How to code qualitative data

By Dr. Rafiq Muhammad, MD, PhD · Updated June 2026

Coding means tagging segments of your data — a sentence, a passage — with short labels that capture their meaning in relation to your question. Those codes let you find and compare related material across everything you collected, and they’re the raw material your themes or categories are built from. Done consistently, coding is what makes a qualitative analysis defensible.

Inductive vs deductive coding

• Inductive (data-driven): codes emerge from the data as you read. Best when exploring something new.
• Deductive (concept-driven): you start from a codebook built from theory or prior work. Best when testing or extending an existing framework.
• Blended: most real studies start with a few a priori codes and add inductive ones as they appear — just say so.

Two cycles

Coding usually happens in passes. First-cycle coding stays close to the data (descriptive, in-vivo using participants’ own words, process codes). Second-cycle coding groups and abstracts those into categories and patterns. Don’t expect your first codes to be your final ones — they evolve.

Build a codebook

A codebook keeps coding consistent — within your own work over weeks, and across coders. Each code gets:

Software (CAQDAS)

Tools like NVivo, ATLAS.ti, MAXQDA, Dedoose, and the free, open-source QualCoder organise codes and retrieval — but they don’t interpret for you. The analytic thinking stays yours; the software just keeps it tidy and auditable.