Executive Guide to Forecasting in Manufacturing, Part 2: Mastering the Workflow

Many executives describe forecasting as a “black box.” Data flows in, models produce numbers, and reports are circulated - but too often the process feels disconnected from the business decisions it is meant to support. Without structure, forecasts risk being produced at the wrong level of detail, at the wrong frequency, or for the wrong purpose.

In Demand Forecasting for Executives and Professionals by Stephan Kolassa, Bahman Rostami-Tabar, and Enno Siemsen, the authors emphasize that forecasting should not begin with data or algorithms, but with the decisions that require a forecast. From that starting point, a structured workflow ensures that forecasts are relevant, transparent, and decision-ready. For enterprise manufacturers, adopting such a workflow is critical to move from reactive forecasting toward reliable, value-driving planning.

Key Insights from the Book

The authors outline forecasting as an iterative workflow, not a one-off calculation. Each stage builds on the previous one to link forecasts directly to business decisions:

1. Identify the decision. Forecasts have no value on their own - they matter because they support strategic, tactical, or operational decisions.
2. Define requirements. Establish forecast horizon, frequency, and level of detail. A network design decision might need a 5-year market forecast, while replenishment requires daily SKU-level projections.
3. Gather data. Use historical demand, predictor variables (deterministic like promotions or holidays, stochastic like weather), and domain knowledge from experts.
4. Prepare data. Clean and structure the dataset by fixing errors, handling missing values, and ensuring consistency.
5. Visualize data. Plot the data to detect trends, seasonality, or anomalies that could impact model choice and interpretation.
6. Choose and train models. Select the right methods for the decision at hand, test multiple models, and balance accuracy with interpretability.
7. Produce forecasts. Generate outputs for the required horizon - as point forecasts, prediction intervals, or full probability distributions. Causal models can also incorporate predictors with known or forecasted values.
8. Evaluate quality. Measure not only forecast accuracy but also relevance, speed, and usability for decision-makers.
9. Communicate results. Go beyond single numbers - share uncertainty through ranges, distributions, or scenarios.
10. Incorporate judgment. Adjust forecasts when new information emerges, but document changes to learn which inputs add value.

The book makes one point especially clear: discipline in following the workflow matters more than complexity of methods. Without this structure, forecasts risk being misaligned with the decisions they are supposed to guide.

Quantics Perspective

Forecasting is an iterative process, not a one-off exercise. From our perspective at Quantics, two elements are particularly important: getting the initial forecasting setup right and establishing a structured process for continuous improvement.

1. Get the forecasting setup right

The starting point should not be the forecasting model, but the business decision the forecast is intended to support. This determines the key requirements: What forecast horizon is needed? At what level of aggregation? What time intervals and units of measure should be used? How frequently does the forecast need to be updated?

This alignment is particularly important in manufacturing. Demand forecasts provide a critical input for a wide range of decisions and stakeholders – from production and inventory planning to procurement, logistics, and finance. A good forecasting setup therefore needs to reflect the requirements of the relevant stakeholders while providing a common basis for planning across the organization.

The next step is to systematically assess the available data. What historical data is available? Which additional internal or external drivers could add value? How should the data be cleaned, structured, and visualized? Only then should forecasting methods be selected, trained, and tailored to the specific business problem.

Evaluating forecast quality should also go beyond a single metric. A model might achieve a lower WMAPE while still exhibiting a strong and persistent negative bias. Bias and relevant accuracy metrics such as WMAPE should therefore be assessed together – across different forecast horizons, time intervals, and levels of aggregation.

A meaningful benchmark is equally important. More advanced methods should demonstrate the Forecast Value Added they deliver compared with a simple baseline. Ideally, this should not be evaluated using a single test period, but across multiple rolling forecasts. This helps determine whether an approach can consistently outperform relevant benchmarks and deliver robust results even for difficult-to-forecast demand patterns – an especially important requirement in B2B manufacturing environments.

2. Establish a process for continuous improvement

Getting the initial setup right is only the beginning. Forecasting needs to become a continuous, collaborative process in which new information is incorporated, assumptions are challenged, and results are continuously improved.

Ideally, the system first generates a statistical baseline forecast. Planning teams can then enrich it with relevant information and business context, assess risks, and evaluate different scenarios. Manual adjustments should be transparently documented and their impact measured. This makes it possible to understand which process steps and additional inputs actually add value to the forecast.

It is equally important to review forecasts and their implications at different levels – from individual Demand Forecasting Units (DFUs) to plants and product groups, all the way up to the company level. Effective visualizations and reporting make it easier to understand relationships and uncertainty and translate them into better planning decisions.

For large manufacturing companies with distributed teams, forecasting is therefore also a matter of process coordination and governance. Forecast updates need to be coordinated, additional information incorporated in a structured way, and responsibilities clearly defined.

Forecasting as a continuous planning process

Quantics is specifically designed for the requirements of large manufacturing companies operating partly or entirely in B2B environments. The solution supports not only forecast generation but the broader forecasting process: coordinating forecast updates across distributed teams, incorporating additional information and scenarios, measuring the value added of individual process steps, and providing transparency across different planning levels.

The result is a continuous improvement process that promotes strong process discipline while still allowing relevant expert knowledge to be incorporated. Ultimately, the goal is not simply to generate the most accurate forecast possible, but to establish a forecasting process that consistently enables better planning decisions.

Practical Takeaways

Executives can strengthen forecasting effectiveness by:

• Starting with decisions. Define what decision the forecast will inform before choosing models or data.
• Tailoring to the decision level. Match horizon and granularity to strategic, tactical, or operational needs.
• Integrating diverse inputs. Blend historical data with predictors and expert judgment.
• Communicating uncertainty. Present ranges and scenarios, not just single numbers.
• Documenting adjustments. Track judgmental overrides to see what truly adds value.
• Aligning functions in one workflow. Ensure sales, operations, and finance work from a shared process.
• Using dedicated software. Platforms like Quantics enforce workflow discipline, automate manual steps, and scale demand planning across complex organizations.

In Conclusion

Forecasts that lack a clear workflow are just numbers without context. By embedding a structured process, executives ensure that forecasts are directly connected to decisions, transparent across teams, and continuously improved. For manufacturers, this discipline reduces politics, builds trust, and strengthens resilience across the supply chain.

At Quantics, we see that organizations with structured forecasting workflows not only improve forecast accuracy but also align teams faster, strengthen cross-functional collaboration, and make more reliable planning decisions.

In our next post, we will explore how to manage uncertainty and make smarter decisions under risk - turning forecasts into powerful tools for balancing costs, service levels, and resilience.

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.

FAQ

1. Do all forecasts need the same level of detail?

No, the right level of detail depends entirely on the decision the forecast supports. Forcing every forecast to the same granularity wastes effort where detail isn't needed and produces unreliable numbers where it is. Match horizon and detail to the decision, not the other way around.

2. What's the difference between a strategic, tactical, and operational forecast?

Strategic forecasts support long-term decisions like network design or capacity investment, typically spanning years at an aggregated level. Tactical forecasts inform medium-term choices like workforce and capacity planning, usually months ahead at a product-group level. Operational forecasts drive daily or weekly decisions like replenishment and scheduling, at the most granular SKU or location level. Each needs a different horizon, frequency, and data setup.

3. Who should own the forecasting process: sales, supply chain, or finance?

No single function should own it alone; forecasting works best as a shared process with clear accountability, often coordinated by supply chain or demand planning. Sales contributes market and customer knowledge, finance brings budget and revenue context, and operations adds capacity constraints. Without a shared workflow, each function tends to build its own number, which is exactly the misalignment structured forecasting is meant to prevent.

4. Is it possible to run a structured forecasting workflow in Excel, or do you need dedicated software?

Yes, Excel can support a structured workflow for a small number of products or a single business unit. It becomes impractical once a company needs to forecast across many SKUs, locations, or business units, test multiple methods, and reconcile results across levels. At that scale, dedicated forecasting platforms enforce the same workflow discipline automatically, without the manual rebuild Excel requires each cycle.