The inadequate classification of retiree sub-groups ultimately results in misaligned policy. To generate sets of sub-groups that may be appropriately targeted for policy and interventions, variables are used that reflect the social structure of retirees, particularly the options of partial and complete retirement, marital status, gender, as well as the retirement status of the spouse, where relevant, and disability. Three sets of longitudinal Australian data were combined, each reflecting a four-year period (2003–2007, 2007–2011, 2011–2015) during which the individuals aged 45 to 69 retired (n = 1179). A multiway frequency analysis was performed to develop an inductive, combinatorial model of retirement from work. The resulting parsimonious taxonomy of sub-groups of the newly retired reflected main effects and interactions of key social-structural variables. Notably, a key driver of the pattern of results was that couples tend to coordinate their retirement behavior in both the decision to retire and form of retirement. Non-partnered retirees were more likely to be women. Disability was also a driver of retirement for non-partnered retirees, regardless of gender. Identifying sub-groups based on combinations of retiree characteristics can better inform policy design, appropriate health promotion interventions and potential specific triggers for enacting those policies. Overall, marital status, spousal retirement behavior and disability may each present a more useful basis for a taxonomy of retirement than more individually oriented age- and wealth-based systems.