Forty hours of interview transcripts. A cursor blinking in an empty spreadsheet. No idea where to actually start coding.
This moment catches a lot of DBA and MBA students by surprise, because reading the transcripts feels like progress — you’ve done the interviews, you’ve read them closely, surely the hard part is behind you? It isn’t, and I want to be direct about why: reading is not coding, and the gap between “I’ve read this closely” and “I have a coding scheme I actually trust” is exactly where this overwhelm lives.
Why the Standard Advice Doesn’t Quite Work
The common advice is to start coding transcript one, line by line, and let a structure emerge as you go. This works for some people, some of the time. More often, it produces a specific, predictable problem: by transcript ten, you have somewhere between eighty and a hundred fine-grained, inconsistent codes, several near-duplicates with slightly different names, and no clear sense of which ones actually matter to your research question. Reorganising that mess after the fact takes longer than doing an initial structural pass would have.
The core issue is that formal, systematic coding and initial theme discovery are two different tasks, and doing them simultaneously, from transcript one, tends to do both badly.
A Two-Pass Method
Pass one: scouting, not final analysis. Read through four or five transcripts — ideally ones that seem to represent some variety in your sample — and instead of formally coding, simply note candidate themes in the margin, in your own words, without worrying about consistency or completeness. The goal here isn’t a coding scheme. It’s a rough map of what’s actually in your data, built from a manageable subset rather than the whole set.
Consolidate before continuing. Take your scouting notes from those four or five transcripts and group them into a smaller number of provisional themes — typically somewhere between five and ten, not eighty. This becomes your first working coding scheme. It will be imperfect and it should be treated as provisional, not final.
Pass two: code against the scheme, and let it evolve deliberately. Now work through your full data set using that scheme, but track when a new transcript surfaces something the scheme doesn’t capture. When that happens, make a conscious decision — does the scheme need a new theme, or does an existing theme need to be split or renamed — rather than silently adding another one-off code.
This is slower at the very start than diving straight into full coding. It’s considerably faster overall, because you’re not redoing the coding of your first ten transcripts once you realise, at transcript twenty, that your original scheme doesn’t actually hold.
Why This Matters More Under Time Pressure
Students often skip the scouting pass specifically because they feel behind and want to see “real” analysis progress immediately. This is understandable, and it’s usually counterproductive — the recoding that follows an ungrounded coding scheme typically costs more time than the scouting pass would have taken. Slowing down at the very start of analysis is one of the few places in a dissertation where doing so genuinely speeds up the overall timeline.
FAQ
How many transcripts should the scouting pass cover? Enough to capture likely variation in your sample — often four to six, or roughly 15–20% of your total, whichever is more practical for your sample size.
Do I need special software for this? No. The scouting pass works fine with margin notes on printed transcripts or a simple document. Software like NVivo or Atlas.ti becomes more useful once you’re doing the formal, full-dataset coding pass (and even then, they are not always essential).
What if my themes keep changing throughout the whole analysis? Some evolution is normal and expected, particularly early on. If your scheme is still changing substantially by the final quarter of your data, that’s worth a direct conversation about whether the scheme needs a more fundamental rethink.
Next Step
If you’re staring at a pile of qualitative data and don’t know where to begin, that first structural decision is worth getting right early — it’s the decision that determines how much rework you’ll be doing later. I offer a free 15-minute introductory call.