Recently, Involvation and aim10x ran a roundtable discussion on the evolution of forecasting in the retail industry. Question on the table was how retailers can leverage emerging forecasting capabilities.
Forecasting promotion volumes
Hans van der Drift, partner at Involvation, and Simon Joiner, Digital Forecasting and Demand Planning Expert at o9 Solutions led the discussion that focused on the challenges of forecasting promotion volumes in a retail environment.
Consumers expect 100% on shelf availability during promotions; a challenging requirement given the high demand uncertainty during these promotions. The panacea is to improve forecast accuracy, but there are no easy answers to this challenge.
Promotion Management Game .jpg?width=198&name=Ontwerp%20zonder%20titel%20(1).jpg)
Participants of the Roundtable first played Involvation’s Promotion Management Game, that clearly demonstrates how difficult this challenge really is. The human mind simply cannot process all the available data from past promotions to accurately forecast the next ones; planners are therefore forced to use simple heuristics and ensure supply via high stock levels.
Involvation subsequently demonstrated an advanced causal forecasting model developed by them, that was used successfully by a leading Dutch retailer for many years. Hans van der Drift ranked the model as “classic” now, because the engine is based on hard coded logic.
Machine learning: less effort, better results
The hypothesis that was discussed, was that thanks to the rapid developments in Machine Learning, this same problem can now be tackled with less effort and with similar or even better results. The roundtable discussion focused on this hypothesis and there was clear consensus that this should indeed be the case. No hard evidence like e.g., a recent case study was presented in the session though, so something to look forward to in the next roundtable!
Want to read more about Digital Planning? Read this article about Digital Planning; Hype or Trend?
