Find the friction pattern across 5 user interviews

User interviews produce data that is overwhelmingly qualitative and almost impossible to summarize honestly. Five stories, five contexts, five sets of complaints. The temptation is to either cherry-pick the quote that fits your hypothesis, or write a 12-page synthesis nobody reads. This use case finds the pattern without either failure mode.

The input

Five interview transcripts. The format does not matter:

• Otter, Fathom, or Granola exports
• Loom transcripts
• Hand-typed notes
• A mix of all three

Each one ideally tagged with the participant's role and team size.

The flow

1. Paste each transcript into the composer. Open the Evidence tab and paste one transcript at a time. A long paste is extracted into candidate rows, each carrying a quote and a type. You keep the rows worth keeping and drop the rest before anything is saved.

2. Everything you keep lands Unreviewed. Extracted rows go to the Inbox with the participant attribution and source intact. Extraction is a capture step. It does not decide what is true.

3. Name the pattern as a claim. Start an intent and write the friction as an outcome or a constraint in plain words: "the system requires me to know my own role before it will help me." Two participants can complain about different features and still be hitting that one claim.

4. Test the claim against the Inbox. Walk the signals. Each one gets Use as backing, Keep unreviewed, or Dismiss. Backing a signal asks you which exact claim it supports, so you find out quickly whether four participants really support the same statement or whether you were reading four different problems as one.

5. Read the answer off the claim. A claim carrying reviewed backing from four independent participants is a pattern. A claim carrying one unreviewed note is a hypothesis you have not tested yet. The spec says which is which without you having to remember.

The output

A research record that:

• Names the pattern as one claim, not 12 pages
• Shows what backs it with participant attribution and your review
• Distinguishes signal from noise, because outliers stay visible as unreviewed rather than disappearing into a summary
• Hands off to engineering without losing the qualitative texture

Five interviews become one artifact your team can check rather than take on faith.

Why this beats traditional synthesis

Traditional research synthesis lives in a doc nobody reads after the readout. By the time the friction reaches engineering it is detached from the original quotes, and the original interviews might as well not exist.

Pathmode keeps the chain intact. Every backed claim points at the specific quote from the specific participant, and at the person who judged that the quote supports the claim. The research investment compounds instead of evaporating.

Try it yourself

1. Gather 5 recent interview transcripts, or 3, since the flow works at small N
2. Open Pathmode, choose a product, go to the Evidence tab
3. Paste each transcript and keep the extracted rows worth keeping
4. Start an intent that states the friction as a claim
5. Work the Inbox and use the supporting signals as backing for that exact claim

Related

• Playbook: Interview to Evidence
• Use case: Turn 30 support tickets into a prioritized spec
• Use case: Anchor every outcome to user evidence