Executive Guide to Forecasting in Manufacturing, Part 7: Forecast Hierarchy

Every planner knows the conversation: the finance team wants an annual forecast, operations needs weekly demand, and sales insist on SKU-by-customer visibility. All are “right”, and all depend on hierarchical forecasting: producing coherent views of demand across organizational levels and time granularities.

Hierarchical structures are everywhere: by product, by region, by customer, or by time (daily, monthly, quarterly). The challenge is that different forecasts at each level rarely add up, creating confusion and rework.

As part of our book review series based on Demand Forecasting for Executives and Professionals by Stephan Kolassa, Bahman Rostami-Tabar, and Enno Siemsen, we will explore why these inconsistencies happen, the main ways organizations deal with them, and what business leaders should keep in mind when aligning forecasts across levels.

Key Insights from the Book

Below are some of the key insights discussed in the book:

Different hierarchies, different needs

Forecasting is rarely one-dimensional. A retailer may need store-SKU forecasts for replenishment and brand forecasts for marketing. A manufacturer plans materials by bill of materials, not by SKU. Recognizing crossed hierarchies (product × region, customer × brand) is the first step toward coherence.

Trade-off between detail and accuracy + how to deal with it?

There is an inherent trade-off between detail and reliability in forecasting. Highly detailed forecasts provide more operational insight, for example at the level of individual products or locations, but they are often noisier and less stable. Forecasts on a more aggregated level are smoother and easier to predict, but they may lack the granularity needed for operational decisions.

Organizations typically address this challenge in several ways. A bottom-up approach forecasts at the most detailed level (such as product variant by plant and customer) and then aggregates the results. While intuitive, it can be sensitive to noisy data. A top-down approach forecasts total demand first and then allocates it across products or regions, which works well when demand patterns are stable but struggles when demand patterns change. A middle-out approach forecasts at an intermediate level, such as product group by region, providing a practical compromise between detail and stability.

A more advanced method is optimal reconciliation, where forecasts are generated at multiple levels and then reconciled to ensure consistency across the hierarchy. This can improve accuracy, but it is more computationally demanding and harder to manage in complex organizations.

Coherence is not always desirable

For risk-related decisions such as safety stock, forcing numbers to add up across levels can be misleading. When demand fluctuates at individual products or locations, the noise is often cancelled out at higher levels. As a result, the buffer needed at a warehouse or regional level is usually lower than the sum of buffers planned for individual stores or products.

Quantics Perspective

In large manufacturing organizations, forecasting hierarchies are not just a technical matter — they are an organizational one. Executives often ask:

‍“Why don’t the numbers from finance match those from operations?”

‍ The answer usually lies in how forecasts are structured. A demand plan built bottom-up from SKUs and a financial plan built top-down from divisions rarely align. Connecting these perspectives requires a clear hierarchical backbone linking all planning levels.

Our experience shows that coherence builds trust. When forecasts reconcile across products, regions, and time horizons, teams spend less time debating numbers and more time planning production, inventory, and capacity from a shared baseline.

In short, hierarchical forecasting ensures that planners, supply chain leaders, and executives are working from the same view of future demand.

Hierarchical Forecasting for Complex Manufacturing Environments

Quantics’ forecasting solution is built for large and complex manufacturing environments. Forecasts are generated across all relevant hierarchy levels, after which reconciliation techniques ensure consistency between them. Our platform supports best-in-class hierarchical forecasting with optimal reconciliation, which has proven effective in improving accuracy and reducing bias across aggregation levels. At the same time, other reconciliation approaches can be applied when they better fit a company’s data structure or planning process.

Selecting the right hierarchy requires both data analysis and business understanding. Together with domain experts from the customer organization, Quantics experts evaluate forecast performance across aggregation levels and test different reconciliation strategies to determine the structure that best supports production planning and supply chain decision-making.

Because forecasting needs vary across the hierarchy, our AI-driven forecasting engine does not rely on a single approach. It can apply bottom-up forecasting where operational detail matters, top-down methods where stability is key, and reconciliation techniques to ensure a consistent enterprise view.

Practical Takeaways

Start by linking forecasts to decisions, then choose the appropriate level

• Top level: budgets, revenue targets, cash flow
• Middle level: capacity planning, workforce planning, campaigns
• Detailed level: ordering, replenishment, daily execution

Decide where numbers must align

• Make consistency mandatory where teams need to agree (for example, budget vs. supply plan).
• Allow flexibility where local context matters (for example, individual stores or customers).

Be cautious when breaking totals into parts

• Using past sales shares works when customer buying patterns remain stable.
• When products, channels, or regions gain or lose importance, relying on old percentages can introduce bias.

Handle highly irregular demand carefully

• If detailed data contains many periods with no sales, forecasting directly at that level can be unreliable.
• Forecasting at a slightly higher level (for example, weekly instead of daily) and then breaking it down often works better.

Think in terms of forecasting hierarchies, not just organizational hierarchies

• Reporting structures are not always the best way to capture demand patterns.
• Re-grouping products or locations for forecasting can improve results without changing how performance is reported.

In Conclusion

Forecasting hierarchies bridge the gap between strategic alignment and operational precision.
As Kolassa and co-authors remind us, coherence is not a constraint, it’s an enabler. By understanding how data aggregates across products, regions, and time, organizations can turn fragmented forecasts into a unified planning system. The right approach depends on your data, your organizational complexity, and above all, the decisions you are trying to support.

At Quantics, we help enterprise manufacturers implement hierarchical forecasting frameworks that reconcile forecasts across levels, improving accuracy, reducing bias, and aligning planning from the production floor to the executive team.

Disclaimer

In this post, we share highlights from the book “Demand Forecasting for Executives and Professionals” by Stephan Kolassa, Bahman Rostami-Tabar, and Enno Siemsen, together with our own reflections on how these ideas apply to today’s manufacturing supply chains. This is not a replacement for the book, but rather a guide to spark thought and discussion. For a deeper dive, we encourage you to explore the book itself - available here.

‍