One thing that struck me as very true when reading through the history of Ericsson website was that when making predictions of the future, it is not only one parameter that changes but many simultaneously. This is what makes predicting things so difficult. A prime example is the forecast of GSM subscribers in the mid-80s and even throughout the 90s which were always wrong in orders of magnitude. And an additional difficulty that comes to mind: Knowing which parameters have an influence in the first place. Even if you make assumption about several parameters you could still miss important one for the forecast.