ZnO nanorod field effect transistor (FET) biosensors enable ultrasensitive, rapid and label free detection of biomolecules. However, the fabrication of these nanorods without any design guideline results in poor signal to noise ratio, limiting their commercial deployment. The various forces operating within the spacing of the nanorods may prevent complete penetration of the liquid for large aspect ratios and hence mere enhancement of the geometric surface area may not yield the optimum output which indicates the intricacies involved in the design of such sensors. The present work consistently models the analyte and the device regions of the sensor by transforming the understanding of the transduction mechanism of biomolecule capture in nanowire FETs into nanorod FET structure and integrating the spacing dependent liquid penetration depth along with the various temporal fluctuations of the sensor to enhance the signal to noise ratio ( SNR ) and lower the molecule detection limit(MDL). Simulation results, verified experimentally for prostate specific antigen (PSA) shows that the optimal geometry does not correspond to the maximum surface area. Instead it is dependent on the developed surface potentials after receptor immobilization and ligand capture which vary non-monotonically with nanorod radius due to nonlinear dependence of buffer penetration for high aspect ratios. This approach has significantly increased the SNR leading to reliable quantification of PSA in human serum with MDL reduced by almost three orders of magnitude in comparison to recent reports. In summary, this work brings the ZnO nanorod sensors a step closer towards practical application for early diagnosis.