Thank you, Gianfranco for your promt reply!
I was afraid to start my practice from so wide question as a culture. I already tried to explain the concept to several people in my organization and I was amazingly puzzled when my Development director said - "That is what we are doing today", meaning "counting" man hours for a set of functions to be delivered. Not saying that the method of counting is far from any method.
My first intention of FPA is to get abstract figures of new functions to be injected to current system and then I convert them to cost assotiated. I do not really want to use FP to count the project effort (probably later). The problem I face - we deal with a multi platform and multi application environment where one project is usually delivered by many different teams. Even we collect historical project data it becomes useless because Project Manager never knows what teams will deliver their project. What is more, a project can be delivered having different technical solutions. And I want to abstract from this fact. So Business wants to see cost of the coming deal first of all. Gradual estimates like Small, Medium, Big does not wrok here...
Another problem, even today every deal I have been working with has different system components involved and historical based effort method takes lots of time to compose final (and still high-level) estimate. Once people accustomed to FP I can apply the practice to other project related tasks and teams. And so on. That is my roadmap. So, my goal at the beginning is to provide abstract (from team, geographic, platform, experience) figures to assess the cost.
Apart of all intentions, I believe there is some information of implementation FPA practices in different organizations. What is "good" project to start with? Steps, maybe prerequisites? I do not believe it is more difficult than current outdated methods we use. I just need to find right way to start.
Or you want to say the old empirical counting is still more accurate and faster?