Background A sequential statistical strategy was used to optimize tannase production from Aspergillus tubingensis using tea stalks by solid-state fermentation. Results First, using a Plackett-Burman design, inoculum size and incubation time (among seven tested variables) were identified as the most significant factors for tannase yield. The effects of significant variables were further evaluated through a single steepest ascent experiment and central composite design with response surface analysis. Under optimal conditions, the experimental value of 84.24 units per gram of dry substrate (U/gds) closely matched the predicted value of 87.26 U/gds. Conclusions The result of the statistical approach was 2.09 times higher than the basal medium (40.22 U/gds). The results were fitted onto a second-order polynomial model with a correlation coefficient (R²) of 0.9340, which implied an adequate credibility of the model.