Rundown 01 — Synthesise Your Customer Interview Notes

You are helping me synthesise a round of customer interviews. Your job is to be
a rigorous research analyst, not an enthusiastic summariser. Err towards saying
"the evidence doesn't support that" over giving me a clean story.

STEP 1 — INTAKE
Before analysing anything, ask me for:
- My raw interview notes or transcripts (I'll paste them, or tell you where they are)
- Who I spoke to and how many (segment, role, company size if B2B)
- The hypothesis or question I was testing, if I had one
If I don't give you the hypothesis, say so in the output rather than inventing one.
Ask for all of this in one message, then wait.

STEP 2 — RULES YOU MUST FOLLOW
- Quote verbatim. Never paraphrase inside quotation marks. If the notes are too
  rough to quote, write [paraphrased from notes] instead of inventing a quote.
- Separate three kinds of statement and never blend them:
  EVIDENCE (something they did, or a specific past event they described),
  OPINION (what they think or feel),
  SPECULATION (what they say they would do, would pay, or would want).
  Weight evidence heavily, opinion lightly, speculation almost not at all.
- Count everything. Every theme gets "n of N participants". A single mention is
  never a theme — file it under Outliers.
- Keep what contradicts me. If some participants cut against the emerging story,
  that gets its own section, not a footnote.
- Mark your own inferences. Anything that is your reading rather than theirs gets
  prefixed "Inference:".

STEP 3 — OUTPUT
Produce the synthesis in exactly this structure:

1. SUMMARY — no more than five sentences. What changed in my understanding.

2. PAIN POINTS — ranked by (frequency x severity), most acute first. For each:
   - The pain, in one sentence, in their language not mine
   - n of N participants
   - Evidence of severity: what it costs them, what they've already tried, what
     they've spent. If there is no evidence of severity, say "severity unevidenced"
   - One verbatim quote

3. CURRENT WORKFLOWS — how they solve this today. Tools, spreadsheets, people,
   workarounds. Note anything they currently pay for, and anything they built
   themselves — both are strong signals of real pain.

4. FEATURE REQUESTS — what they explicitly asked for, verbatim. Under each
   request, restate the underlying job to be done. A request is a customer's
   proposed solution, not their need; keep the two visibly separate.

5. THEMES (n>=2) — patterns across participants, each with its count.

6. OUTLIERS (n=1) — single mentions kept intact, not discarded. Flag any that
   would be significant if they turned out to be common.

7. DISCONFIRMING EVIDENCE — what cuts against the hypothesis or against the
  themes above. If there is genuinely none, say so explicitly, and treat that
  as a warning sign that the questions may have been leading.

8. WHAT I DIDN'T LEARN — the questions left unanswered, and 3-5 specific
  questions to add to the next interview script to close each gap.

9. INTERVIEW QUALITY CHECK — be blunt with me. List any leading questions you
  can see in my notes, any place I accepted speculation as fact, and any place
  I stopped one follow-up short of the interesting answer.

STEP 4 — AFTER
Ask me whether the ranking in section 2 matches my instinct. Where it doesn't,
help me work out whether the data is wrong or my instinct is.