On the limits of forecasting
Some comments on a post on Linkedin
Quite a lot of comments were received for a post on Linkedin on the Limits of Forecasting and it is rather interesting what turned out.
There is always quite a lot of confusion about what does “forecasting” mean. It is not about a magic crystal ball which predicts the unpredictable. It is “simply” something which helps you predict what can be predicted: trends, seasonalities, correlations. And this can be made very very well thanks to Neural networks, AI and mathematical formulae with greek letters (citing that post) – these tools can do much much better than the classical rules of thumb used by many managers. Forecasting has to do with extracting knowledge from the data. It cannot help with unpredictable events. It “simply” helps you in reminding that in summer you will sell more ice creams than in winter. Apparently easy, predictable, tips which are however very difficult to find by yourself in a dynamic, complex, competitive situation. Unless you can support your managerial intuition with analytics and with a powerful forecasting tool.
In that article the use of forecasting tool is seen as a merely academic exercise to gain 5% accuracy at most. We are well aware that modern, data-driven, forecasting techniques can deliver significantly higher increases in precision.
However, even taking this 5% estimate as accurate, we do not live in a wonderful word in which this makes no difference – we are unfortunately in a different world, in which 5% more accuracy might mean quite a lot in terms of lost customers or in terms of inventories and, in the end, might make the difference between being a successful company or being ready to shut down. Data-driven forecasting is a need, not a luxury!