If forecasting Is the science of predicting outcomes then I would say that in the IT industry it’s a very loose science. On the one hand you have companies such as IBM would like very much for you not to forecast and rely on them, and on the other you have technology arguably vanguards putting a stake in the ground and often getting it wrong before getting it right

I have had experience in forecasting but mostly as a secondary user (I'm usually not the one collecting the data, but rather observing and interpreting the result). The reaesrch firm name that mostly pops up is Gartner. My experience has been mixed; I've been thankful for some sound bytes and with, I hope, some healthy skepticism, I feel they are a bit horse before the cart. I do think they get ahead of the game by some 3-5 years but my experience is in the mid-term they get it right. However being such a dominant force many weigh in on their assumptions following and perhaps paving the way. See Gartner giving advice on uptake of Windows Vista. http://www.theregister.co.uk/2006/12/27/2006_review_year_ahead/

Product forecasting (thanks to Wikipedia) is the science of predicting the degree of success of a new product (or in our case technology) will enjoy in the marketplace. I would estimate that the degree of success in IT forecasting is limited particularly in the long term. Many companies invest in packaged analytics such as Vanguard to review historical performance data such as sales and applies trends and patterns in the data that can be used to predict future performance. http://www.regdeveloper.co.uk/2006/10/09/visual_numerics/

I think with IT this is a little bit optimistic (the happy-path scenario) . The usage, or more specifically ease of usage equates to the value add of the software\hardware (relative to its cost) and therefore often define its success. Text messaging was a instant success because it was cheap and easy, familiar and had a broad reach. Its easier to project the extended use of SMS (to say MMS) but its harder to tell if a technology generation will be skipped and something new and more prolific will take its place.

To Gartner's credit they do quantify market segment, share and impact historically very well also. http://www.theregister.co.uk/2007/02/22/gartner_2006_sales/. I think forecasting companies stick to one industry because of the sheer volume of data required - also time in the business equates to trust, if proven correct. I was also assert that forecasting, beyond the quantitative mathematics is specific to an industry and needs qualitative real world input.

More reading on the subject:

Forecasting for Technologists and Engineers: A Practical Guide for Better Decisions (I E E Management of Technology Series)