Strategy researchers are often faced with the convenience or the necessity of using existing data bases that were assembledfor purposes that differ from the research being conducted. Inferences drawn from this research can be overstated or can be simply wrong because researchers have ignored the pitfalls of translating results based on aggregated data to specific situations. Although these issues have been addressed in political science and sociology, they have not been discussed at length in the context of strategy research. This paper, there fore, explores "unit of analysis" problems specifically as they can occur in strategy research. Issues of independent and dependent variable grouping, contextual effects and examples of aggregation practices and errors typical in strategy research are presented.