Actual orders "consume" the forecast, meaning that the forecasted demand is adjusted to avoid double-counting. This is typically called forecast consumption.
For example, if a forecast predicts 100 units of a product for a month and 70 units have already been ordered, the forecast for the remainder of the month will be reduced to 30 units.
However, if a sudden spike in demand occurs during the month, it can complicate matters for the next months as well because perhaps this is an indication of increased overall demand rather than just forecast consumption. This is where demand sensing comes in.
(Forecast consumption is explained well by Frepple in this article.)
Using real-time data like sales orders and POS data allows demand sensing to fine-tune the short-term forecast.
Sales orders will give you immediate feedback on the accuracy of your forecast. As orders come in, they offer a real-time view of what’s happening in the market. If you notice a steady increase in orders from certain regions or for specific products, it’s a clear signal that you may need to get more inventory into those areas to meet rising demand.
Incorporating sales orders as a demand signal is especially valuable in fast-moving industries like consumer goods, where shifts in buying patterns can occur quickly. When you combine this real-time data with other external sources, like POS data from stores, you can see where demand/trends are increased compared to what the original forecast indicated.
Some companies include using external signals like indexes as part of demand sensing, but if it isn't focused on the shorter term (which such signals usually aren't doing), then it isn't demand sensing.
Demand sensing according to Logility for the shorter term.
Demand sensing according to O9