What kinds of products can this demand forecasting workflow handle?

The workflow can be used for any manufactured products with historical demand data, such as components, finished goods, or spare parts, provided that the data can be read into KNIME.

Can I use external drivers in my demand forecasts?

Yes. You can include additional variables such as promotions, prices, macroeconomic indicators, or customer segments in your models by joining them to the historical demand data.

How do I choose the right forecasting model?

You can train multiple candidate models in KNIME and compare them using error metrics on holdout periods or cross-validation, then select the one that best fits your business requirements.

Can forecasts be generated automatically every period?

Yes. By deploying the workflow on KNIME Business Hub, you can schedule it to run regularly and automatically update forecasts for each cycle.

How can I share forecast results with planners?

You can export forecast outputs to spreadsheets, databases, or data apps, or integrate them into existing planning tools to make them easily accessible to planners and stakeholders.

Manufacturing Demand Forecasting

How This Workflow Works

This workflow forecasts weekly demand for key automotive products by analyzing historical manufacturing data. It enables users to select a product, explores time series patterns, builds and evaluates SARIMA forecasting models, and visualizes both the forecasts and their accuracy.

Key Features:

Allow users to select and predict demand for specific products
Forecast future product demand using advanced time series models (i.e., SARIMA)
Extend forecasts beyond available data to support long-term planning
Visualize trends, seasonality, and forecast accuracy

Step-by-step:

1. Select and Explore Product Demand:

Users choose a product of interest, and the workflow aggregates production data to a weekly level. It then explores the time series, highlighting trends and seasonal patterns that may affect demand.

2. Analyze Seasonality and Stationarity:

The workflow examines the selected product’s demand for repeating seasonal patterns and tests whether the data is suitable for forecasting. This includes visualizing autocorrelation plots and running statistical tests (e.g., ADF test) to check if a time series is stationary or non-stationary.

3. Train and Apply Forecasting Models:

The historical data is split into training and testing periods. The workflow trains SARIMA models on past data, generates forecasts for the test period, and then extends predictions into future years to support proactive production planning.

4. Visualize and Score Forecasts:

Forecast results are combined and compared to actual demand using visual plots and quantitative metrics. This step helps users assess forecast accuracy and understand how well the model captures real-world demand patterns.