This paper presents a new concept for the decomposition of fuzzy numbers into a finite number of α-cuts. Instead of subdividing the µ axis in an equidistant way, we suggest to subdivide the x axis equidistantly leading to a more efficient decomposition of the µ axis. Considering the interpolation error as a measure for the loss of information during the decomposition, our concept leads to the minimal information loss of the decomposed fuzzy numbers.