Forget for a moment everything you thought you knew about supply chain management. In the midst of a tumultuous era marked by supply chain challenges, global pandemics, geopolitical conflicts, and numerous disruptive events, it's time to bring back the F-word: Forecasting. Yes, forecasting - the fundamental practice that has been unjustly banished to the shadows amidst the chaos.
As the business landscape drastically transformed following COVID-19 and subsequent disruptions, urgent priorities such as maneuvering around capacity constraints, and allocating limited resources such as materials and talent overshadowed basic tenants of supply chain: forecasting and demand planning.
The profound impact of the pandemic on supply chains exposed vulnerabilities and weaknesses that were previously overlooked, and businesses are still grappling with the aftermath, attempting to rebuild and fortify their supply chains against future uncertainties—a resilient supply chain.
We all remember how some products experienced an unprecedented surge in demand. Anticipating higher usage and to reduce the frequency of shopping trips, consumers reacted to the crisis by stocking up on essential CPG goods. However, this buy-ahead behavior created a perceived shortage of essential items like toilet paper and hand soap, which fueled panic-buying among consumers. Moreover, products like bidets and white vinegar witnessed inflated demand as consumers sought alternative solutions. Meanwhile, other businesses witnessed a complete evaporation of demand. These dynamics created a ripple effect that permeated across various sectors and industries.
For a number of companies, for example Peloton, the demand was unusually inflated during lockdown restrictions and the closure of gyms, fitness centers, and recreation leagues. Additionally, target consumers were unable to travel, shop or dine out, causing a unique intersection of both available discretionary cash and a desire to feel more connected. The opposite occurred with brick-and-mortar stores, airlines, restaurants, venues and more. As an example, US passenger miles flown in 2020 Q2 were 12% of 2019 Q2 totals.
Peloton’s unexpected sharp increase in sales created an imbalance of supply and demand, and the company struggled to keep up with orders. Peloton started expanding its manufacturing facilities in anticipation of multi-year growth, but the company’s luck turned all at once after 2021. The drop was nearly as sudden and steep as the company’s rise, and Peloton stock went straight down, losing approximately 80% of its value. Drastic and sudden changes in demand like the one experienced by Peloton highlight the need for demand sensing and effective forecasting – we’ll elaborate more on that below.
Other stories of big companies struggling during this period come to mind. For General Motors, the main issue has not been demand, but supply. Shortages, production issues and bottlenecks caused by the pandemic have had serious repercussions for the company, and a mere two vehicles of their hyped Hummer EV trucks have made their way to customers so far in 2023.
Let’s take a moment to reflect on past experiences, and the ever-important need to adapt demand planning approaches in the face of changing market dynamics.
Understand Demand Drivers
The pandemic triggered significant shifts in consumer behavior, both short-ranging and long-term, requiring a comprehensive understanding of demand drivers. Across different sectors and as the world gradually returned to normality, certain products and services that experienced a surge in demand, were expected to return to pre-pandemic levels, but that hasn’t necessarily been the case. The home office category boomed during the pandemic, as millions of people around the world switched to working from home. Even though it has slowed down considerably, the demand remains strong as there continues to be a need for accommodating on-demand work areas at home.
Recently, inflation and fears of a recession emerged as new factors impacting demand. Rising prices and changes in purchasing power can significantly influence consumer behavior, making it crucial for businesses to understand and anticipate these shifts to make informed decisions.
To navigate these challenges, supply chain planning technology equipped with AI and machine learning (ML) has become increasingly essential, enabling companies to analyze vast amounts of data and identify the causal variables that drive changes in demand.
These advanced technologies have the capability to detect patterns, correlations, and trends that human analysis alone may struggle to uncover. They enable businesses to develop accurate demand models that take into account factors such as the impact of inflation on consumer preferences, macroeconomic indicators, and overall demand patterns – delivering powerful insights to make informed strategic decisions. According to McKinsey, applying AI-driven forecasting to supply chain management can reduce errors by between 20 and 50%, and result in a significant reduction of up to 65% in lost sales and product unavailability.
Don’t Forget About Demand Sensing
To effectively anticipate and respond to demand, companies must engage in demand sensing, which responds to market shifts and changes in buying behavior and is typically focused on a shorter time horizon such days or weeks. Accurate demand sensing is crucial in predicting and planning for demand fluctuations, as it involves considering causal variables, demand drivers, and macroeconomic impacts that influence consumer behavior and purchasing patterns. By relying on outside-in signals, such as point-of-sale data and other external sources, businesses can gain valuable insights that go beyond their internal supply chain operations.
The incorporation of external data sources allows companies to understand the broader market dynamics and factors that impact demand, while identifying emerging trends, changes in consumer preferences, and shifts in buying patterns. This provides a deeper understanding of true demand to adjust planning strategies, as required.
Additionally, demand sensing allows organizations to consider the impact of inventory trade-offs on consumer behavior. By aligning inventory levels and availability with customer demands, businesses can optimize their supply chain operations and improve their ability to meet customer expectations.
The Value of Causal Modeling
Causal modeling is an essential tool for success in demand planning. This methodology provides companies with insights for understanding true demand, and account for inventory imbalances or one-time buys that may create black swan events.
It is important to look beyond the immediate future and consider the broader impact of factors like housing markets, energy costs, inflation, and global financial markets. Leveraging advanced techniques, such as AI and ML, enables organizations to discern the underlying drivers behind these impacts, ensuring a more accurate data-driven perspective rather than relying on vague and aggregated signals and do so without an army of data scientists.
In the past, for example, causal modeling of store-level data for numerous SKUs across multiple stores was seen as requiring a team of data scientists or statistical forecast analysts to fine-tune the models. Time-series forecasting was the conventional approach adopted by many companies due to the large investment of time and resources required. However, with the advent of more powerful planning platforms, organizations can leverage advanced technologies to gain a more accurate and data-driven perspective, moving away from relying solely on vague and aggregated signals.
AI, for example, can empower companies to perform data cleansing, data transformation and enrichment, as well as feature selection across hundreds of thousands or even millions of combinations, yielding robust predictors and avoid overfitting. A well-designed AI system enables businesses to re-forecast products on a daily basis, moreover, AI enables detailed analysis of each SKU and store, allowing the identification and adjustment of influential variables, while also pinpointing stores that may have been affected by stock-outs, pricing fluctuations, or other non-seasonal impacts that could potentially cloud demand predictions. AI allows for a 'Think Locally' approach, considering the impact of regulations, regional holidays, weather conditions, socio-economic factors, population data, and even census information that remains concealed when data is aggregated for forecasting and prediction purposes.
Get Your Demand Planning Back on Track – Embrace the Supply Chain F-word Once Again
Consumer behaviors continue to change rapidly, while macroeconomic conditions remain uncertain. These times are challenging businesses to rethink the status quo and reassess their demand planning strategies. The lessons ought to serve as an important reminder that forecasting deserves its rightful place as part of your comprehensive supply chain strategy, to help you effectively navigate an unpredictable future.
Forecasting and demand planning require constant vigilance and adaptability, even (and especially) during turbulent times. Traditional methods that rely solely on only outside-in historical sales data and basic statistical methods for demand forecasting fail to accurately predict demand when patterns fluctuate significantly or when dealing with unforeseen events. So, it’s now crucial to shift the approach and embrace a data-driven paradigm.
To adjust and regain track, it’s key to explore a variety of data sources and understand how they interrelate; and validate data across multiple channels to capture the true demand within each product category. The abundance of diverse data sources paves the way for the use of ML technology to improve future demand predictions. By employing predictive models powered by AI, companies can make real-time adjustments to demand forecasts, thereby minimizing the financial impact of flawed predictions.
Let us learn from past experiences and embrace the importance of demand planning in shaping resilient and prosperous supply chains.