It’s no secret that the COVID-19 pandemic generated historically high demand volatility – Many manufacturers struggled to satisfy surges in demand for various products, while numerous retailers are still sitting on vast amounts of unsold inventory.
This worldwide event revealed a key shortcoming with demand planning today: The vast majority of businesses rely solely on historical demand patterns for their demand planning, and do not understand how economic activity impacts demand for their products.
In this post, I illustrate how the pandemic impacted consumer behavior, and how it revealed process gaps in the demand planning function. I also show some simple ways that businesses can incorporate external market factors into their demand planning methodology. Not only will this make the planning cycle more agile, this provides business leaders with invaluable insight into demand for their products.
Unexpected Shifts in Consumer Demand
Given that the virus spreads more quickly in crowded areas, it makes sense that consumers have shifted away from buying merchandise at retail stores and shopping malls. The pandemic accelerated the trend to online shopping, and companies like Amazon recognized a surge in demand for products across categories. Already struggling to compete with e-commerce, physical retailers have been hit especially hard, many left with vast amounts of unsold inventory.
Some Industries Realized a Surge in Demand
Consider the following scenario: Nearly all spectator events are cancelled, and popular tourist destinations and cruises are similarly shut down. You are working from home, staring at the same four walls every day. There is nowhere to go, and it’s Spring. How will you spend your time? You guessed it – Home improvements. Big box retailers reported surges in sales in the Summer of 2020, far beyond the typical seasonal shift.
Stuck at home, consumers searched for anything and everything to make their living spaces a little more pleasant. The supply chain was not prepared for this onslaught of consumer demand, and many manufacturers simply had never contemplated this.
The Apparel Industry Was Hit Hard
At the other end of the spectrum, the COVID-19 pandemic yielded rapid declines in consumer demand for many other industries. Much of the white-collar workforce began working from home in March 2020. As a result, there was very little demand for items related to officework. For instance, demand for office attire dropped dramatically, and the clothing industry is still stuck with months of unsold inventory.
Why Were Planning Teams Caught Off Guard?
When the COVID pandemic first took root in China, country officials acted quickly to shut down large portions of the country. Factories were idled overnight, and much of the workforce was quarantined in their residences. Shortly after, European countries and the US experienced the outbreak and also tried to contain the outbreak by idling huge swaths of their economies. It is rare that we see a complete shutdown of large portions of the economy, but this is precisely what happened in March 2020.
Most businesses were caught off-guard with this impact, largely because their demand planning processes are based almost exclusively on historical data. Simple demand forecasting models likely utilize a rolling average of recent demand, adjusted for seasonality. Complex forecasting models utilize complex autoregression algorithms to predict future demand. Regardless of the complexity of the model, if you are using only internal data to forecast future demand, you would have had no way to predict the impact of the COVID-19 pandemic.
Inventory Levels Jump
As noted above, most businesses were simply caught off guard with the pandemic, as their forecasting models are autoregressive, meaning that they use historical data to predict future demand. As an example, I looked at Dillards, a brick and mortar retailer that is focused largely on selling business attire. Because it conducts business at a physical store, and due to its reliance on business attire, this business is highly impacted by the pandemic.
Dillards SEC filings reveal the business held 73 days of Inventory in August 2019. This increased to 180 days in August 2020, an increase of over 100%. This is an extraordinary increase for this large retailer.
Solution: Use External Data to Predict Demand Shifts
By incorporating external indicators into their demand planning methodology, businesses can better predict the outcomes based on known market events. This allows the demand planning function to incorporate events that are not predicted within historical data set.
As early as January 2020, there were already signals that the pandemic could seriously jeopardize the retail industry. If the demand planning teams at our largest retailers utilized a regression with national retail sales data, they would be better able to make predictions based on potential outcomes of the pandemic.
For instance, if you knew your business was highly correlated to national retail sales, perhaps you could use industry-based retail forecasts as an input to your demand forecast. As another example, if your business is highly correlated with new construction, then you may use new construction starts as an indicator of demand.
Step 1: Identify Factors That Drive Customer Demand
You know your business better than anyone – What factors do you think would impact sales to your business? Do you produce agricultural products that may be impacted by weather trends? Do you sell athletic apparel that may be impacted by current fitness trends? Once you identify some market factors, you can identify if there is quantifiable data that will measure these trends.
Early in my finance career, I valued an equity interest for an owner of one of the largest festivals in the United States. To normalize cash flows, I developed a correlation between weather patterns and fair attendance. In fact, I was able to quantity the dollar impact for each day of rain, or mostly cloudy days. The best correlations are very simple and intuitive.
Step 2: Prove Correlation with Regression
Once you identify a potential correlation to your actual historical demand, you need to test your theory. Statisticians use a simple regression to identify correlation between two sets of time series data. A simple regression takes two sets of time series data and identifies whether one set of data is a predictor of the other. It’s that simple. Microsoft even has a free add-in for Excel that allows a simple regression analysis. I won’t go into the detailed mechanics of how to develop a regression analysis, but this tutorial provides a nice overview.
Step 3: Incorporate Linear Regression in Demand Planning
Using search data as a predictor of demand
Social signals provide near real-time indicators of consumer behavior. Imagine being able to capture instantaneous snapshots of trends in networks like Google, Pinterest, and Twitter. In fact, these platforms indeed share historical trend data that shows historical search trends for given key words.
In the below, I share an example of how businesses can revise their demand planning methodology to incorporate real-time Google trend data.
Example: Search Trend Reveals Spike in Remodeling Interest
The above figure illustrates Google search trend data for the search term “Remodeling”. I highlighted in green the obvious peaks in search activity over the past five years. Typically, searches for remodeling peak around January to February each year. While search interest remains elevated in early spring, it gradually drops down after the peak. In 2020, it is obvious that there was a second peak in May/June. As I noted earlier in this article, demand skyrocketed for home remodeling products in the Summer of 2020.
Let’s assume you were working in the demand planning function at a manufacturer of products used in home remodeling (i.e plumbing fixtures, etc). If you incorporated this data into your plan in March, you may have correctly anticipated a spike in demand in mid-Spring.