Please read the piece below which I have taken from Fortune Magazine:

"It was mid-May the last time FBR Capital market analyst Craig Berger checked with his contacts in Apple's supply chain, and what he heard then was bad news: orders for iPhones for the second quarter had just been been cut 25%.

But FBR went back to those same sources earlier this month, and on Thursday he and Robert Pikover reported "big positive revisions" in Apple's so-called build forecast.

"Our latest checks show forecasted calendar 3Q and 2008 iPhone build volumes have been revised significantly higher, with more than 15 million 3G iPhones plus two million old 2G iPhones forecast for 2008," they wrote in a note to clients.

Apple's official target is to sell 10 million iPhones in 2008."

One of the interesting things in the piece above is drastic fluctuation of the demand estimation. In such an environment, the task of the forecasters would be very difficult as the product is new, historical data does not exist that much and as we see, market preference is changing.

Do you have any idea how they do the forecasting for such a product?