WorkflowWork · Research

Run a complete user research cycle from question to synthesized report

Most research fails at the seams — between question and study design, or between notes and findings.

Outcome: A sharpened research question, a study design that could actually answer it, and a real synthesis at the end instead of a pile of session notes.
Time: Spans the full research cycle, likely 1-3 weeks

The guided runner is a Pro feature

Every step is fully readable and copyable below, free. Pro adds step sequencing, carried inputs, checkpoints, and resumable runs.

Unlock with Pro →

Research cycles often fail not at any single step but at the seams between them — a fuzzy question produces an unfocused study design, and good session notes get compiled instead of synthesized. This playbook runs the full cycle as connected steps, each one built on the last, rather than treating question-design, execution, and synthesis as separate efforts that happen to follow each other.

When not to use this

For a quick, informal check-in with a couple of users where formal rigor isn't warranted, this is more process than the situation needs. This is for a real research cycle meant to inform a specific decision.

Before you start

A topic or rough question you want to research
Some ability to actually talk to or survey real users

01 02 03

Turn the topic into an answerable research question

Force the topic into a specific question tied to a real decision before designing any study around it.

Fill in

{{topic}} — The topic{{decision}} — What decision this needs to inform

Prompt

Topic: {{topic}}
Decision this needs to inform: {{decision}}

Help me turn this into an answerable question: push back if it's
still too vague, propose 2-3 specific versions tied to the decision,
and for the strongest one, tell me what a clear answer would
actually look like.

Why it works

Tying the question to a specific decision is what keeps the whole cycle aimed at something actionable instead of interesting-but- unused findings.

Checkpoint

You have one specific, answerable research question tied to a real decision.

Common mistakes

×Accepting a vague topic as the research question instead of sharpening it first.
×Not naming the decision it needs to inform, leaving no way to judge if the question is useful.

Design a study that could actually answer the question given real access

Match the study design honestly to who you can actually reach, not an idealized sample.

Fill in

{{access}} — Who you can actually talk to or survey

Prompt

Given this question and my real access: {{access}}

Recommend a study design (interviews, survey, usability test, or a
mix) that could actually produce a confident answer given this
access — and be honest if my access doesn't support real confidence.

Why it works

Checking the design against real access, not a wish-list sample size, prevents planning a study you can't actually execute as designed.

Checkpoint

You have a study design matched honestly to your real access.

Common mistakes

×Designing for a sample size you don't actually have access to.
×Ignoring the honest confidence-level caveat and treating results as more certain than the access supports.

Synthesize findings across sessions into real themes

Turn raw notes into evidence-backed findings, not a session-by- session compilation.

Fill in

{{notes}} — Your raw notes from all sessions

Prompt

Here are my raw notes from all sessions: {{notes}}

Identify themes that recur across sessions (not what any single
session said), state each as a finding with supporting
sessions/quotes, flag any genuine disagreement between sessions
rather than smoothing it over, and connect back to the original
research question — what does this actually answer, and what's
still open?

Why it works

Requiring themes to recur across sessions before calling them findings, and flagging disagreement rather than smoothing it, is what makes the output real synthesis instead of a compiled summary that leaves the pattern-finding work to the reader.

Checkpoint

You have a small number of clear findings, each backed by evidence from multiple sessions, tied back to the original question.

Common mistakes

×Organizing findings by session/participant instead of by recurring theme.
×Smoothing over genuine disagreement between sessions into one artificially unified finding.

Follow-ups

Turn this synthesis into a one-page summary for a stakeholder who won't read the full findings.
What follow-up research would confirm or challenge the strongest finding here?

Did this work?

You end with a small number of clear, evidence-backed findings tied directly back to the decision the research was meant to inform.

Tested on claude-opus-5. Evidence status is draft; it moves to battle-tested only on recorded runs, never by hand.