I have 20 interview transcripts. Where do I start coding?

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

Do not start with transcript one and grind through to twenty. Start small and deliberate: a few transcripts coded carefully give you a codebook, and the codebook makes the other seventeen fast.

Step 1: read two or three transcripts without coding

Pick transcripts that differ: different roles, a long one and a short one, an enthusiastic and a reluctant participant. Read them through and write a short memo on each: what surprised you, and what keeps coming up.

Step 2: code those few line by line

Give short labels to meaningful chunks, staying close to participants' words at first. Expect 40–80 rough codes. That is normal; you will merge them later. Saldaña (2021) lists first-cycle methods such as descriptive, in vivo and process coding; pick one or two and say which in your methods chapter.

Step 3: turn your codes into a codebook

Merge duplicates. For each code that survives, write a name, a one-line definition, when to use it and when not to, and one example quote. This one table is what makes your analysis consistent and auditable.

Step 4: apply the codebook to the rest, and let it change

Code the remaining transcripts with the codebook. When something does not fit, add or split a code and note the date and reason. Recode the early transcripts at the end so all twenty use the final version.

Step 5: move from codes to themes

Group related codes into candidate themes and test each against the data: does it hold across participants, and does it answer your research question? Braun and Clarke's (2006) six phases of thematic analysis are the most widely cited route. Keep your memos; they are your audit trail for trustworthiness.