Many fewer small satellite galaxies are seen around the Milky Way and M31 than the number of small dark matter halos seen in semianalytic models  and high-resolution simulations [52, 53] of such systems. But for CDM the discrepancy arises only for satellites smaller than the LMC and SMC , and such small satellites are expected to form stars very inefficiently [54, 55]. Semianalytic models taking both reionization of the universe and local feedback from supernovae into account appear to be in good agreement with observations of the relative numbers of faint and bright galaxies in environments ranging from the local group to clusters . This is encouraging, but it is important to check whether such models and high-resolution simulations  are capable of accounting for more detailed properties of low-mass galaxies, such as their radial distribution, metallicity and ages , and the "fundamental line" of dwarf galaxy properties .
"Millilensing" by DM halo substructure may be required to account for anomalous flux ratios in radio lensing of quasars by galaxies producing multiple images . A concern is whether there enough halo substructure in the inner ~ 10 kpc of galaxies, where it appears to be needed to account for such lensing, although a recent analysis suggests that substructure along the line of sight to these galaxies can help account for the observations . Spectroscopic observations can determine the mass of the lensing perturber by comparing the magnification of different regions, for example whether the tiny broad line region nearest the AGN is lensed, or also the much larger narrow line region . Lensing of AGN jets can also be a useful diagnostic for substructure . Further observations are needed.