Case studies in error analysis and new product forecasting


Gerhard SvolbaMy colleague Gerhard Svolba (Solutions Architect at SAS Austria) has authored his third book, Applying Data Science: Business Case Studies Using SASĀ®." While the book covers a broad range of data science topics, forecasters will be particularly interested in two lengthy case studies on "Explaining Forecast Errors and Deviations" and "Forecasting the Demand for New Products."

Explaining Forecast Errors and Deviations

This case study deals with the evaluation of forecast quality. Gerhard provides guidance on preparing results (i.e., forecasts and actuals) for analysis. He then illustrates several descriptive statistics and visualizations that go well beyond typical forecast performance reporting.

Linear regression is then used to investigate the influence of input factors on forecast error. While some of the statistical methods may be beyond the comfort zone of a typical demand planner, the book provides plenty of SAS code examples to apply the approach and generate easy to interpret results.

The final section analyzes the effect of manual overrides in forecasting. This material will be especially helpful to anyone doing Forecast Value Added (FVA) analysis of their forecasting process. It shows methods for digging more deeply into FVA results, so you can answer questions like:

  • Do manual overrides have a better effect on products with only a short sales history?
  • Are larger overrides more beneficial than smaller overrides?

Forecasting the Demand for New Products

This case study shows how forecasts can be generated for products that have no or only a short history of known demand. The study begins with discussion of the preparation and organization of data used for new product forecasting. Gerhard shows how to use Poisson Regression to model the relationship between different product features and the demand in a particular period.

The study also illustrates an alternative approach using similarity search. From all existing products that have sufficient demand history (e.g., one full year of sales), a set of reference products is defined based on similarity to the new product's features. Forecasts for the new product are based on the known demand values of the reference products. (This approach is commonly referred to as "forecasting by analogy." The book provides a rigorous illustration of how to implement it using SAS.)

Cover of Applying Data ScienceAvailable Now at SAS Books

You can purchase Applying Data Science online at SAS Books.


About Author

Mike Gilliland

Product Marketing Manager

Michael Gilliland is a longtime business forecasting practitioner and currently Product Marketing Manager for SAS Forecasting. He is on the Board of Directors of the International Institute of Forecasters, and is Associate Editor of their practitioner journal Foresight: The International Journal of Applied Forecasting. Mike is author of The Business Forecasting Deal (Wiley, 2010) and editor of the free e-book Forecasting with SAS: Special Collection (SAS Press, 2020). He is principal editor of Business Forecasting: Practical Problems and Solutions (Wiley, 2015) and Business Forecasting: The Emerging Role of Artificial Intelligence and Machine Learning (Wiley, 2021). In 2017 Mike receive the Institute of Business Forecasting's Lifetime Achievement Award. In 2021 his paper "FVA: A Reality Check on Forecasting Practices" was inducted into the Foresight Hall of Fame. Mike initiated The Business Forecasting Deal blog in 2009 to help expose the seamy underbelly of forecasting practice, and to provide practical solutions to its most vexing problems.

Leave A Reply

Back to Top