You might find some research that Dean Karlan and Bram Thuysbaert have done to be especially relevant. Here's a working paper that explores that question:
They look at various approaches to selecting the poorest households (participatory rural appraisals, proxy means test, proxy means + verification), and compare the results against a more comprehensive survey on consumption. The main takeaway is that while all of those methods perform reasonably well at picking the poorest households and excluding the wealthiest households, beyond that, none of the options do especially well. My takeaway from this research is that a more parsimonious approach to targeting is likely better in terms of cost-effectiveness. It also presents a fairly strong case for randomizing within a fairly large pool of potentially eligible individuals.