Top left: predicted evolution of the uncertainty in , that is, the L_IR inferred from a single broadband photometric measurement, each band corresponding to a different color and line style (see legend). This uncertainty is derived by measuring the standard deviation of the difference between the true L_IR that was put in the simulated catalog and the observed monochromatic rest-frame luminosity. This is an optimal uncertainty, assuming 1) no error on the measured flux, 2) knowledge of the best average L_IR/L_ν conversion factor (i.e., the best average SED), 3) perfect subtraction of the stellar component (which only matters for 16 and 24 μm at high-redshift), and 4) no contamination from AGNs. For comparison, we show with a black solid and dashed lines the logarithmic scatter in L_IR/M_dust and L_IR/L₈ in the sample at each redshift, which are the main drivers of SED variations at λ> 30 and λ< 30 μm in our model, respectively. Top right: predicted systematic error on the L_IR of starburst galaxies (selected here with R_SB> 2) normalized to the galaxies’ offset from the main sequence. In other words, a value of x on this plot means that the L_IR will be wrong on average by a factor . Bottom left and right: same as top, but for .