Does ML/AI forecasting enforce quarter-end Rallyes?

Well, first of all I think the underlying real problem here is the short-term thinking triggering these quarter-end rallyes, caused by a questionable incentivation mostly created by capital market pressure. Privately owned often face such issues to a much lesser extent. But that's not the point here. As I see it, forecasting should never be reduced to a pure mathematical exercise in search of the best accurateness. It needs to have a notion of "target setting" and something like a guidance role attached to it as well. Otherwise it is nothing better than an autonomous driving vehicle just replicating the driving faults and imperfections of the car or driver in front of it. What we really need is a prescient AI taking into account the driving behavior of more than one car ahead (=increasing the scope of input data) and anticipating the consequences of own behavior to the other drivers on the road (=increasing the logical power of the algorithm by 1-2 levels) - which will, at worst, be a traffic jam otherwise easily avoidable. Forecasting sales is very similar. The AI needs to be able to anticipate and assess its own impact on its environment before finalizing and releasing the forecast.

Dear Knut Alicke. I think the first of your mentioned challenges is very important. Improving your companies' performance with the help of AI means making data-driven decisions which are based on facts and not on emotions. It is a challenge for people to evaluate which of their tasks are easy for machines to take over so humans can focus on more sophisticated tasks. If we get this right, we improve our processes and the overall performance of our supply chain.

Knut I think you are right that pure application of AI without process and capability change will automate what exists today. A few thoughts Like the hockey stick – how far does forecasting it, get a Supply Chain; if the perfect algorithm gets you to a brilliant forecast incl. hockey stick would it not eliminate a lot of today’s noise. So as a start like it, forecast it, and then level it Have the right discussion – forecast accuracy vs demand spikes; if a perfect forecast exists could the discussion change to “..can we discuss how we get to the last week of the quarter being only 20% above the average of the quarter...” People are too used to discuss accuracy and not drivers. Forecast accuracy should be obsoleted by AI and the discussion should be on necessary and real flexibility Cut out the hockey stick – sell in vs sell out; in go to markets with channel partners sell in forecast would predict the hockey stick, a sell-out forecast would help; especially enriched with external data sets on “real” trends. Having a good sell-out forecast would help planners to challenge the internal behaviors. While this is old for industries which have Point of Sales data, it is still interesting for other businesses

Thank you very much, Knut Alicke, for raising these questions. I would fully agree, these are key challenges to be tackled. To me, an extremely important key to overcome these challenges is - leadership. Now, while the transactional leadership style answer to the topics raised may be pretty clear ("Do not allow rallies at year end." "Sanction dirty tricks."), I am convinced this will not suffice. It is the transformational leadership competence that is required in the Supply Chains: Enabling, inspiring cross-functional teams to deal with the ML/AI input. Motivate the team in the value network to find the right steering models in order to foster "E2E"-oriented, business driven and customer focused decisions based on the data gained. Build trust. Help the team to take responsibility. So, transformational leadership competence to me is a key here. Not easy at all. Yet inevitable.