The COVID pandemic is said to be the biggest economic disruption since WWII, and its impact on supply chain operations has been enormous for almost all industries. In this article, we discuss the main challenges that the pandemic has imposed on supply chain planning and examine case studies on how two Fortune 500 companies responded to the disruption using data science methods.
The COVID-19 outbreak has impacted supply chain operations in many ways. First, most countries announced lockdown measures in March 2020, and this immediately resulted in major shocks for all retail sectors. Some sectors, such as supermarket stores, experienced a dramatic demand surge due to consumer panic and had to urgently reconfigure their supply chains to focus on high-demand items. Many other sectors, such as apparel and department stores, were ordered to shut down their brick-and-mortar operations, which completely disrupted their merchandise procurement processes, which are typically planned seasons in advance. Many large retailers, such as Macy’s, Kohl’s, and Gap, had to delay payments to their suppliers or terminate contracts to accommodate.
Second, lockdown has sharply changed the long-term demand patterns. Similar to in the initial panic phase, some sectors were impacted positively while others were impacted negatively. For instance, video game publishers experienced a major long-term surge in demand that generally invalidated their regular supply and promotion plans. Moreover, the surge patterns could be different across countries, requiring multinational companies to develop specialized supply chain response strategies for each country.
Finally, the reopening period and secondary lockdowns imposed their own challenges. The demand and operation patterns after the lockdown were very different from the normal patterns, so companies had to figure out how to factor new considerations into their inventory, price, and workforce planning and management models. We next consider two case studies that show how two large companies navigated through the challenges of supply chain management in a pandemic using advanced analytics.
A leading developer of media products used a demand forecasting model to optimize pricing and promotional decisions across more than 70 countries, multiple product lines, and multiple digital and brick-and-mortar retail channels. The model was able to forecast weekly demand numbers for up to 24 months ahead with reasonable accuracy, and the system generated multiple forecasts for different discount levels. Pricing managers used these forecasts to understand how price elasticity would change over time and to determine optimal promotional times and depths.
The company faced an unprecedented problem when the COVID-19 outbreak in China and subsequent quarantine measures rapidly changed the demand pattern, so the actual sales numbers quickly diverged from the forecast. In the case of media products, demand numbers sharply surged compared to the forecast. In a few weeks, it became clear that COVID-19 had turned into a massive global disruption that invalidated demand forecasts for all geographies.
The company made a decision to immediately update the forecasting models. After the initial analysis and experiments, the data science team decided to modify the existing solution as follows:
●A secondary demand forecasting model was introduced to estimate the forecasting error of the primary model.
●Additional signals, such as changes in quarantine measures and university closures, were incorporated into the models, which helped to improve accuracy.
●Global data were used to enable transfer learning across countries and leverage the experience of geographies where the pandemic had started earlier to forecast the trajectories in countries that were impacted later.
The model adjustments were implemented in about two weeks, which helped the forecasting accuracy to return to acceptable levels and stabilized the revenue management process. This was a major contribution toward the resilience of the company in this time of crisis. The analysis and modeling of demand shocks also produced useful insights for operations and leadership teams. For example, it turned out that the demand in Asian countries (e.g., China, South Korea, and Japan) surged much more rapidly after the introduction of quarantine measures compared to in European countries, the United States, Canada, and Australia. This helped to segment countries into several groups and facilitate the analysis and operationalization of the insights produced.
A leading clothing company that operates more than 900 stores worldwide faced massive store closures in March 2020 as the lockdown measures were introduced. The management quickly realized that the demand and operational patterns after the end of the lockdown would be very different compared to the regular patterns, and rapid accommodation to the new environment would be critical for survival. The company deployed several data science teams that were tasked to develop optimization models for post-lockdown operations. One of the use cases was the optimization of the store workforce, that is, the number of store associates on duty and their schedules.
Workforce optimization is generally based on store traffic forecasts, and the company had traffic forecasting models for this. However, it was clear that these models would not work for the opening period. Consequently, the data science team responsible for this track researched and implemented the following extensions:
●The team built a proxy model that estimated the impact of regular seasonal influenza on store traffic. This model then was recalibrated using limited COVID data.
●The second extension was the model that quantified the impact of macroeconomic factors. This model helped to establish the link between the patterns of the economic crisis of 2008, which was the latest major disruption, and the patterns of 2020.
●Finally, Google Community Mobility reports were used to incorporate the ongoing changes in consumers’ mobility.
The above approach helped to adjust the legacy traffic forecasting model and produce a reasonably accurate workforce optimization policy.
The above two case studies show that advanced analytics can really help to stabilize supply chain and pricing operations in case of major shocks, but they require setting realistic goals and acting quickly rather than pursuing high accuracy, leveraging innovative data sources and modeling techniques, and involving an experienced data science team or technology partner to implement the solution in time.
Ilya Katsov is Head of Data Science at Grid Dynamics, a global data science and engineering services company. He leads the effort to help major retail, manufacturing, and technology companies become successful AI adopters and deliver innovative AI solutions. Prior to joining Grid Dynamics, Ilya worked at Intel Research on emerging wireless communication technologies. He is the author of several scientific articles and international patents, and also authored a book, “Introduction to Algorithmic Marketing: Artificial Intelligence for Marketing Operations” (2017).