Thanks for the response, Khebab -- this is basically what I was asking. Do points prior to 50 Gb ever get used in the estimation. I guess I am just curious to how long you have to wait to choose a "linear" portion of the profile. Also, how effective is this if the peak hasn't occurred yet?

For instance, what if you started estimating with the first two observations and then updating your estimates as new observations came in. You would get wildly varing answers. Then, eventually, you would have to decide where to quit using early points to capture the linear part of the profile. In the graph above, if the estimation was done using the points between 40 Gb and 70 Gb, we would have estimated the cumulative Gb to be around 85.

In the end, I think this is a decent way to model existing data, but may be poor in making predictions of any accuracy. I just think we have to be careful on how this is presented -- especially when addressing scientifically-minded non-believers.