The strongest thing you can publish about your work is a specific customer with a specific result and an honest caveat. Most case studies do the opposite: they blur the customer, inflate the result and remove every qualification, which produces something nobody believes and nobody can verify. The structures below keep the credibility in.
The spine every version shares
- Who. Named if permitted, otherwise described precisely enough to be useful: size, sector, situation.
- The problem, in their terms. What was going wrong, and what it was costing or blocking.
- What was actually done. Specific enough that a peer recognises the work.
- The result, with a baseline and a timeframe. "32% reduction from 118 hours to 80 per month, measured over three months."
- The caveat. What else changed, what did not improve, who this might not work for.
The six structures
Replace every [bracket].
1. Problem → intervention → result
Use when: the result is the story.
[Customer] was dealing with [specific problem], which meant [consequence in money, time or risk]. They had already tried [thing], which did not work because [reason]. We did [specific work] over [period]. Result: [figure, with baseline and measurement period].
2. The before-and-after table
Use when: the change is measurable across several dimensions.
| Metric | Before | After | Measured over |
|---|---|---|---|
| [metric] | [value] | [value] | [period] |
| [metric] | [value] | [value] | [period] |
Every figure needs a source you can point to. If you cannot say where a number came from, leave it out.
3. The obstacle narrative
Use when: the interesting part is what nearly went wrong.
The obvious approach was [thing], and it failed because [specific reason]. What actually worked was [alternative], after [what was learned]. This took [period] longer than expected, and the result was [outcome].
4. The day-in-the-life
Use when: the change is about how work feels and flows.
Before: describe one real task end to end, in the customer's own words where possible. After: the same task, described the same way. Difference in time: [figure]. Difference in steps or handoffs: [number].
5. The short quote-led version
Use when: you have a customer willing to say something specific.
One quote that states a concrete outcome, followed by two sentences of context supplied by you. Avoid quotes that praise you without stating anything verifiable — they read as favours.
6. The internal case note
Use when: building a library before anyone has approved anything external.
Written for your own team, not for publication: what happened, what was learned, what you would repeat and what you would change. Several of these can later be combined into one external version once permission exists.
Writing numbers honestly
A figure without a baseline and a period is decoration. Three rules that keep results defensible:
- State the starting point. "47% faster" than what? Give the before value.
- Name the measurement window. A comparison across six weeks of one quarter against the same six weeks of the previous one, not "since we started".
- Declare what else changed. If they also hired two people or changed their pricing, say so. Omitting it is the version that gets caught later.
Getting approval without three rounds
Case studies stall in approvals, not in writing. Two things reduce the cycles: send the numbers in a separate, plain document so the customer can check figures without reading prose, and agree on anonymity upfront — ask whether they want naming, description, or full anonymity, and write to that answer the first time.
Editing time, from our 30-day benchmark
Our 30-day benchmark measured roughly 8 minutes of editing per 1,000 words on a general assistant and about 22 minutes on a low-cost dedicated writer. Case studies were the format where the difference showed most, because they depend on holding several specific facts steady at once — and the cheaper drafts had a habit of quietly rounding numbers into claims.
Make the customer the subject
Baseline, period, caveat — that is what makes a result believable.
Try Rytr →Frequently Asked Questions
How long should a case study be?
Between 400 and 900 words in most cases. Long enough for a specific problem, the actual intervention, a result with its baseline, and one honest caveat — that is what carries credibility.
Should I name the customer?
If they agree, yes — a named account is worth considerably more than an anonymous one. If not, describe them precisely enough to be credible: sector, size, and the specific situation.
What makes case study numbers untrustworthy?
A percentage with no baseline, no measurement window, and no mention of what else changed at the same time. Each one is fixable in a sentence.
Can AI write a case study?
It can shape the narrative once given the facts. It does not know your results, and if left to fill gaps it will produce plausible numbers, which is the failure mode to avoid.
How do I get approval faster?
Send figures separately from prose so they can be checked independently, and settle naming versus anonymity before writing rather than after.