Five themes, each carrying its coverage. One participant saying something twice is not a pattern, and the bar is there to keep that honest.
The problem
Recruiting takes a week, interviews take a week, and synthesis — reading, tagging, clustering, arguing about whether two phrasings are the same theme — takes another week if you are rigorous. The decision is usually made before the deck is ready.
Speeding that up is easy and mostly wrong: a fast summary of five transcripts is an impression, and an impression that arrives with quotes attached is more persuasive than one that does not, whether or not the quotes are real.
What it does
Transcripts
Five interviews, anonymized. The raw material, kept where you can read it.
Observations
Each one anchored to a specific quote, so nothing floats free of its source.
Themes
Clustered, and each one says how many participants it rests on.
Quote check
Every quote verified against its transcript, word for word.
What it showed
Themes, from five transcripts
5
Each with the number of participants behind it shown next to it.
Contradictions named
2
Kept as findings rather than rounded into a majority view.
Coverage is the load-bearing feature. “The second week has no owner” rests on four of five participants; “it loses to the shift, not to other apps” rests on two. Both are worth knowing and only one of them is a pattern, and a synthesis that prints them at the same weight has destroyed the distinction.
The other one is the contradictions tab. Both humans and tools round toward consensus, and the disagreement is usually the finding — here, participants who could repeat the anonymity guarantee accurately and still refused to use the product, which makes it a trust position rather than a comms gap.
What it cannot do
It verifies that a quote exists in the transcript. It cannot verify that the quote means what the theme says it means, and that is the judgment the tool is there to inform rather than replace.
Fast synthesis of bad interviews is fast garbage. Nothing here improves the questions that were asked.
The long version of the argument behind it is The 2026 product leader’s toolset.