Photometric redshift surveys are currently one of the best tools for cosmologists to probe vast swathes of the Universe with a large number density of tracers. The many opportunities they provide, however, come with a large number of data analysis challenges that the community has been successfully tackling. In this talk I will first briefly overview the some of these opportunities and challenges. Then, I will focus specifically on the problem of characterising redshift distributions and marginalising over their uncertainties. I will showcase 3 different approaches, particularly relevant in the context of near-future multiwavelength experiments covering overlapping parts of the sky.