The International Cancer Benchmarking Partnership is investigating cancer survival differences between six high-income nations using population-based cancer registry data. Differences in overall survival are often explained by differences in the stage at diagnosis and stage-specific survival. Comparing stage at diagnosis using cancer registry data is challenging because of different regional practices in defining stage, despite the existence of international staging classifications such as TNM. This paper describes how stage data may be reconciled for international analysis. Population-based cancer registry data were collected for 2.4 million adults diagnosed with colorectal, lung, breast (women) or ovarian cancer during 1995-2007 in Australia, Canada, Denmark, Norway, Sweden and the United Kingdom. The stage data received were coded to a variety of international systems, including the TNM classification, Dukes' for colorectal cancer, FIGO for ovarian cancer, and to national "localised, regional, distant" categorisations. To optimise comparability for analysis, a rigorous and repeatable process was defined to produce a final stage variable for each patient. An algorithm was also defined to map TNM, Dukes' and FIGO to a "localised, regional, distant" categorisation. We recommend how stage data should be recorded and processed to optimise comparability in population-based international comparisons of stage-specific cancer outcomes. The process we describe to produce comparable stage data forms a benchmark for future research. The algorithm to convert between TNM and a "localised, regional, distant" categorisation should be valuable for international studies, until global consensus is achieved to adhere to a single staging system like TNM.