aIn the biomedical field, end-point detection bioassays such as enzyme-linked immunosorbent assays (ELISAs) are essential tools because of their simplicity, high-throughput, and suitability for their use at the point-of-care. End-point bioassays are significantly constrained by the need of sample labeling with fluorescent or colorimetric tags for subsequent detection. A promising strategy to overcome these limitations is to harness recent advances in label-free biological nanosensors. Here we analyse the potential of nanomechanical biosensors based on surface stress for the label-free end-point detection of horseradish peroxidase. We address the variability of the sensor responsethrough the analysis of 1012 cantilevers with different antibody surface densities, two blocking strategies based on polyethylene-glycol (PEG) and bovine serum albumin (BSA) and stringent controls. The study reveals that the performance of the assay critically depends on both antibody surface density and blocking strategies. We find that the optimal conditions involve antibody surface densities near but below saturation and blocking with PEG. We find that the surface stress induced by the antibody-antigen binding is significantly correlated to the surface stress generated during the antibody attachment and blocking steps. The statistical correlation is harnessed to identify immobilization failure or success, and thus enhancing the specificity and sensitivity of the assay. This procedure enables achieving a rate of true positives and true negatives of 90% and 91% respectively. The detection limit is of 10 ng/mL (250 pM) that is similar to detection limit obtained in our ELISA