A novel strategy for calibrating Indicator Displacement Assay (IDA)-based sensors is presented herein. The main idea is to replace the instrumental measurement responses by the equilibrium concentration of spectroscopically active species which can be obtained by the Classical Least Squares (CLS) method. Also, coupling the Indirect Hard Modelling (IHM) and CLS methods for the calibration model resulted in a reduction of matrix effects. According to Beer's law, the measured multivariate spectrum of a mixture is the sum of contributions of all spectroscopically active components via their concentrations and pure spectra. Concentrations of a few components are usually the fundamental variables in a measured spectrum in several sensors or wavelengths. In IDA systems, the equilibrium concentrations of indicator and indicator-receptor species are the fundamental variables that can be an alternative for instrumental responses as the input data for regression methods. These fundamental variables can be exploited from the recorded spectra of the mixtures when the pure spectra of the active species are known. Using dramatically reduced number of input variables without the need for any variable selection method is the main advantage of this idea over conventional calibration methods that use variable selected spectroscopic signals. This strategy can be applied for systems in which the active species are known. Accordingly, for IDA-based sensors with determined pure response to indicator and indicator-receptor complex, the free concentrations of active species can be resolved by the Classical Least Squares (CLS) method. The pure analytical responses are altered in mixtures due to the intermolecular interactions caused by matrix effect in real experimental dataset. So, the free concentrations of spectroscopically active species are not resolved correctly by the CLS method. To tackle the issue of nonlinearity of data due to the matrix effect, the Indirect Hard Modelling (IHM) can be applied to correctly resolve the fundamental variables. The applicability of the presented idea is successfully validated by simulated and real sensor array systems.