Journal of Business Forecasting Articles Issues

As AI in demand forecasting begins to gain traction, we must manage trade-offs between improvements in forecast accuracy and explainability. In this article I reveal the importance of being able to explain how a forecast is derived so executives can make informed decisions without relying on a black box approach. I provide a matrix revealing the trade-offs between performance, explainability, interpretability and accuracy for different forecasting methods of varying complexity. I also introduce three frameworks to explain statistical models, black models, and advanced AI models to facilitate understanding and adoption of your forecasts among decision-makers.

For many of us in the demand planning profession, the pursuit of consensus has become futile. It is a political morass, often unmoored from reality and centered on presenting a number that everyone wants to believe in, even if it makes no sense at all. In this article I reveal the problems with aiming for consensus and how most adjustments to the unconstrained forecast are negative value-add. I provide practical advice to take control of S&OP meetings, including bringing a data-driven forecast, being forearmed with product and business knowledge, holding commercial teams’ assumptions to account, and focusing only on high-value SKUs.

Artificial Intelligent (AI) driven by machine learning has generated significant excitement in the realm of demand forecasting due to advancements in data collection, storage, and processing, as well as techno-logical improvements. Machine learning (ML) can certainly be integrated as another method within the suite of existing forecasting approaches. It’s important to note that there isn’t a single tool— whether a mathematical equation or an algorithm—solely dedicated to demand forecasting. This article delves deeper into Recurrent Neural Networks (RNNs) and their variants, which have gained significant popularity in recent years due to their remarkable ability to handle unstructured sequential (time series) data. These models are called “recurrent” because they process data that unfolds in a sequence, such as text and time-stamped information.

There is currently a trend away from global free trade as more countries impose tariffs on imports. The World Trade Organization (WTO) has expressed concern because it is seeing a bifurcation of trade into two geographically divided blocs. It is estimated that were this to happen, it would result in a loss of 6.4% in global GDP. Previously, as part of the findings from the MIT Supply Chain 2020 Project, I postulated there might even be three or four major trading blocs in the future. This column discusses a strategic planning approach based on “Decision-Making Under Uncertainty”. It recommends using three future scenarios developed by the SC2020 project team as the basis for strategic scenario planning projects aimed at aligning a company’s future global supply chains to the trading blocs that might arise.

I am a Gen Z supply chain professional. Being a part of Gen Z has always had its fair share of negative connotations, “All you kids want to do is stick your nose in a cell phone”, or “You guys have no idea what it was like without a GPS to tell you where to go”, are common among the phrases I heard growing up. While some of these statements may have an element of truth, they don’t define the Gen Z workforce. With the Gen Z population entering and maturing into the work force at a surging pace, it is an appropriate time to understand the nuances of my generation’s values and psychology. In this article I reveal what makes this generation tick, and how to better manage them.

The relationship between sell-in and sell-out in the pharmaceutical industry is a critical one. Sell-in represents the volume of product sold to manufacturers, retailers or distributors while sell-out represents the actual volume sold to the end consumer. Minimizing the delta between the two is key to avoiding burdening different channels with excess stock and maximizing profitability. This article reveals how S&OP and planning tools can be used to minimize the difference. By collaborating with marketing and conducting an effective demand review, the impact of marketing promotions and other demand factors can be identified, allowing for better and more agile inventory decision making.

This article details machine learning for demand forecasting in the cyber security industry. I reveal a standard approach to using machine learning for demand forecasting and compare it to the improved methodology I employed at a leading cyber security firm, which integrates Hierarchical Time Series Forecasting. I discuss how this methodology works and the KPI improvements it yielded relative to the existing approach, plus the strategic decision making benefits it provides.

The new presidential administration is set to bring change, with the rapid pace of its directives likely to impact the economy’s fundamentals. According to Consensus, the nation’s GDP growth rate is expected to increase by 1.45% from Q2 2025 to Q1 2026. Similarly, Wells Fargo projects a slowdown in real GDP growth to 1.78% over the same period, as the inflationary effects of tariffs reduce real income growth, thereby weighing on consumer spending.

Artificial Intelligence (AI) is revolutionizing demand planning. The key objective of demand planning is to grow and manage demand. AI provides tools to achieve both. To increase demand, AI provides tools such as dynamic pricing, personalized offerings, new channels of distribution, and Augmented Reality. To manage demand, we need to balance supply and demand, make the supply chain more agile, retain more customers, and optimize the product portfolio. In this article, I discuss how we can use AI tools to do all these much more effectively.

There is little debate as to the efficacy of AI-based demand forecasting when implemented sucessfully, but most implementations fail. Failure is due in large part to organizations being unprepared for such a wide-ranging and transformative initiative and not understanding the foundational elements required to support it. Three pillars underpin successful AI/ML planning transformation, namely Data, Technology, and Organizational Readiness. In this article I discuss the nuances within these pillars and how they combine to form an effective AI-forecasting and planning ecosystem. Key questions to ask are provided, designed to help you assess your own readiness for AI / ML-based forecasting.

Applying AI to forecasting and planning can seem a daunting, even overwhelming, challenge. However, applying this game-changing technology is often far simpler than many Demand Planners believe. Tools like Power BI and Excel features allows us to start small with AI forecasting, gaining small wins and allowing us to integrate AI into our existing planning processes over time. In this article, I reveal how to take your first steps in AI, how new platforms no longer require data science or coding skills, and how this technology, far from replacing Demand Planners, elevates our value.

Forecast Value Added (FVA) Analysis, a tool for assessing the effectiveness of demand forecasting processes, has traditionally focused on understanding how each step in a forecast process contributes to or detracts from overall accuracy. With the advent of generative AI and large language models (LLMs), new possibilities for enhancing FVA analysis are emerging. This article examines how generative AI and LLMs can be integrated into FVA to improve forecasting accuracy and process efficiency. Importantly, it considers current limitations and potential future developments which may remedy those limitations.

AI will shape everything we do, and an organization’s AI payoff hinges on doing the hard work of changing its culture. Becoming an AI-driven company requires a business-led roadmap, a purpose-driven vision, the right organizational culture, talent management, agile processes, and data and technology. It is imperative that executive leadership lead AI transformations and that use cases focus on measurable business impact. In this article I reveal a practical framework for becoming an AI-driven company capable of supporting effective AI implementation.

In an environment where demand planning is increasingly difficult, machine learning (ML) models help generate accurate forecasts and allow us to better plan the business. Machine learning for demand fore- casting goes beyond traditional statistical models using just sales data. In this article, I discuss how ML models can use a variety of data inputs including consumption data, promotions, social media, weather, economic data, and more. I also reveal the importance of ‘forecast explainability’ when using ML. This avoids a ‘black box’ approach and allows us to understand the individual components of the forecast and their contribution, helping us to understand and better communicate the demand drivers.

There are numerous challenges facing the current supply chain sector, including unpredictable demand and a lack of stability among suppliers, both of which impact efficiency. Accurate prediction of supplier Lead Time is crucial for efficient inventory management, production scheduling, and service level improvement. This article presents an innovative solution by combining multi-source information with artificial intelligence (AI) algorithms to enhance prediction accuracy, to generate more reliable predictions of supplier delivery times. This integration improves inventory management and production planning, ultimately enhancing overall supply chain resilience.

There are numerous subcategories of Artificial Intelligence. Robotics, machine vision and natural language interfaces are all subsets of AI. Expert systems are also AI. I could argue most supply chain planning solutions are expert systems by nature. When I hear folks talking about AI in planning, the ‘function’ or ‘task’ most often mentioned is the leveraging of machine learning algorithms to statistically forecast or optimize inventory, or to integrate and analyse large volumes of data into logical constructs.