We report here on iron determination in sea matrix by electrothermal atomic absorption spectrometry. We optimised the experimental conditions of the spectroscopic factors, i.e. entrance monochromator slit height, slit width and lamp current, and of the electrothermal factors, i.e. calcination temperature, calcination duration, atomisation temperature and injected volume, since these two groups are independent. Because of the technical constraints occurring on our experimental domain of study we built a D-optimal experimental design to conduct these optimisations. We monitored three experimental responses, i.e. the standard devlation of the instantaneous specific absorbance, the specific integrated absorbance and the non-specific integrated absorbance. The three of them clearly showed an optimum in three opposed experimental conditions. Consequently, a desirability transformation of the experimental responses permitted us to obtain a better compromise. We, finally, computed a new experimental response which took the raw measurements into account to directly estimate the limits of detection. The optimisation of this computed response showed us that the lowest detection limit was obtained under the same experimental compromise that the one suggested by the desirability study. This result allowed us to validate the computed response as an interesting optimisation criterion in electrothermal atomic absorption spectrometry.We report here on iron determination in sea matrix by electrothermal atomic absorption spectrometry. We optimised the experimental conditions of the spectroscopic factors, i.e. entrance monochromator slit height, slit width and lamp current, and of the electrothermal factors, i.e. calcination temperature, calcination duration, atomisation temperature and injected volume, since these two groups are independent. Because of the technical constraints occurring on our experimental domain of study we built a D-optimal experimental design to conduct these optimisations. We monitored three experimental responses, i.e. the standard deviation of the instantaneous specific absorbance, the specific integrated absorbance and the non-specific integrated absorbance. The three of them clearly showed an optimum in three opposed experimental conditions. Consequently, a desirability transformation of the experimental responses permitted us to obtain a better compromise. We, finally, computed a new experimental response which took the raw measurements into account to directly estimate the limits of detection. The optimisation of this computed response showed us that the lowest detection limit was obtained under the same experimental compromise than the one suggested by the desirability study. This result allowed us to validate the computed response as an interesting optimisation criterion in electrothermal atomic absorption spectrometry.