Cluster analysis has been applied to quantitative data in educational research over several decades and has been a feature of the Maurice Galton's research in primary and secondary classrooms. It has offered potentially useful insights for teaching yet its implications for practice are rarely implemented. It has been subject also to negative criticism. Acknowledging cluster analysis to be an exploratory technique, this paper outlines different approaches to cluster analysis and reports examples of cluster analysis applied in a range of published studies. It goes on to pose five research questions about the process of achieving valid clusters which provide meaningful and practicable implications for teaching. These questions are addressed with reference to the analysis of a major quasi-experimental study of science pedagogy in Pakistan. The paper exposes both the perils of misapplied cluster analysis, and the practical value of pursuing a rigorous application process. (Contains 4 tables and 1 figure.)