Short-Term Demand Forecasting During Disruption — Know Your Customer
The COVID-19 pandemic radically impacted customer demand and supply chain operations around the world. As the virus spread, consumer behavior continued to shift, causing a ripple effect in demand for virtually all industries.
In the beginning of the pandemic, products such as shelf-stable groceries, daily hygiene, bottled water, and pet food experienced a demand surge due to consumer panic-buying and stockpiling. At the same time, the demand for products like cosmetics, luxury beauty, and skin care plummeted. This volatility, coupled with disruptions in supply chain, posed supply chain planning and operational challenges ranging from inventory allocation and stockouts to transportation and production constraints.
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. It’s exactly these abilities that we foster in our Coupa demand solutions and digital supply chain tools.
Demand Forecasting During and After Disruption
Knowing how consumers are impacted by a crisis such as COVID-19 is a key step toward developing a short-term demand forecast to make better response decisions. That might sound like a daunting task, but businesses already have much of the necessary information available to identify those impacts. They must collect things like:
- POS sales data
- Customer order history
- Inventory levels
- Demand changes in SKU-region combinations
- Supplier shortages
Once fed into a model, businesses can easily see shifting trends that may impact their operations. Such forecasting models can likewise be augmented with publicly available data. In the case of COVID-19 that might include virus trajectory, lockdown periods, hospitalization data, and relevant macroeconomic indicators. Put together, businesses gain a predictive power even when using a model that only has a few weeks of demand history reflecting the current environment.
Why Use Consumer Behavior Data?
Traditional forecasting models are built with historical data, and their error rate rises over time. They’re effective for understanding long-term trends or making historically contextualized long-term plans.
However, a sudden crisis like a once-in-a-century pandemic can throw a monkey wrench into these models. A model that works based on historical data would not have access to relevant data with which business owners can make decisions.
That’s where consumer behavior comes in. In most markets, it follows a very similar pattern influenced by specific external events. These included things like:
- Infection rate
- Government health and safety campaigns
- Border closures
Since consumer behavior was so reactive to these events, it provided a means for businesses to assess potential scenarios and consequences using data that was already readily available to them.
Businesses could even take it a step further. Analyzing demand in countries that were ahead of the infection curve provided a baseline for demand planning in other markets regarding spending patterns that were about to arise. For example, one of our clients in the food and beverage industry uses this method to continuously monitor consumer demand data and automatically detect if there are signals that suggest a regional disruption. Advanced machine learning models will generate COVID-19 adjusted forecasts in the short term that guide vital scenario planning decisions.
Getting the Most Out of Short-Term Demand Forecasting
Short-term demand forecasting in times of disruption gives businesses insight to keep pace with markets and events. However, to get the most out of these tools, it’s necessary to do two things.
1. Match Consumer Reactivity and Proactivity
Amid ongoing demand volatility, forecasts should be updated frequently. For example, consumers are monitoring the COVID-19 situation closely and adjusting their buying behavior according to the latest news. A government that announces the reopening of its economy can change demand patterns drastically — as can one that announces border closures or a renewed lockdown.
Short-term demand forecasting can help companies capture these changes as quickly as possible to avoid over/under production and stocking. However, they should also take a proactive approach by performing extensive scenario analysis to gauge the impact of possible future changes. This will lay the groundwork to build a playbook of contingency plans.
2. Segment Your Products and Supply Chain
Like forecasting models, a disruptive event requires a new way of looking at product segmentation that is driven by consumer demand patterns. Specifically, it’s now crucial to focus on products and markets that need more attention.
The 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 category require the most level of modeling. In contrast, demand for products with low revenue and impact can be forecasted using automatic models. For more sophisticated segmentation that includes additional attributes such as margin, price, and volume, clustering models can be used to develop multi-dimensional segmentation.
Segmenting supply chains is one element of a sourcing maturity model. It’s also among the best ways to prepare for — and mitigate — disruption.
Work With Understanding, Not Unknowns
Demand forecasting has taken on new importance for businesses and supply chains, boosting the importance of more accurate supply chain analytics. COVID-19 has exposed the vulnerabilities of many traditional demand forecasting and planning processes. Using carefully aggregated historical data simply doesn’t work in situations where the last historical precedent happened a century ago.
Yet, without demand forecasting, businesses cannot hope to navigate the rapidly shifting situation that has impacted every level of operations. That’s led many companies to turn to consumer behavior data as a means to predict what step they must take next.
As companies continue to recover from this crisis and adapt their operations, they’re waking to the reality that the next disruption is inevitable. The time to transform and evolve processes to develop supply chain resiliency is now.