A direct comparison between hydrodynamical simulations and observations is needed to improve the physics included in the former and to test biases in the latter. Post-processing radiative transfer and synthetic observations are now the standard way to do this. We report on the first application of the SKIRT radiative-transfer code to simulations of a star-forming cloud. The synthetic observations are then analyzed following traditional observational workflows. We find that in the early stages of the simulation, stellar radiation is inefficient in heating dust to the temperatures that are observed in Galactic clouds, thus the addition of an interstellar radiation field is necessary. The spectral energy distribution of the cloud settles rather quickly after ∼3 Myr of evolution from the onset of star formation, but its morphology continues to evolve for ∼8 Myr due to the expansion of H ii regions and the respective creation of cavities, filaments, and ridges. Modeling synthetic Herschel fluxes with one- or two-component modified blackbodies underestimates total dust masses by a factor of ∼2. However, spatially resolved fitting recovers up to about 70% of the intrinsic value. This “missing mass” is located in a very cold dust component with temperatures below 10 K, which does not contribute appreciably to the far-infrared flux. This effect could bias real observations if this dust exists in large amounts. Finally, we tested observational calibrations of the SFR based on infrared fluxes and concluded that they are in agreement when compared to the intrinsic SFR of the simulation averaged over ∼100 Myr.