Counterintuitively, adding noise can improve signal detection in nonlinear systems. A weak subthreshold signal alone cannot push the system over a detection threshold — but the right amount of noise assists it, causing threshold crossings that encode the signal. Too little noise: no crossings. Too much: random crossings swamp the signal. The SNR peaks at an optimal noise level — stochastic resonance.
First proposed by Benzi et al. (1981) to explain ice-age periodicity, it has since been found in sensory neurons, photoreceptors, and engineered detectors. The red dot on the SNR curve shows your current operating point.