At this beginning stage of analysis, most experts suggest unitizing the data —identifying “units of information that will, sooner or later, serve as the basis for defining categories. Units come from interview transcripts, observation notes, or documents. A unit can be a phrase, a sentence, a paragraph. Each unit of information can be put onto a separate index card and coded according to any number of categories ranging from situational factors (who, what, when, where) to categories representing emerging themes or concepts.

Developing categories. In addition to coding units of data by obvious factors such as who, what, when, and where, analysis involves the development of conceptual categories that interpret the data for the reader. Devising categories is largely an intuitive process, but it is also systematic and informed by the study’s purpose, the investigator’s orientation and knowledge. Developing categories or themes involves looking for recurring regularities in the data. Which units of information go with each other? It is a task of comparing one unit of information with the next. Devising categories involves convergent thinking. Convergence is determining what things fit together — which pieces of data converge on a single category or theme. The categories that one constructs should be internally homogeneous. That is, all items in a single category ought to be similar. The number of categories one constructs depends on the data and the focus of the research. In any case, the number should be manageable.

There should be a minimum of unassignable data items, as well as relative freedom from ambiguity of classification. Moreover, the set of categories should seem plausible given the data from which they emerge, causing independent investigators to agree that the categories make sense in light of the data. Much of the work in category construction is a form of content analysis.

There are three guidelines by which to judge the efficacy of categories derived from content analysis. The first guideline is looking at the content of the data in developing categories. The categories should reflect the purpose of the research. Care should be taken to ensure that categories are congruent with research goals and questions. The second is that the categories should be exhaustive — that is, all relevant items in the sample of documents under study must be capable of being placed into a category. The third is that the categories should be mutually exclusive — no single unit of material should be placed in more than one category.