Would it work if you took out the absolute references ($)? =$D9 + (A10-$A9) * ($D9-$D8) / ($A9-$A8). Where all of the extrapolated points are collinear with the last two entered values, linear extrapolation based on any two of the points (including the previous two in the series), should yield the same result. Then when someone copies in a larger table than original, the formula won't need to be adjusted. The tricky situation is how to deal with the case when someone copies in a smaller table than given. I don't know exactly how I would handle that in your situation -- maybe build a "template spreadsheet" with the smallest possible entered table (two rows). Then train users to always start with that template when entering their own data.