Case Study Skill

A Claude skill that drafts customer case studies in an impact-first format — the measurable result in the title, hero metrics up top — from the messy inputs teams actually have: customer interview transcripts, CSM and AE notes, Slack threads, product usage data, CRM notes, and web research. Honest-metrics guardrails are built in: every number must trace to a source that can support it, correlation is never written as causation, and qualifiers travel with their numbers. When impact metrics or specifics are missing, the skill doesn't guess — it hands you targeted questions, grouped by who can answer them.

Distilled from case-study work Daniel Glickman contributed to across enterprise SaaS — including a published customer case study on AI-adoption measurement — and from the claim-qualifier discipline that runs through his own published record.

What it does

• Impact-first structure: title with the measurable result ("How [Customer] lifted X 62% with [Product]"), a hero-metrics block, then About → Challenge → Solution → Results → Quote
• Consumes the real input mess — interview transcripts, CSM/AE interviews, Slack messages, usage data, Salesforce notes, web research — and classifies each by what it can honestly support
• Internal hearsay is a lead, not a source: a CSM's "they said pipeline doubled" becomes a question for the customer, never a claim in the draft
• Audits every claim: source, verb honesty, qualifier — with a claims-audit table shipped alongside every draft
• Asks instead of guessing: missing impact metrics, baselines, and quote approvals come back as forward-ready questions grouped by owner (customer, CSM/AE, analytics)
• Verbatim quotes only, each marked approved / pending / internal-only; consistent anonymization on request
• Audit-only mode: point it at an existing draft and get the corrections without a rewrite

The verb rule (the part most case studies get wrong)

Adoption, coverage, and reach get honest verbs — spans, reaches, adopted across. Causal verbs — drove, generated, produced — are reserved for results whose source actually establishes causation. When in doubt, the weaker verb wins. A case study that overclaims doesn't survive the skeptical read, and the skeptical read is the one that matters: the customer's legal team, an analyst, or a prospect who has seen too many vendor stories.

When it triggers

Phrases like:

• "Draft a case study from these inputs"
• "Turn this customer win into a case study"
• "What do we still need to make this a case study"
• "Audit the claims in this case study"

Installation

Option A — Claude.ai (no command line — where most marketers work)

1. Download case-study-skill.skill from this repo (top level, one click).
2. In claude.ai, open Settings → Capabilities, find Skills, and upload the file.
3. Done — next time you ask for the job, the skill runs.

Option B — Claude Code (if you work in the terminal)

Copy the skill folder into your Claude skills directory:

# Personal (all projects)
cp -r case-study-skill ~/.claude/skills/

# Or project-scoped
cp -r case-study-skill /path/to/project/.claude/skills/

Restart Claude Code (or start a new session) and the skill will be available.

Usage

Dump in whatever you have and ask:

Draft a case study from these inputs: [interview transcript] [CSM notes] [Slack thread] [usage data] [CRM notes]. Customer agreed to be named: yes/no.

Or audit-only:

Audit the claims in this case study draft: [paste]

See examples/ for a worked multi-source example (invented, illustrative data) — including internal hearsay converted to customer questions and an unsourced metric held out of the draft.

Repository contents

• case-study-skill/SKILL.md — the skill
• examples/ — worked example (illustrative, invented data)
• case-study-skill.skill — packaged for upload

Part of the PMM Skill Stack

One of four free Claude Skills for product marketers: cmoconfessions.com/skills. The method behind the stack: Kill the Battle Card.