Nonresponse and other sources of bias are endemic features of public opinion surveys. We elaborate a general workflow of weighting-based survey inference, and describe in detail how this can be applied to the analysis of historical and contemporary opinion polls.