Turn a usability test into a spec the team can build

A usability test is the highest-signal research you can run, and the most likely to evaporate. You watch five people fail at the same step, you feel it in your gut, and then it becomes a slide. By the time engineering picks up the work, the why is gone — they're building from a one-line ticket, not from the moment a user said "wait, where did it go?" This flow keeps the observation attached to the fix.

The input

The raw output of a few sessions, in whatever shape you have it:

• Observation notes ("P3 clicked Save twice, didn't trust it worked")
• Verbatim quotes from think-aloud
• Timestamps or clip descriptions
• The task each participant was attempting

No need to clean it up first — messy notes are fine.

The flow

1. Paste the findings into the Evidence Board. Open the Evidence tab and drop your notes in. Each observation becomes an evidence item; the AI pulls out the quote and the friction type (confusion, workaround, unmet expectation).

2. Cluster by underlying problem, not by screen. Switch to the Clusters layout and generate clusters. Two participants failing on different screens for the same reason — "the system never confirms my action" — land in one cluster. That's the real finding.

3. Pick the cluster that's costing the most. Each cluster carries a one-line summary and its underlying observations, attribution intact. Choose the one with the heaviest, clearest friction.

4. Synthesize an intent from it. Select the items and click Synthesize Intent. The resulting spec opens with its outcomes already anchored to specific participant moments — so the "why" rides along into engineering instead of being left in the readout.

The output

A spec that carries the test's evidence forward:

• Outcomes tied to observed failure, not to a hunch
• Quotes preserved with attribution, so the fix is defensible
• The pattern named once, instead of one ticket per screen
• A handoff that keeps the texture — engineering sees what the user actually did

Why this beats a findings deck

A deck is where research goes to be admired and then ignored. The gap between "here's what we saw" and "here's what we'll build" is where the insight leaks out. Synthesizing straight from the evidence closes that gap: the spec is the synthesis, and every outcome traces back to a clip. The research compounds instead of expiring at the readout.

Try it yourself

1. Gather notes from your last 3–5 usability sessions
2. Open Pathmode → product → Evidence tab and paste them
3. Switch to Clusters, generate clusters
4. Select the heaviest cluster and click Synthesize Intent

Related

• Use case: Find the friction pattern across 5 user interviews
• Use case: Catch the experience gaps in a spec before build
• Playbook: Interview to Evidence