Star Formation Rate (SFR) inferences are based in the so-called constant SFR approximation, where synthesis models are require to provide a calibration; we aims to study the key points of such approximation to produce accurate SFR inferences. We use the intrinsic algebra used in synthesis models, and we explore how SFR can be inferred from the integrated light without any assumption about the underling Star Formation history (SFH). We show that the constant SFR approximation is actually a simplified expression of more deeper characteristics of synthesis models: It is a characterization of the evolution of single stellar populations (SSPs), acting the SSPs as sensitivity curve over different measures of the SFH can be obtained. As results, we find that (1) the best age to calibrate SFR indices is the age of the observed system (i.e. about 13 Gyr for z = 0 systems); (2) constant SFR and steady-state luminosities are not requirements to calibrate the SFR ; (3) it is not possible to define a SFR single time scale over which the recent SFH is averaged, and we suggest to use typical SFR indices (ionizing flux, UV fluxes) together with no typical ones (optical/IR fluxes) to correct the SFR from the contribution of the old component of the SFH, we show how to use galaxy colors to quote age ranges where the recent component of the SFH is stronger/softer than the older component. Particular values of SFR calibrations are (almost) not affect by this work, but the meaning of what is obtained by SFR inferences does. In our framework, results as the correlation of SFR time scales with galaxy colors, or the sensitivity of different SFR indices to sort and long scale variations in the SFH, fit naturally. In addition, the present framework provides a theoretical guideline to optimize the available information from data/numerical experiments to improve the accuracy of SFR inferences. More info en Cerviño, Bongiovanni & Hidalgo A&A 588, 108C (2016)