Foresight begins the new year with our 44th issue since the journal began publishing in 2005, and in this Winter 2017 collection we’re showcasing a broad range of incisive and entertaining pieces. We’re looking at new research on the effectiveness of collaboration in forecasting and planning, fresh perspectives on our “forecasting paradigm,” a pair of reviews of a commendable new textbook on forecasting for managers, and a road map for harmonizing research and practice.
And then there’s our feature article, Recoupling the Forecasting and Stock-Control Processes by Aris Syntetos and Steve Morlidge. The problem with the application of forecasting to inventory, the authors say, is that
Too often, the forecasting process is decoupled from the stock control process: forecasters have little knowledge of how their forecasts are being used, and inventory managers have little knowledge about where the forecasts are coming from.
In consequence, critical errors are often made in the calculation of safety-stock requirements—mistakes readily corrected with better communication among forecasters, inventory controllers, and suppliers of the software in use.
Collaborative arrangements in forecasting and planning have been tried now for more than 20 years, with guidelines on the implementation of collaborative processes provided periodically by VICS (Voluntary Interindustry Commerce Solutions Association) since its initial publication in 1998. The lessons are mixed: in this issue’s Hot New Research Column, Paul Goodwin reviews recent studies on the conditions for and challenges of successful collaboration. Among his observations:
Collaboration sounds like a win-win situation and many research papers provide evidence of significant gains from it in organizational performance. But developing a collaborative scheme can require major changes in the ways that participating organizations go about their business. Success is never guaranteed, so the costs of implementing collaborative arrangements need to be carefully weighed against the expected benefits.
Foresight Associate Editor Stephan Kolassa has teamed up with Professor Enno Siemsen, Executive Director of the Erdman Center for Operations and Technology Management at the University of Wisconsin, to write the newly published volume, Demand Forecasting for Managers. The authors aim their presentation at the nonexpert, such as a manager overseeing a forecasting group, or an MBA student not specializing in analytics. You’ll find two reviews here, one by an academic and the other by a practitioner.
The 2016 Foresight Practitioner Conference was held in October in partnership with the Institute for Advanced Analytics at North Carolina State University. The theme was “Worst Practices in Forecasting: Today’s Mistakes to Tomorrow’s Breakthroughs.”
Foresight Editor for Forecasting Practice Mike Gilliland delivered the keynote address, which has now been expanded into his article Changing the Paradigm for Business Forecasting. In it, Mike questions the merit of the “offensive paradigm” in which we attempt to extract every last degree of accuracy from our forecasts, offering instead a “defensive paradigm” that acknowledges the limits of what forecasting can deliver and recognizes the foolishness of unreasonable accuracy expectations. It’s far more valuable, he argues, to devote resources to eliminating harmful forecasting practices—such as overly complex models—than to attempt to further refine already effective models and procedures.
In a Commentary on the Gilliland article, forecasting philosopher David Orrell notes that while he agrees completely with the conclusion that simple models are usually preferable to complicated models, he believes the problem is less an obsession with complexity than with building detailed mechanistic models of complexity. With the new defensive paradigm, models should be simpler, but they can also draw on a range of techniques that have developed for the analysis of complex systems.
Almost a year ago, in the Spring 2016 issue, we printed a note from Sujit Singh describing a gap between what business wants and needs forecasters to do and what researchers spend their time examining. We followed that with a feature section in the Summer 2016 issue—“Forecasting: Academia vs. Business”—that offered more than a half-dozen perspectives, from academics and practitioners, on why the gap exists and what it might take to close it. Now Robert Fildes, founder of the Lancaster Centre for Forecasting, takes a deeper look at the limited reach to date of Research to Improve Forecasting Practice. We certainly do have a gap and, in Robert’s view, to close the gap:
All three sides of the triangular relationship among academics, software suppliers, and users need to work together. Business users need to demand software changes, software people need to understand that they add value, and academics need to deliver the necessary realistic algorithms.
Software developers, argues Tom Willemain, a cofounder of SmartSoftware, face constraints that make it difficult to integrate the three sides of the triangle.
We developers have an ecology marked by intense competition and distressingly long sales cycles. Within our own councils, the bias we must confront is toward giving customers what they want instead of what’s good for them. We cannot realistically expect customers to themselves suggest promising algorithmic improvements, so we have the double burden of innovation and then documentation of benefit.
It’s not hopeless to close the academic vs. business forecasting gap, but it will need new mind-sets among all the players.