Right, extrapolation from the present and past where the growth has been exponential is not promising. But if somehow know the saturation level, upper asymptote, then what are doing is more like interpolation than extrapolation. Then the sigmoid curve is the solution to the differential equation I posted here a few minutes ago here at
https://news.ycombinator.com/item?id=31429475
and there is just the one free parameter k. Sometimes in practice have a shot at estimating k.