Reading 30 interviews in an afternoon

Thirty interviews. One afternoon. No, really.

The conventional wisdom says synthesis takes as long as the fieldwork did, maybe longer. That's often true — but it's true because most people synthesize badly, not because synthesis is inherently slow. With a workflow that respects how your brain actually processes qualitative data, a round of 30 interviews can go from raw transcripts to an exec-ready summary in about four hours.

Here's the workflow. It's the one we've converged on after making every mistake in the list at the bottom of this post.

Step 1 — Read the first five without taking notes

This is the step everyone wants to skip. Don't.

Before you tag, theme, or categorize anything, you need the gestalt of the study. Read the first five transcripts all the way through. No highlighting. No notes. No tagging. Just read.

What you're doing is letting your brain build an ambient model of the space — the kinds of people, the kinds of problems, the language they use — before you impose structure on it. If you start tagging on interview one, you'll bake in whatever was salient about that specific conversation as the frame for the other twenty-nine. That's how studies ship the first interviewee's opinion dressed up as a consensus.

Twenty minutes per transcript, five transcripts. You've spent an hour and a half. Resist the urge to feel unproductive. You're about to move four times faster than you would have.

Step 2 — Jot the themes that are already obvious

Now open a blank doc. Without re-reading anything, write down the five to ten themes that are sitting in your head. Rough language is fine. "People confused about billing" is a theme. "Something about handoff with design" is a theme. "Everyone hates the Tuesday sync" is a theme.

You will feel uncertain. That's correct. These are hypotheses, not conclusions. The next step either confirms, refines, or invalidates each one.

Ten minutes.

Step 3 — Read the remaining 25 with the themes in hand

Now you're reading with purpose. For each transcript, your job is:

Tag the moments that match one of the existing themes.
Add a new theme only when an existing one truly doesn't fit.
Promote a theme from "possible" to "confirmed" when a second, third, fourth independent interview supports it.

The "truly doesn't fit" bar is important. The natural instinct is to create a new theme for every small variation, and you end up with forty themes after twenty-five interviews. That's worse than having five. If a moment is 80% about billing confusion, tag it "billing confusion" even if it has a unique twist. The twist gets captured in the quote, not the taxonomy.

Ten minutes per transcript, twenty-five transcripts. Four hours — except most transcripts go much faster once the themes are stable. In practice, this step takes an hour and a half to two hours.

Step 4 — Collapse and prioritize

You now have a messy theme list with some overlapping categories. Merge the obvious duplicates. Rank what's left by two axes:

Frequency: how many distinct interviews mentioned it?
Pain: how severe was it for the people who mentioned it?

A theme that shows up in 25 of 30 but is mildly annoying ranks similarly to a theme that shows up in 6 of 30 but caused people to cancel. Both deserve attention. A theme that shows up in 3 of 30 and is mildly annoying probably doesn't.

Fifteen minutes.

Step 5 — Write the executive summary first, not last

This is the step everyone does backwards. Don't write the long report and then summarize it. Write the summary first, in three bullet points, while the data is fresh. Then flesh out the supporting detail.

Forcing yourself to commit to three bullets — before you've had time to soften them — is the single most clarifying act in synthesis. If you can't get it to three, you haven't finished thinking. Keep going.

The long report is much easier to write once the summary exists. Each bullet becomes a section. Each section has its frequency count, its representative quotes, its implications. The report writes itself.

Thirty minutes for the summary, an hour for the full writeup.

Total: roughly four hours

Step 1: 90 min. Step 2: 10 min. Step 3: 90 min. Step 4: 15 min. Step 5: 90 min. Call it four hours with coffee. For 30 interviews. An afternoon.

Mistakes we kept making (and stopped)

Reading to confirm. The biggest one. After interview three, you have a theory. Then interview four is just "evidence" for the theory. You stop noticing the thing that doesn't fit. The fix is Step 1: refuse to theorize until you've read five.

Over-tagging. Every theme that fits 90% is tempting to split into two themes. Don't. A messy coarse taxonomy is more useful than a clean fine one, because coarse themes get acted on and fine ones get ignored.

Writing the summary last. By the time you finish a 3000-word report, you are too deep in the details to summarize clearly. Write the summary while you can still see the shape of the thing. Revise it at the end if you need to.

Tagging during the first read. You tag what catches your attention. What catches your attention is a function of your priors. Your priors are wrong more often than you think.

Closing

We run this workflow in Honne's insights view, because it's where our transcripts, tags, and themes already live. But the method is tool-agnostic. It runs just as well in Notion, in Dovetail, in a stack of printed transcripts and a pencil. What matters is the sequencing — read, then theme, then tag, then rank, then summarize — and the discipline not to skip steps to feel productive sooner.

Thirty interviews is a lot. An afternoon is enough. The workflow is the difference.

Reading 30 interviews in an afternoon — a workflow