Demand sensing = marketing tosh; nothing new or clever . . . I can't see evidence to the contrary.

Not to say, of course, that there is no benefit in getting an earlier feed of data on something such as demand, if - and only if - it can result in a better decision that is executed earlier than otherwise - giving an economic benefit.

The use cases do exist though. My (real) example is in the supermarkets world, specifically on the day that a new set of weekly promotions start. On this 'day one' of the promotional week regular and promotional pricing are reset, promotional 'aisle ends' are built as well as other changes in layout being implemented. For all the forecasting that is done beforehand, the actual sales by 11:00 on the first day provide a very accurate view of how all the changes - overlayed with each other - are being interpreted by the customer, resulting in sales. This early read on sales - which is net of all the inter-SKU relationships - 'could' result in a different POs being placed with suppliers a day earlier than if the daily sales were interpreted in an overnight batch process. This 'could' result in availability and sales better in line with demand, and fewer overestimates/overstocks in store which are also negative in terms store productivity. The challenge in this example is not so much 'the software' but rather the data flow and the timing of the execution of the calculation (e.g. purchase order calculation), in line with other time windows that need to be hit in order to get a PO out a day earlier. In this sense there is no difference between 'real time' and 'timely' - e.g. a batch process is perfectly OK as long as it can be executed quickly and when you need it.

Acknowledgement from that the probabilistic approach is superior, or at least will offer more resilient decisions. How long to get to market in a core product that actually works? 10 years?

I am looking forward to some good discussions here.