Case study builder

Case Study Builder

Produces case studies a skeptical buyer believes: a real customer in a specific situation, the exact intervention, and results stated as verified numbers with their source — or not stated at all. Every study runs on the situation → intervention → result spine, and every pull quote earns its placement by carrying either emotion or a number.

When to Use

• The user has a customer win (interview notes, call transcript, metrics, or just a name to interview) and wants a publishable case study.
• Sales needs proof assets and raw material exists.
• The user asks for "interview questions for a case study" — step 2 stands alone.
• Not for: placing the finished proof into page copy (landing-page-copywriter), spinning it into social posts (content-repurposer, linkedin-content-pack), or general market research (deep-research-brief).

Quick Reference

Action	Rule
Spine	Situation (before, stakes) → Intervention (what changed, how) → Result (measured after)
Metric rule	A number ships only with source + timeframe + customer confirmation; else [UNVERIFIED — do not publish]
Interview	Use the step 2 question list; record answers verbatim where possible
Pull quotes	2–3 max; each carries emotion or a number; verbatim, trimmed with "…" only
Approval	Customer sign-off on quotes + numbers before publication — human manages
Length	500–900 words body + headline + summary box

Procedure

1. Gather — confirm: the customer (name, role, company, permission status), what the product/intervention was, raw material available (transcript, notes, metrics, emails), and where the study will live. No raw material yet → deliver step 2's interview kit as the output and stop there.
2. Interview kit — supply this question list (adapt names, keep the shape):
   • Situation: "What was breaking before?" / "What did that cost you — time, money, morale?" / "What had you already tried?"
   • Intervention: "What made you pick this?" / "What did the first week look like?" / "What almost stopped you?"
   • Result: "What number moved, from what to what, over what period?" / "Where does that number come from?" / "What surprised you?"
   • Quote bait: "If a peer asked whether this was worth it, what would you tell them?"
3. Extract the spine — from the material, write one sentence per stage: situation, intervention, result. If the result sentence has no number, go get one (ask the user/customer) or plan an outcome-described-honestly study — do not decorate with fake precision.
4. Verify every metric — for each number: source (dashboard, report, customer statement), timeframe, and baseline. Log them in the metric ledger. Anything unconfirmed is tagged [UNVERIFIED — do not publish] in-text and listed with exactly what confirmation is needed. "Roughly doubled" confirmed beats "103%" invented.
5. Select pull quotes — from verbatim material only. Pick 2–3 that carry feeling ("I stopped dreading Mondays") or numbers ("we cut review time from 3 days to 4 hours"). No verbatim quotes available → mark [QUOTE PENDING customer approval]; never compose a quote and attribute it.
6. Draft — headline states the result ("[Customer] cut [metric] by [X] in [time]"), summary box (3 bullets: who, what, result), then the spine in 500–900 words. Specifics over superlatives; the customer is the hero, the product is the tool.
7. Deliver + approval route — assemble the Output Template including the metric ledger and an approval checklist. Publication waits for customer sign-off; the human runs that step.

Output Template

# [Customer] [verb] [metric change] in [timeframe]

**Summary** — Who: [customer, role, company] | Challenge: [one line] | Result: [verified number + timeframe]

## The situation
[before-state with stakes, in the customer's words where possible]

## The intervention
[what changed, how adoption actually went, including one honest friction point]

## The result
[verified numbers with baseline + timeframe; qualitative wins after the numbers]

> "[Pull quote 1 — verbatim]" — [Name, Role]
> "[Pull quote 2 — verbatim]" — [Name, Role]

---
**Metric ledger:** [metric → source → timeframe → confirmed by]
**Approval checklist (human):** [ ] quotes approved  [ ] numbers approved  [ ] logo/name usage approved

Pitfalls

• Invented precision — "40% improvement" appears because it sounds better than "meaningfully faster". Recovery: run the step 4 ledger; any number without source + timeframe + confirmation is replaced with the honest qualitative phrasing or tagged [UNVERIFIED — do not publish].
• Composed quotes — a quote gets "cleaned up" into words the customer never said. Recovery: diff every quote against the transcript; trims marked with "…" are fine, new words are not — revert or mark [QUOTE PENDING].
• Missing before-state — the study opens with the win, so the win has no weight. Recovery: rebuild from the step 3 spine; if the situation sentence is empty, ask the situation questions from step 2 before drafting further.
• Product-hero syndrome — the draft reads like a brochure; the customer disappears. Recovery: count sentence subjects in each section; if the product leads more than the customer, rewrite with the customer as actor and the product as instrument.
• Publishing before sign-off — the study ships and the customer objects to a number or their name. Recovery: the approval checklist is part of the deliverable; if published early, notify the human to pull it and route approval immediately.

Verification

• Spine complete: situation, intervention, and result each land in one clear sentence before the draft exists
• Every published number appears in the metric ledger with source, baseline, timeframe, and confirmation
• Zero [UNVERIFIED] numbers in publishable text; pending items listed with what would confirm them
• 2–3 pull quotes, all verbatim-traceable or explicitly [QUOTE PENDING]
• Headline states the verified result; body 500–900 words
• Approval checklist attached; publication marked as blocked on customer sign-off