Survey response rates are important process quality indicators and are used for many purposes. However, attempts to compare response rates - across surveys, years, organisations and countries - are severely hampered by inconsistencies in the use of survey outcome categories and in the calculation of response rates based upon these categories. In this article, we highlight some of the main issues. With regards to outcome categories, these include the structure of the coding schema, the definitions of the categories and field implementation. With regard to response rate calculation, main issues include the suitability of different rates for different purposes, how to treat uncertainty regarding eligibility of sample units and whether or not to weight the data. We illustrate some of these issues using data from UK surveys and we suggest the possible form of some solutions.