We have studied the yield of 58-nuclotide run-off transcripts obtained from in vitro SP6 RNA polymerase transcription using incomplete factorial and response surface methods. Incomplete factorial experiments were first used to estimate the relative impact of eight variables, a kind of buffer, pH of buffer, concentrations of buffer, DTT, NTP, DNA templates, and SP6 RNA polymerase on the yield of 58-nucleotides run off transcripts. Sixteen trials were performed according to a balanced and randomized design. We calculated the contrast coefficient From these coefficients we selected the significant variables. The concentration of SP6 RNA polymerase, DNA template, and NTP proved to be significantly correlated with the yield of RNA. We then optimized the yield with two designed, response surface experiments. Eighteen experiments were performed to sample the quadratic surface relating the yield to be three significant variables. Coefficients of the quadratic function with all two-factor interactions were and significant coefficients were retained. Partial differentiation of the resulting quadratic model showed it to possess an optimum. In first response surface experiments we obtained the optimal value of each variable. The optimal concentration of NTP and DNA were within the range of experimental variation but the optimal value for SP6 RNA polymerase was out of the range. So we again designed the second response surface experiments which made the optimal point as the center the new range. We obtained the new optimal values. And we optimized the concentration of $MgCl_2$. Transcription performed at the corresponding conditions yielded about 2-fold more that other conditions, confirming the predictive value of the experimentally determined response surface method.