in Germany. Retrieved 1 February 2021, from https://www.deutschland.de/es/the-timeline-corona-virus-germany.
• Dolgui A., Ivanov D., Rozhkov M. Does the ripple effect influence the bullwhip effect? An integrated analysis of structural and operational dynamics in the supply chain“ Int. J. Prod. Res. 2020;58(5):1285–1301. [Google Scholar]
• Dolgui A., Ivanov D., Sokolov B. Ripple effect in the supply chain: an analysis and recent literature. Int. J. Prod. Res. 2018;56(1–2):414–430. [Google Scholar]
• Dolgui A., Ivanov D., Sokolov B. Reconfigurable supply chain: The X-Network. Int. J. Prod. Res. 2020;58(13):4138–4163. [Google Scholar]
• Dubey R., Gunasekaran A., Childe S.J., Wamba S.F., Roubaud D., Foropon C. Empirical Investigation of Data Analytics Capability and Organizational Flexibility as Complements to Supply Chain Resilience. Int. J. Prod. Res. 2021;59(1):110–128. [Google Scholar]
• Dubey R., Gunasekaran A., Childe S.J., Papadopoulos T., Blome C., Luo Z.W. Antecedents of resilient supply chains: an empirical study. IEEE Trans. Eng. Manage. 2019;66(1):8–19. [Google Scholar]
• El Baz J., Ruel S. Can supply chain risk management practices mitigate the disruption impacts on supply chains’ resilience and robustness? Evidence from an empirical survey in a COVID-19 outbreak era. Int. J. Prod. Econ. 2021;233 [Google Scholar]
• Fragapane G., Hvolby H.H., Sgarbossa F., Strandhagen J.O. Autonomous mobile robots in sterile instrument logistics: an evaluation of the material handling system for a strategic fit framework. Eur. J. Oper. Res. 2021;294(2):405–426. [Google Scholar]
• Fürstenhans M., Grosse E., Glock C. Smart lighting systems: state-of-the-art and potential applications in warehouse order picking. Int. J. Prod. Res. 2021 doi:10.1080/00207543.2021.1897177. [CrossRef] [Google Scholar]
• Ghadge A., Dani S., Chester M., Kalawsky R. A systems thinking approach for modelling supply chain risk propagation. Supply Chain Manage. Int. J. 2013;18(5):523–538. [Google Scholar]
• Govindan K., Mina H., Alavi B. A decision support system for demand management in healthcare supply chainsconsidering the epidemic outbreaks: A case study of coronavirus disease 2019 (COVID-19) Transp. Res. Part E: Logist. Transp. Rev. 2020;138 [PMC free article] [PubMed] [Google Scholar]
• Frazzon E.M., Freitag M., Ivanov D. Intelligent Methods and Systems for Decision-Making Support: Toward Digital Supply Chain Twins. Int. J. Inf. Manage. 2021;57 [Google Scholar]
• Hosseini S., Ivanov D., Dolgui A. Review of quantitative methods for supply chain resilience analysis. Transp. Res. Part E: Logist. Transp. Rev. 2019;125:285–307. [Google Scholar]
• Hosseini S., Ivanov D., Dolgui A. Ripple effect modeling of supplier disruption: Integrated Markov Chain and Dynamic Bayesian Network Approach. International Journal of Production Research. 2020;58(11):3284–3303. [Google Scholar]
• ILO, 2020. COVID-19 and food retail. International Labour Organization. Retrieved from https://www.ilo.org/wcmsp5/groups/public/ed_dialogue/sector/documents/briefingnote/wcms_741342.pdf.
• IMF, 2020. Germany: Growth rate of the real gross domestic product (GDP) from 2009 to 2021. In Statista – The Statistics Portal. Retrieved 4 February 2021, from https://www.statista.com/statistics/375203/gross-domestic-product-gdp-growth-rate-in-germany/.
• Ivanov D. Disruption tails and revival policies: A simulation analysis of supply chain design and production-ordering systems in the recovery and post-disruption periods. Computers and Industrial Engineering. 2019;127:558–570. [Google Scholar]
• Ivanov D., Das A. Coronavirus (COVID-19 / SARS-CoV-2) and supply chain resilience: A research note. International Journal of Integrated Supply Management. 2020;13(1):90–102. [Google Scholar]
• Ivanov D., Rozhkov M. Coordination of production and ordering policies under capacity disruption and product write-off risk: An analytical study with real-data based simulations of a fast moving consumer goods company. Ann. Oper. Res. 2020;291(1–2):387–407. [Google Scholar]
• Ivanov D. Viable Supply Chain Model: Integrating agility, resilience and sustainability perspectives – lessons from and thinking beyond the COVID-19 pandemic. Ann. Oper. Res. 2020 doi:10.1007/s10479-020-03640-6. [PMC free article] [PubMed] [CrossRef] [Google Scholar]
• Ivanov D. Supply Chain Viability and the COVID-19 Pandemic: A Conceptual and Formal Generalisation of Four Major Adaptation Strategies. Int. J. Prod. Res. 2021 doi:10.1080/00207543.2021.1890852. [CrossRef] [Google Scholar]
• Ivanov D. Springer Nature; Cham: 2021. Introduction to Supply Chain Resilience. [Google Scholar]
• Ivanov D. Exiting the COVID-19 Pandemic: After-Shock Risks and Avoidance of Disruption Tails in Supply Chains. Ann. Oper. Res. 2021 doi:10.1007/s10479-021-04047-7. [PMC free article] [PubMed] [CrossRef] [Google Scholar]
• Ivanov D. Simulation-based Ripple Effect Modelling in the Supply Chain. Int. J. Prod. Res. 2017;55(7):2083–2101. [Google Scholar]
• Ivanov D. Predicting the impacts of epidemic outbreaks on global supply chains: A simulation-based analysis on the coronavirus outbreak (COVID-19/SARS-CoV-2) case“ Transp. Res. Part E: Logist. Transp. Rev. 2020;136 [PMC free article] [PubMed] [Google Scholar]
• Ivanov D., Dolgui A. Viability of Intertwined Supply Networks: Extending the Supply Chain Resilience Angles towards Survivability. A Position Paper Motivated by COVID-19 Outbreak. International Journal of Production Research. 2020;58(10):2904–2915. [Google Scholar]
• Ivanov D., Dolgui A. A digital supply chain twin for managing the disruption risks and resilience in the era of Industry 4.0. Prod. Plan. Control. 2021;32(9):775–788. [Google Scholar]
• Ivanov D., Dolgui A. OR-methods for coping with the ripple effect in supply chains during COVID-19 pandemic: Managerial insights and research implications. Int. J. Prod. Econ. 2021;232 [PMC free article] [PubMed] [Google Scholar]
• Ivanov D., Dolgui A., Sokolov B. The impact of digital technology and Industry 4.0 on the ripple effect and supply chain risk analytics. Int. J. Prod. Res. 2019;57(3):829–846. [Google Scholar]
• Ivanov D., Dolgui A. Stress testing supply chains and creating viable ecosystems. Oper. Manage. Res. 2021 doi:10.1007/s12063-021-00194-z. [CrossRef] [Google Scholar]
• Ivanov D., Sokolov B., Dolgui A. The Ripple Effect in Supply Chains: Trade-off ‘Efficiency-Flexibility-Resilience’ in Disruption Management. Int. J. Prod. Res. 2014;52(7):2154–2172. [Google Scholar]
• Ivanov D. Springer International Publishing; Berlin, Germany: 2018. Structural dynamics and resilience in supply chain risk management. [Google Scholar]
• Li Y., Chen K., Collignon S., Ivanov D. Ripple effect in the supply chain network: Forward and backward disruption propagation, network health and firm vulnerability. Eur. J. Oper. Res. 2021;291(3):1117–1131. [PMC free article] [PubMed] [Google Scholar]
• Kinra A., Ivanov D., Das A., Dolgui A. Ripple effect quantification by supply risk exposure assessment. Int. J. Prod. Res. 2020;58(19):5559–5578. [Google Scholar]
• Llaguno A., Mula J., Campuzano-Bolarin F. State of the art, conceptual framework and simulation analysis of the ripple effect on supply chains. Int. J. Prod. Res. 2021:1–23. doi:10.1080/00207543.2021.1877842. [CrossRef] [Google Scholar]
• Lohmer J., Bugert N., Lasch R. Analysis of Resilience Strategies and Ripple Effect in Blockchain-Coordinated Supply Chains: An Agent-based Simulation Study. Int. J. Prod. Econ. 2020;228 [PMC free article] [PubMed] [Google Scholar]
• Loske D. The impact of COVID-19 on transport volume and freight capacity dynamics: An empirical analysis in German food retail logistics. Transp. Res. Interdiscipl. Perspect. 2020;6 [PMC free article] [PubMed] [Google Scholar]
• Macdonald J.R., Zobel C.W., Melnyk S.A., Griffis S.E. Supply chain risk and resilience: theory building through structured experiments and simulation. Int. J. Prod. Res. 2018;56(12):4337–4355. [Google Scholar]
• Minner S., Transchel S. Periodic review inventory-control for perishable products under service-level constraints. OR Spectrum. 2010;32(4):979–996. [Google Scholar]
• Nagurney A. Supply Chain Game Theory Network Modeling Under Labor Constraints: Applications to the Covid-19 Pandemic. Eur. J. Oper. Res. 2021 doi:10.1016/j.ejor.2020.12.054. [PMC free article] [PubMed] [CrossRef] [Google Scholar]
• OECD Food Supply Chains and COVID-19: Impacts and Policy Lessons. 2020. http://www.oecd.org/coronavirus/policy-responses/food-supply-chains-and-covid19-impacts-and-policy-lessons-71b57aea/ Retrieved 16 August 2020, from.
• Olivares-Aguila J., ElMaraghy W. System dynamics modelling for supply chain disruptions. Int. J. Prod. Res. 2020;1–19 doi:10.1080/00207543.2020.1725171. [CrossRef] [Google Scholar]
• Paul S.K., Chowdhury P. A production recovery plan in manufacturing supply chains for a high-demand item during COVID-19. Int. J. Phys. Distrib. Logist. Manage. 2021;51(2):104–125. [Google Scholar]
• Paul S., Venkateswaran J. Designing robust policies under deep uncertainty for mitigating epidemics. Comput. Ind. Eng. 2020;140 [Google Scholar]
• PwC, 2020. How COVID-19 is changing food retail. https://www.pwc.de/en/retail-and-consumer/how-covid-19-is-changing-food-retail.html. Retrieved 4 February 2021, from https://www.pwc.de/en/retail-and-consumer/how-covid-19-is-changing-food-retail.html.
• Queiroz M.M., Ivanov D., Dolgui A., Fosso Wamba S. Impacts of Epidemic Outbreaks on Supply Chains: Mapping a Research Agenda Amid the COVID-19 Pandemic through a Structured Literature Review. Ann. Oper. Res. 2020 doi:10.1007/s10479-020-03685-7. [PMC free article] [PubMed] [CrossRef] [Google Scholar]
• Rathore R., Thakkar J.J., Jha J.K. Impact of risks in food grains transportation system: a system dynamics approach. Int. J. Prod. Res. 2020:1–20. [Google Scholar]
• Ruel S., El Baz J., Ivanov D., Das A. Supply Chain Viability: Conceptualization, Measurement, and Nomological Validation. Ann. Oper. Res. 2021 doi:10.1007/s10479-021-03974-9. [PMC free article] [PubMed] [CrossRef] [Google Scholar]
• Schmitt T.G., Kumar S., Stecke K.E., Glover F.W., Ehlen M.A. Mitigating disruptions in a multi-echelon supply chain using adaptive ordering. Omega. 2017;68:185–198. [Google Scholar]
• Shen B., Li Q. Market disruptions in supply chains: A review of operational models. Int. Trans. Oper. Res. 2017;24:697–711. [Google Scholar]
• Simchi-Levi D., Schmidt W., Wei Y., Zhang P.Y., Combs K., Ge Y., Gusikhin O., Sander M., Zhang D. Identifying risks and mitigating disruptions in the automotive supply chain. Interfaces. 2015;45(5):375–390. [Google Scholar]
• Singh S., Kumar R., Panchal R., Tiwari M.K. Impact of COVID-19 on logistics systems and disruptions in food supply chain. Int. J. Prod. Res. 2021;59(7):1993–2008. [Google Scholar]
• Sodhi M., Tang C., Willenson E. Research opportunities in preparing supply chains of essential goods for future pandemics. Int. J. Prod. Res. 2021 doi:10.1080/00207543.2021.1884310. [CrossRef] [Google Scholar]
• Svoboda J., Minner S., Yao M. Typology and literature review on multiple supplier inventory control models. Eur. J. Oper. Res. 2021;293:1–23. [Google Scholar]
• Guardian The. A disastrous situation': mountains of food wasted as coronavirus scrambles supply chain. 2020. https://www.theguardian.com/world/2020/apr/09/us-coronavirus-outbreak-agriculture-food-supply-waste accessed on 5th March 2021.
• Veselovská L. Supply chain disruptions in the context of early stages of the global COVID-19 outbreak. Problems Perspect. Manage. 2020;18(2):490. [Google Scholar]
• Wamba S.F., Queiroz M.M. Industry 4.0 and the supply chain digitalisation: a blockchain diffusion perspective. Prod. Plan. Control. 2020:1–18. [Google Scholar]
• Wilson M.C. The impact of transportation disruptions on supply chain performance. Transp. Res. Part E: Logist. Transp. Rev. 2007;43(4):295–320. [Google Scholar]
• WHO, 2020. Rolling updates on coronavirus disease (COVID-19). Who.int. Retrieved 4 February 2021, from https://www.who.int/emergencies/diseases/novel-coronavirus-2019/events-as-they-happen.
• Winkelhaus S., Grosse E.H. Logistics 4.0: a systematic review towards a new logistics system. Int. J. Prod. Res. 2020;58(1):18–43. [Google Scholar]