Qualitative coding

Qualitative coding is the foundational technique in qualitative research for converting unstructured language into structured data. The analyst (human or, increasingly, model) reads each unit of text — a sentence, a paragraph, a response — and assigns one or more codes drawn from a code book.

Codes can be inductive (emerging from the data, refined across passes) or deductive (drawn from a pre-defined framework). The Frustration Condition Tracker uses a hybrid: deductive coding against the five Frustration Architectures, with an inductive layer beneath each Architecture for the specific sub-pattern the team is naming.

Quality of coding is judged on three axes: consistency (the same input gets the same code on a re-read), coverage (every unit gets at least one code or an explicit "unclassified"), and discriminability (codes are distinct enough that an item rarely belongs to two of them).