B2B case study production

Applies when real client results exist and proof assets do not. Produces a series of case studies a skeptical buyer believes and a sales team actually sends.

The job of a case study

A case study succeeds when a specific buyer recognizes their own situation in the first hundred words and finds the causality plausible all the way down. It is written for the skeptical read: a buyer who has been burned before, scanning for the seam where marketing took over from reality. Applause is a side effect. Recognition closes deals.

Real numbers, few of them

One or two verifiable anchor numbers per story. Everything else is narrative: what changed operationally, who stopped doing what, which meeting got shorter. A page carpeted in metrics reads as invented even when every figure is real, because buyers know real engagements produce two clean numbers and a lot of texture. And the rule under the rule: never fabricate, never round up, never "roughly double" a 1.7x. One invented figure, discovered, poisons every honest story in the set.

Anonymize without bleaching

Clients who refuse to be named still produce usable stories, and the disguise follows a strict trade: drop the identifiers, keep the texture. Describe the client by durable attributes, a size band, a specialty or vertical, a region, a situation, specific enough that a peer thinks "that is exactly us", general enough that a determined guesser cannot confirm the identity. The full attribute table, rounding rules, and the re-identification check are in references/anonymization-rubric.md. The mediocre version anonymizes by deleting everything interesting, leaving "a mid-sized company improved efficiency", which protects the client by making the story worthless.

Open on the pain the reader owns

The problem section is written in the buyer's own vocabulary, mined from sales calls, and aimed at the pain the reader personally absorbs, the escalations they field, the report they dread, the vendor they cannot reach. Loss framing needs care: told bluntly "you are losing money", buyers who feel fine bounce off. The version that lands describes quiet underperformance, the money that disappears a little at a time, because it matches how the reader actually experiences the problem. Steal their phrases verbatim; a buyer's own words in the opening paragraph does more work than any statistic on the page.

Intake before writing

Interview the delivery team and, where possible, the client, using the full question set in references/intake-questionnaire.md. The single highest-value answer is the moment the engagement nearly failed, because a story with a wobble in the middle reads true and a straight line reads like an ad. Writing from memory or from a dashboard export produces the straight line every time.

Build a series, not a trophy

One case study is an anecdote. Five or six, each on its own page, is evidence. Plan the set as coverage: one story per segment, per specialty, per problem shape, so every priority buyer finds a mirror. Sequence production so the strongest verifiable numbers ship first. Structure and a worked example are in references/story-structure.md.

Deploy where buyers are in motion

Case studies are conversion assets. They belong where an evaluating buyer already is: the site's proof section, the sales follow-up, the proposal appendix, the outbound proof line. Pushed weekly at an existing audience of followers, the fifth case study lands as repetition and the format sours internally, a predictable failure that gets misread as "case studies don't work". They work on buyers in motion. Distribution to a general audience is a bonus channel, rationed and never the plan.

What good looks like

The best operators collect one buyer phrase per sales call into a swipe file and open every problem section from it. The mediocre version is a wall of unverifiable metrics over a straight-line narrative with a logo the client never approved. Good output survives three tests: a skeptical stranger finds the causality plausible, the sales team sends it without being asked, and the featured client reads it and confirms every word, comfortable that nobody else can prove it is them.

Rules

• MUST get explicit written approval for every name, logo, and exact figure before publication; anonymized drafts speed this instead of skipping it.
• MUST anchor each story on at most two verified numbers, sourced from the engagement record.
• MUST run the re-identification check on every anonymized story before it ships.
• NEVER invent, inflate, or "estimate" a metric; a story without a clean number ships as narrative with no number.
• NEVER imply causality the delivery team will not defend in the room.