If you’ve spent any length of time in demand planning, you know the drill: Forecasts built on historical data, endless Excel manipulation, and the elusive white whale of “the perfect plan.”
I bet you’ve also noticed how much the supply chain planning landscape has evolved over the years. Not just the rise and adoption of advanced planning systems (APS), but also the need for more scenario-based modeling, adapting to operational constraints, and the strategic gaps left by traditional forecasting methods.
As a supply chain consultant, I am often asked about AI and how it will impact our collective work. Having just wrapped up a project where we integrated AI forecasting with a leading APS, I thought it might be helpful to explain how and why companies are leveraging AI in demand planning.
If you don’t make it until the end, here’s the big idea: AI forecasting is not a silver bullet. But it can be used to accelerate demand planning decisions. It can help your company move beyond reactive planning and into a space where you can actually anticipate change, not just respond to it. In my mind, the future of demand planning includes AI.
The challenges of traditional time series forecasts
If necessity is the mother of invention, it’s not surprising that AI is being used to fill some of the demand planning gaps that exist today. Traditional time series-based forecasting is a powerful asset for any company, but it’s not without some challenges:
- Data quality and availability: Accurate time series forecasting requires high-quality historical data. Incomplete, inconsistent, or inaccurate data can lead to poor forecasts. Collecting and maintaining quality data is often a significant challenge for many companies.
- Short product lifecycles: Many consumer goods and hi-tech products have short life cycles, requiring rapid adjustments to forecasts. Plus, time series models struggle with new product launches where there is no historical data to base predictions on.
- Promotions and advertising effects: Promotional events and advertising can cause sudden spikes or drops in demand that are not easily predicted or understood by traditional time series models, which rely heavily on historical data patterns without understanding the context.
- External factors and sudden shifts: Factors such as economic shifts, regulatory changes, or unforeseen global events (e.g., health crises) can quickly render historical data irrelevant if those contexts dramatically affect consumer behavior.
- Forecast granularity: Time series models can sometimes struggle to provide forecasts at the granularity needed (e.g., by store or by product variant) due to data limitations or increased complexity, leading to a pretty big trade-off between detail and accuracy.
- Scalability: As the business grows, managing and updating multiple time series models becomes more complex—requiring automated systems that can scale and adapt without sacrificing accuracy.
If you’re experiencing one or more of these challenges, then you might benefit from AI-based forecasting—wholly or in part. Time series forecasting can be combined with machine learning and other AI-driven methods to create more flexible, adaptable, and context-aware forecasting processes.
The benefits of integrating AI forecasts with an APS
The benefits of AI in supply chain are pretty straightforward at this point: AI can help companies improve the accuracy, efficiency, and responsiveness of their supply chain and business operations.
But let’s be more specific. Why AI in demand planning? What are the benefits of integrating AI forecasts directly with an APS?
- Improved forecast accuracy: AI models can analyze large volumes of historical data and identify complex patterns that traditional methods might miss. This results in more accurate demand forecasts, reducing the risk of overstocking and stockouts.
- Connected decision-making: By providing precise demand projections, AI enables more informed decision-making across the supply chain. This includes inventory management, production scheduling, and supply chain logistics, ensuring resources are allocated efficiently.
- Dynamic and real-time forecasting: AI tools can process data in real-time, allowing businesses to adjust their forecasts promptly in response to sudden changes in demand or supply chain disruptions.
- Customization and segmentation: AI can provide insights at a granular level, allowing businesses to tailor forecasts for specific products, locations, or customer segments. This enables more personalized inventory strategies and promotions efforts.
- Lower operational costs: By optimizing inventory levels and reducing waste, AI forecasts can lower holding costs, minimize markdowns, and reduce expedited shipping expenses.
- Improved collaboration and communication: Accurate AI forecasts can align different departments within an organization, such as marketing, sales, and operations, fostering better collaboration and communication. This alignment ensures all stakeholders work towards common goals based on reliable data.
- Scalability: AI systems can easily scale as a business grows, accommodating increased data volume and complexity without a loss in performance. This flexibility makes it easier to consider expanding operations or enter new markets segments.
- Risk mitigation: By providing early warnings about potential demand spikes or drops, AI forecasts help businesses proactively manage risks associated with supply chain disruptions, ensuring continuity and reliability.
- Enhanced customer satisfaction: By ensuring product availability and reducing delivery times, AI-enhanced demand planning can lead to greater customer satisfaction, loyalty, and repeat business.
Simply put: AI in demand planning can transform your entire supply chain from being reactive to proactive, driving not only operational efficiencies but also strategic growth and competitive advantage.
How to integrate AI forecasts with an APS
APS tools like Blue Yonder, Kinaxis, and ketteQ help optimize production schedules, allocate resources, and formulate demand and supply plans. When combined with AI-based modeling to not only generate a forecast but also support demand sensing efforts, APS unlock higher levels of precision and flexibility across the planning horizon.
Leveraging an improved and reliable baseline forecast allows for a more efficient consensus forecasting process that focuses the cross-functional team on those forecast intersections that require product or market insights to drive an intelligent demand signal.
Based on the work we do at Spinnaker SCA, here’s how I recommend getting started:
- Define what “success” looks like. Clearly outline what you aim to achieve with AI forecast integration into the demand planning process, such as improving base forecast accuracy, creating a low touch demand signal or enhancing responsiveness to market conditions or consumer behaviors.
- Set the baseline. Assess your existing APS tool to understand its capabilities, limitations, and integration possibilities. Ensure the solution can seamlessly receive and process input from an AI forecasting tool.
- Determine your AI forecasting requirements. Consider AI tools that provide forecast explainability within their solution (e.g., primary drivers and percent contribution) and ensure it aligns with your expectations for level of detail (e.g., by forecast intersection or time/planning bucket). Also consider AI models or supporting tools that align with your specific forecasting requirements, like time series forecasting, machine learning models, or hybrid approaches—depending on the complexity and quality of your data.
- Integrate the AI data. Identify what information is needed by the AI tool to improve the relevancy of the generated forecast (e.g., discontinued dates). Then establish a robust data pipeline for integrating AI forecasts with the APS system(s). This involves data extraction, transformation, and loading (ETL) processes to ensure data consistency and quality.
- Iterate and fine-tune. Decide on the need for real-time data processing based on the business context. Some operations might require immediate updates (e.g., new products or customers), while others can be handled in batch processes (e.g., weekly demand forecasts).
Change management: The real engine behind AI success
While implementing AI-based forecasting in demand planning can lead to significant improvements, I cannot overstate the importance of effective change management. From people to policies and metrics, change management is what determines the ultimate success or failure of any supply chain transformation—including AI.
Still figuring out how AI fits into your demand planning process? You’re not alone. Whether it’s choosing the right tool or making it work with your APS, I’d love to help you think it through. Let’s have a conversation to see how Spinnaker SCA might be able to help.
