LLamasoft Supply Chain Blog

← Back to Blog

Short-term Demand Forecasting During COVID-19 – Know Your Customer 

By Amir Kalantari  April 22, 2020

By Amir Kalantari, LLamasoft Product Manager of Demand Forecasting Solutions 

The COVID-19 pandemic has radically impacted customer demand and supply chain operations across the world. As the virus continues to expandconsumer behavior will shift, causing a ripple effect in demand for virtually all industries. While products such as shelf-stable groceries, daily hygiene, bottled water and pet food are experiencing a demand surge because of consumer panic-buying and stockpilingdemand for other products including cosmetics, luxury beauty and skin care has plummetedThis volatility coupled with disruptions in the supply chain has posed planning and operational challenges ranging from inventory allocation and stockouts to transportation and production constraints.

LLamasoft has put together this Demand Disruption Checklist to help you manage short-term disruption and evolve your demand forecasting to be better prepared for future events in the long-term.

While it is impossible to predict a pandemic like COVID-19 and all its consequences, maintaining and adapting operations is possible with the right combination of people, processes, data and technology.  

Knowing how consumers are impacted by COVID-19 is a key step towards developing a short-term demand forecast to make better response decisions. This can be done by collecting and analyzing sales data through POScustomer orders or other methods that capture consumer demand changes for each SKU-region combination. Forecasting models can be augmented with publicly available COVID-19related data such as virus trajectory, lockdown periods, macroeconomic indicators as well as internal business data including product hierarchies, attributes, price and promotions. This approach increases the predictive power of models that only have few weeks of demand history reflecting the current environment. 

According to recent market research by Nielsenduring this pandemic, consumer behavior in most markets follows a very similar pattern which is influenced by external events such as identification of the first case, government health and safety campaigns, border closures and other events that offer early signals of spending patterns. Analyzing demand in countries, such as China and Italy which were ahead of the infection curve, can provide a baseline for other marketsOne of LLamasoft’s clients in the food and beverage industry uses this method to continuously monitor consumer demand data and automatically detect if signals are present that suggest regional disruptionAdvanced machine learning models will generate COVID-19 adjusted forecasts in the short-term that guide vital planning decisions. 

Due to the current high demand volatilityforecasts should be updated frequentlyConsumers are monitoring the COVID-19 situation closely and adjusting their buying behavior according to the latest news. A government who announces reopening their economy can change demand patterns drastically. Companies should capture these changes as quickly as possible to avoid over/under production and stocking. They should also take a proactive approach by performing an extensive scenario analysis to gauge the impact of possible future changes and build a playbook to inform contingency plans.  

Another recommendation is to prioritize time by focusing on products and markets that need more attention. Segmentation is a powerful tool that can help with this task. Like forecasting models, COVID-19 requires a new way of looking at product segmentation that is driven by consumer demand patternsThe simplest version of this segmentation is a two-dimensional matrix that ranks products based on revenue and impact level, both positive and negative. Products that fall into the high revenue and impact bucket require the most level of modeling. While demand for products with low revenue and impact can be forecasted using automatic modelsFor more sophisticated segmentation that includes additional attributes such as margin, price and volume, clustering models can be used to develop multi-dimensional segmentation. 

COVID-19 has exposed the vulnerabilities of many traditional demand forecasting and planning processesAs companies recover from this crisis and restore their operations, they should view this event as a learning opportunity to transform and evolve those processes and prepare for the inevitable next disruption 

Learn how the COVID-19 Demand Impact Analyzer can help you understand how your current demand is being impacted and improve short-term forecasts for immediate response plan action.