Case studies that respect confidences: writing proof when you cannot name the client
In 2025 the American Bar Association told lawyers that stripping the names out is not enough. Formal Opinion 511R requires a client's informed consent before a lawyer posts about a matter. That holds even in hypothetical or abstract form, when a reader could later infer who the client was.
That standard is stricter than the one most case studies meet. It is also the correct standard for anyone writing about work a client asked them to keep quiet.
So two questions follow. What can a case study carry when the name is off the table, and how do you ask for permission that holds?
The name is not the proof, though it is not nothing
A case study earns its place by showing a reader how the work happened. The situation, the constraint, the decision, the sequence, and what changed by the end. A name adds credibility to that account. It does not supply the account.
The cost of losing the name is worth stating plainly. A marketing manager at a UK contract manufacturer framed the objection well. Say you produced a large sales increase for an equipment manufacturer. That carries less weight than the same claim attached to a brand and a named person.
A second manufacturer in the same piece supplies the other half. What moves a technical buyer is how the approach was used and what that released. Which company used it matters less.
Checkability is the trust bar now. A claim a stranger can confirm beats one that only sounds impressive. An unnamed case study fails that test at the door. Everything else in it has to work harder as a result.
Anonymizing got weaker while you were not looking
The old test was whether a reader would guess. That test has expired.
In February 2026, researchers from ETH Zurich, Anthropic, and collaborating institutions published work on automated deanonymization. Software agents identified people behind anonymous online profiles from their writing alone. Recall ran between roughly 25% and 67%, at precision between 70% and 90%. The cost came to about $1 to $4 per target.
The study looked at forum users rather than corporate case studies. The mechanism transfers, because it runs on exactly what an anonymized case study leaves in. Sector, rough size, region, timeline, and the shape of the problem.
So the test changed. Ask whether a reader with an assistant and ten minutes could match your description to a company. If the answer is yes, you have published the client's name in a slower format.
That is what Formal Opinion 511R said first, in the language of professional conduct rather than machine learning.
Permission has rungs
Most people ask once, hear no, and stop. The no usually applies to one rung rather than to the whole ladder.
- Named, quoted, and reachable. The client is named, speaks in their own words, and agrees to take a reference call.
- Named, unquoted. The client is named and the results are stated, with no quote and no call.
- Described, not named. Sector, rough size, and results given as ranges or percentages.
- Pattern only. The mechanism, the sequence, and the decisions, with no client-specific figures.
Ask for the top rung. Accept the rung the client can live with, and put the request in writing at every level.
Healthcare providers work to a federal template worth copying outside healthcare. A valid authorization under 45 CFR 164.508 names six things. What information is used, who uses it, where it appears, the purpose, an expiration, and the right to revoke. Six lines on a page, and they remove the argument you would otherwise have in eighteen months.
Timing does more work than wording. Ask in the month after something worked. Willingness falls as the engagement recedes and the client's attention moves on.
At rung one the wording still matters. What a strong client quote is doing is a separate craft from the case study around it.
The composite is the line
Merging three clients into one invented client produces a story that reads as fact and is not one.
If that composite carries a quote, it becomes a testimonial attributed to a person who does not exist. The FTC's rule on consumer reviews and testimonials covers that directly. The exposure is beside the point. The composite spends the only asset an unnamed case study has, which is that the account is true.
One solo practitioner with more than 70 projects behind them puts it in a line. A vague true case beats a specific invented one.
Do not invent a company name either. A descriptor does the job without misleading anyone. A national logistics provider. A twelve-partner litigation firm. When a repeated label is needed, say on first use that it stands in for a name.
Spend candor where the name used to sit
Anonymity removes the client's communications review from the difficult parts of the story. That is worth more than most people notice.
A named case study travels through an approval chain, and the friction gets sanded off on the way. An unnamed one does not. You can write the wrong turn, the month that produced nothing, and the assumption that failed.
The agency Velocity Partners makes this the central argument for anonymous case studies, and it holds up. Candor fills the space the name used to hold. A story with visible friction reads as an account, and a story with none reads as an advertisement.
When the rules are not yours to bend
Lawyers work under Rule 1.6, which requires informed consent before revealing information relating to a representation. ABA Formal Opinion 480 extends that to information sitting in a public record. South Carolina's version reaches generally known client information too. Rule 7.1 adds a separate constraint. A truthful account of a result can still mislead when it invites an expectation the next matter cannot support.
Clinicians need written authorization before any patient story promotes the practice. A de-identified account avoids the requirement, and the deanonymization problem above decides whether it is de-identified.
Investment advisers work under the SEC Marketing Rule, which permits testimonials and endorsements on disclosure conditions. The Division of Examinations issued a risk alert in December 2025. It found advisers falling short on the timing and placement of those disclosures. Check your own bar or regulator before publishing, since the state versions differ.
What this will not do
An anonymized case study converts below a named one. The repair is more permission at the front, not better disguising at the back.
Some engagements cannot be written up at any level. Pattern-only writing has a floor, and below it there is nothing to publish that would help anyone.
And pattern-only pieces read as vague to a portion of readers. No wording fixes that. The trade is a smaller claim in exchange for one that survives a check.
What to do this week
Take your last three engagements. For each one, name the rung you could reach if you asked on Friday, then ask.
The rung available is usually higher than the one assumed, because the ladder above the first no rarely gets tested.
