| name | case-study-writer |
| description | Write customer case studies and success stories using the Situation-Complication-Resolution (SCR) narrative framework and Barbara Minto's Pyramid Principle (conclusion first). Produces case studies where the headline states the result, the complication makes readers recognize their own situation, and every paragraph advances the story. Use when the user asks for a case study, customer story, customer success story, win story, reference write-up, or "turn this customer into a case study". Reads brand voice and service catalog from knowledge/. For a press announcement, see press-release-writer. For pitching the story to media, see pr-pitch-writer. For repurposing it into social, see content-repurposer. |
| metadata | {"grounded_in":["Pyramid Principle - Minto (SCR)"],"reads":["knowledge/brand/voice.md","knowledge/services/","knowledge/icp/personas.md","knowledge/content-library/case-studies/"],"writes":["output/case-study/"]} |
case-study-writer
Turns raw customer interview material into a structured case study. Mandatory SCR (Situation-Complication-Resolution) narrative structure with Pyramid Principle headline. Designed for sales enablement and SEO. The reader is a busy prospect who will decide in 10 seconds whether to keep reading - the conclusion goes first, not last.
When to use
- "Write a case study from these notes"
- "Turn this customer interview into a story"
- "Draft a customer success story for X"
- "We need a reference write-up"
Framework: SCR + Pyramid Principle (McKinsey storytelling / Barbara Minto)
Why SCR, not chronological storytelling:
Chronological case studies bury the hook. Readers are busy. SCR gets to the problem in the first paragraph, which is where readers decide whether this is relevant to them.
The 3 SCR components:
| Component | Length | What it must do | It has failed when |
|---|
| Situation | 1 paragraph | State only what the reader accepts without argument | It reads as a company profile, or the reader has to be persuaded of anything in it |
| Complication | 2 paragraphs | Make the reader recognise their own situation, and show the cost of doing nothing | It lists missing features instead of naming a business or human problem |
| Resolution | 2-3 paragraphs plus results | Answer the complication directly, with specifics | It introduces a benefit the complication never asked for |
SITUATION - establish context. What was true before? Set the scene without drama. This is what the audience already knows or can accept. Keep it short - one paragraph maximum. Do not start the story here.
COMPLICATION - introduce the tension. What changed? What made the situation unacceptable? This is where the reader recognizes their own situation. The complication is the hook. If readers don't see themselves in this section, the case study will not convert.
RESOLUTION - answer the complication. What was done? How did it work? What happened? The resolution is the proof - specific, measurable, credible. Vague resolutions ("they saw great results") are worthless.
Pyramid Principle application (Barbara Minto, The Pyramid Principle):
Minto's core rule is that ideas at any level must summarise the ideas grouped beneath them. Applied here, that means the headline is not a teaser for the story, it is the story's conclusion, and everything below it exists to support that one claim. If a section does not support the headline, it belongs in a different document.
- Lead with the headline conclusion (the result) - never save it for the end
- Support with the 3 SCR arguments
- Each argument supported by evidence: quotes, metrics, specific events
- Every section answers "so what?" - what does this prove for the reader?
- Minto's grouping test: read only the section headings in order. They should tell the whole story without the body text. If they do not, the structure is wrong, not the prose
Proof strength: rank the evidence before you write
Not all evidence carries the same weight, and a case study is only as strong as its weakest load-bearing claim. Rank what the inputs actually contain, and lead with the strongest.
| Rank | Evidence type | Why it ranks here |
|---|
| 1 | A before-and-after metric the customer measured themselves, with the measurement window stated | Checkable, and the customer owns it |
| 2 | A metric from your own product data that the customer has confirmed | Checkable, but the source has an interest |
| 3 | A named customer quote about a specific change | Not measurable, but attributable and human |
| 4 | A timeline fact: how long implementation took, when the change appeared | Weak alone, strong as corroboration |
| 5 | A qualitative statement with no number and no name | Nearly worthless as proof. Do not lead with it |
Decision rules:
- No rank 1 or rank 2 evidence means this is not a case study yet. Say so. Offer a shorter customer story or testimonial instead of stretching a rank-4 fact into a headline.
- The headline metric must be rank 1 or 2. A headline built on a quote is an opinion presented as a result.
- Never state a metric without its window. "Cut onboarding time by 60%" over one week and over one year are different claims.
- If the result has a cause other than your product (they also hired three people, or changed process at the same time), say so in the Resolution. A prospect who discovers the omission later stops believing the whole document, and the customer knows the truth and is reading the draft.
Case study structure using SCR + Pyramid:
1. HEADLINE: conclusion first - "[Metric improvement] for [Customer] with [product/approach]"
2. SITUATION: who they are, what their world looked like before
3. COMPLICATION: what became unacceptable, what they tried, what wasn't working
4. RESOLUTION: what they did (with you), how it worked, the specific steps
5. RESULTS: quantified outcomes - metrics, timeline, before/after comparison
6. QUOTE: customer's own words on the complication and/or resolution
7. SO WHAT: what should the ideal prospect believe or do after reading this?
Inputs needed (ask the user for each)
- Customer: name, industry, size, role of contact
- Problem: what they were doing before, what was broken
- Solution: which product or service they adopted
- Implementation: timeline, scope, who was involved
- Results: numbers, quotes, qualitative outcomes
- Permission: confirmed they can be named, or anonymous
If any are missing, ask. Do not invent.
Stop conditions
Run these before drafting, in priority order. The first three are hard rules: they mean this piece does not get written today, and no amount of good material overrides them.
| Condition | Action |
|---|
| No written permission and no decision to anonymise | Stop drafting a named study. Default to anonymised and set permission-confirmed: false. See the approval gate below |
| No rank 1 or rank 2 metric | Stop. Offer a testimonial or a short customer story instead. Padding a case study with rank-4 facts produces a document sales will not use twice |
| The contact who gave the material has left, or was never authorised to speak | Stop. Confirm with someone who is currently authorised before anything carries their company's name |
| The customer is still in an active escalation, renewal negotiation, or dispute | Flag hard and ask before proceeding. Timing, not writing, is the risk here |
| The result predates a major product change | Flag it and date the study, so a prospect does not buy a version that no longer exists |
| The metric is confidential or competitively sensitive to the customer | Ask what they will allow: a percentage instead of an absolute, a range, or an order of magnitude. Their limit is the limit |
Process
-
Load context. Read knowledge/brand/voice.md, the relevant knowledge/services/<service>.md, and 1-2 past case studies from knowledge/content-library/case-studies/ to mirror structure.
-
Confirm anonymization. If the user has not confirmed permission to name the customer, default to anonymized ("a Series B fintech in EMEA") and flag for review.
-
Map inputs to SCR before writing. Fill this in mentally before drafting:
- Situation:
- Complication:
- Resolution:
- Headline metric:
-
Write using the SCR + Pyramid structure:
# <Headline: conclusion first>
Format: "[Metric] for [Customer] with [product/approach]"
Example: "How Acme cut onboarding time from 14 days to 3 using automated workflows"
Rule: the headline must contain a specific number or measurable outcome.
## Situation
1 paragraph. Who is the customer, what they do, their scale. End with the context that
makes the complication make sense. This is the "given" - what the reader can accept
without argument.
## Complication
2 paragraphs.
- Paragraph 1: What changed or became unacceptable. Specific, not generic.
("Their CSM team was spending 40% of the week on manual data entry" not
"they had operational challenges")
- Paragraph 2: What they tried before and why it didn't work. The cost of inaction:
revenue at risk, hires deferred, churn increasing, competitor gaining ground.
Rule: a reader from the same situation should recognize themselves in this section.
If they don't, the complication is too abstract.
## Resolution
2-3 paragraphs.
- Why they chose this product (the decision, not the features)
- What they implemented and who was involved
- How the rollout went (specific steps, timeline, early signals)
One pull quote from the customer about the decision or early experience.
## Results
- Lead with the headline metric (the one in the title)
- 3-5 supporting metrics as a bulleted list (specific numbers, not ranges)
- Time-to-value: how long from start to result
- One pull quote about the outcome - the customer's own words on what changed
## What this means for you
1 paragraph. "So what" for the ideal reader.
- What should the reader believe is now possible for their team?
- Soft CTA: "If your team is dealing with <same complication>, see how <product> can help"
Rule: this section must name the complication, not the product. The reader should feel
spoken to, not sold to.
-
Voice rules:
- Customer is the hero, not the product
- Use the customer's words verbatim where possible (mark as "[customer quote]")
- Numbers are as specific as the source, never more so. If the customer said "nearly half", write "nearly half". Converting a hedge into a figure is a claim tightened past its source, and the source is a named customer who will read it
Rules
- Headline states the result before the story starts. Never save the punchline for the end.
- The complication is not a list of features that were missing. It is a human or business situation that became unacceptable. Write it that way.
- Never invent metrics, quotes, or implementation details. If the user gave incomplete inputs, mark gaps as
[NEEDS INPUT: <what's missing>] instead of fabricating.
- If permission is not confirmed, anonymize aggressively. Industry, region, and size only.
- Always show the cost of inaction in the complication. A case study without stakes has no tension and no pull.
- If every paragraph doesn't advance situation, complication, or resolution - cut it.
What to cut when it runs long
600-1000 words is the target, and a case study that runs over does not get read by the prospect it was written for. Cut in this priority order, and never cut upward past a step to protect a favourite paragraph:
- Company background in the Situation. Anything beyond what the Complication needs to make sense. This is where almost all overrun lives.
- Implementation detail in the Resolution. Keep the decision and the timeline, cut the configuration steps. A prospect wants to know it worked, not how the sprint was run.
- Supporting metrics below rank 3. Keep the strongest four, drop the rest. A weak metric beside a strong one drags the strong one down.
- The second quote if both make the same point. Keep the one in the customer's plainer words.
- Adjectives in the Results section. The number is the claim, and every adjective next to it reads as an attempt to make it sound bigger.
Never cut the cost of inaction from the Complication, and never cut the measurement window from a metric. Those are the two things that make the document credible, and they are the two most often removed for space. If the piece is still over 1000 words after step 5, the material is two case studies, not one. Say so rather than compressing both.
Minimum bar before this goes to sales
A case study that sales will not use twice was not worth writing. Before handing it over:
- The headline metric is rank 1 or 2, with a stated window.
- At least one direct customer quote, attributed to a real named or role-described person, appears verbatim from the interview.
- The Complication would be recognised by a prospect who has never heard of this customer.
- Frontmatter
approval reads granted DD-MM-YYYY, or the study is anonymised and says so on its face.
If any of the four fails, say which one and hand back a testimonial or a short story instead. Publishing a weak case study costs more than publishing nothing, because it becomes the proof asset everyone points at.
Customer approval before publication
A named case study is a public claim about another company, usually carrying their logo, a quote
attributed to a named employee, and a metric about their business. Every real company requires
sign-off on that, and this is a legal exposure, not a courtesy.
- Send the final draft to the customer contact for written approval before publication.
- Record the result in frontmatter:
approval: pending | granted DD-MM-YYYY | refused.
- Never publish, and never pass to
/pr-pitch-writer or /press-release-writer, while approval
is pending.
- If the customer asks for a metric to be softened or removed, that is their call, not a
negotiation. Change it.
- Anonymise by default until approval is explicit: "a Series B fintech" rather than the name.
What this skill cannot know
These are limitations of the skill, not of the customer's story. Anything below that reaches the draft must be labelled [NEEDS INPUT] or [UNVERIFIED] there, never asserted.
- Whether the customer's legal or communications team permits their name, logo or quote to be used. Verbal agreement from one contact is not company approval, and the person who gave you the interview is often not the person who can grant it.
- Whether the metric quoted is still accurate at publication. Case studies are drafted weeks before they publish and then run for years. Date the measurement, and re-confirm before any republication.
- Whether the named employee still works there or still endorses the quote. A quote from someone who has left is worth re-clearing with their successor, or anonymising to the role.
- Whether the result is actually attributable to the product. Correlation is what the inputs usually contain. Say what the customer says, name anything else that changed, and do not upgrade it to causation.
- Whether the metric is confidential. Revenue, headcount and churn figures are frequently material or competitively sensitive, and a public company has disclosure rules the marketing contact may not have considered.
- Whether the logo usage is within the terms of the contract. Some agreements allow the name but not the logo, or one placement but not paid media.
Related skills
/customer-research for running the customer interview properly, before there is anything to write up
/content-repurposer for turning the approved study into social and email derivatives, but only after approval is granted
/press-release-writer for an announcement version, which needs the same approval and usually a second one from the customer's PR team
/pr-pitch-writer for pitching the story to media, again only after approval is granted
/landing-page-writer when the study becomes a gated asset with its own page
/linkedin-post for the headline-metric post, keeping the number and its window intact
/messaging-framework when three case studies in a row surface the same complication, which is a positioning signal rather than a content one
/ad-campaign-writer for paid use of the proof, noting that ad consent is a separate permission from case-study consent