Randomly designed siRNA targeting different positions within the same mRNA display widely differing activities. We have performed a statistical analysis of 46 siRNA, identifying various features of the 19bp duplex that correlate significantly with functionality at the 70% knockdown level and verified these results against an independent data set of 34 siRNA recently reported by others. Features that consistently correlated positively with functionality across the two data sets included an asymmetry in the stability of the duplex ends (measured as the A/U differential of the three terminal basepairs at either end of the duplex) and the motifs S1, A6, and W19. The presence of the motifs U1 or G19 was associated with lack of functionality. A selection algorithm based on these findings strongly differentiated between the two functional groups of siRNA in both data sets and proved highly effective when used to design siRNA targeting new endogenous human genes.