Carbon nanotube (CNT) network-based sensors have been often considered unsuitable for practical applications due to their unpredictable characteristics. Herein, we report the study of universal parameters which can be used to characterize CNT network-based sensors and make their response predictable. A theoretical model is proposed to explain these parameters, and sensing experiments for mercury (Hg2+) and ammonium (NH4+) ions using CNT network-based sensors were performed to confirm the validity of our model.