In analyzing Snap Spectacles, I discussed the value of early information, and argued how collecting early information is helpful. Production lead times for hardware products are long, and hence firms have to make huge commitments based on forecasts even before the demand starts trickling even as the product is being manufactured.
Ideally, a firm would like its total production volume to match the final demand exactly. Being able to revise this production commitment can be a significant advantage. Even a small amount of initial demand can help correct orders or validate decisions with high confidence.
Some recent data about VA 2017 elections from NY Times is useful in thinking about this problem in a different context.
Below is a snapshot of the voting prediction (a bit after 10 pm). The prediction began with 50-50 probability for blue and red teams when the votes started trickling in. With only 20% of votes in at 745pm, Probability (blue wins) has been revised to almost 95% confidence.
Of course, I am simplifying quite a bit here. For example, where the votes came in from can improve the confidence in prediction in different degrees. Similarly, where or from whom the demand information is coming from can affect the confidence in the production release decision.
However, the main idea is that one needs just a bit of (demand) information to validate or reverse one’s (product) decision with high confidence.