Here is the first of what I hope to be many guest postings from my colleagues here at SAS. Today Snurre Jensen, Business Advisor from SAS Denmark, writes about his recent encounter with a blog about dealing with demand changes in SAP APO. From Snurre:
In my ongoing quest for knowledge of how forecasting is applied in different companies and how different forecasting systems support this work I recently stumbled upon a blog that describes an approach in SAP APO to deal with the effects of the current and expected future decrease in demand that a lot of companies face.
After making sure it wasn’t April 1st I realized that this is what a lot of people working with forecasting on a daily basis have to endure. And this is just the tip of the iceberg. In companies using SAP APO and similar systems to “support” their forecasting process the numbers entered for the effects of promotions and events in general are created, maintained and updated in a separate environment, typically spreadsheets. Quite quickly the forecast process involves a nightmarish sub process where people have to agree on which spreadsheet is the “right” one, agree on an effect of the future promotion/event, update the spreadsheet ending with the pasting of the final value into the forecasting system. As an outsider it’s difficult to see the value in this approach.
The blog also indicates that the statistical forecast have a difficult time picking up changes in the underlying trend in sales. But what kind of support does the statistical forecast engine in these forecast systems then give you? Looking at this project at the Lancaster University Centre for Forecasting, it appears that the answer is “not much”. As Mike Gilliland’s earlier post indicates, the paradox seems to be that actual support for a good forecasting process might not be the primary influencing factor when buying a forecast system.
Luckily there is light at the end of the tunnel. Forecast support systems that automatically take into account trend and seasonal effects, automatically detect outliers, automatically estimates effects of different events (including moving calendar events like Easter), and finally automatically estimates effects of promotions do exist. I’ll leave it to you make an educated guess at a company that delivers a system like this…
You can meet Snurre at the A2009 Analytics Conference in Copenhagen, July 1-2.
For those of you with data quality issues (and that probably includes everyone), SAS is hosting a free webcast today (June 17, 1pm EDT): Got Bad Data?
This is the second in a series of live webcasts focused specifically on this problem, and will be delivered by my colleagues Guarav Verma and Anne Milley. Topics include data quality profiling, continuous improvement, data collection with closed-loop monitoring, and supporting analytical applications. The webcast will be recorded for later viewing, and Part 1 of the series is available for free replay.
Finally, I would like to thank Constance Korol, Senior Marketing Manager at the Institute of Business Forecasting & Planning, for inviting me to contribute to IBF’s blog site a couple of weeks ago. Constance has been able to attract a stellar array of guest-bloggers such as Tom Wallace, Chaman Jain and Larry Lapide...as well as a few like me.