Jo -

You wrote: So, can we say that regression analysis is non-sense for such
applications!? ... So, what logic should be or can be used here, instead of
reg? <

As I wrote earlier: I think it's always important to Look At The Data and
to think about the appropriate functional form before fitting a curve. <<

You need to use your business knowledge to choose the appropriate model.
Regression is just a computational method. Instead of a "naive" model that
uses only time as the explanatory variable, you may need to include other
explanatory factors, e.g., competition, seasonality, etc. For more complex
models, regression may be very useful for sorting out the important factors.

- Mike
http://www.MikeMiddleton.com