Several recent articles seeking to validate the Goldthorpe class schema have implicitly raised some interesting questions about the relation between analytical techniques and theory. The favoured technique of latent class analysis actually brings this variety of class analysis close to the much older tradition of ideal-type analysis. In doing so, it helps to illustrate a much wider problem that arises in any conceptualization in terms of classes. The paper seeks to show how, just as in popular discourse around 'class', stereotypes are used to simplify and summarize a complex set of related characteristics, so similarly ideal types, as used in class analysis, are an over-simplification. As elsewhere in social science, the danger of using them is that they serve to protect a theoretical concept from empirical refutation. An alternative 'research programme' to that of class analysis is presented, based on measures of stratification arrangements that are more closely linked to the reproducing structure of inequalities.