When a case study client is a composite, not a real one
You are reading a case study and the client is “a DTC skincare brand.” No name, no logo, no link out. The numbers are specific and there is nothing to check. One thing that can mean: the result belongs to no single company.
The tell: a client you can’t look up
Two versions of the same sentence. The first: “Client X, a 40-person DTC skincare brand, ran a Q3 campaign and lifted engagement rate by 32 percent.” The second: “A DTC skincare brand lifted engagement rate by 32 percent.” The claims look nearly identical. What you can do with them does not. Three abilities leave with the name: finding the account to see what was posted, checking the campaign window against a public record of launches and seasonal peaks, and asking the client’s marketing lead whether the number matches their reporting.
A generic descriptor is not proof of a composite. Real, checkable engagements get anonymised for ordinary reasons; a confidentiality clause is the most boring explanation. It is a tell, not a verdict, and it sets the next question: one company behind this number, or several?
How a composite figure actually gets built
The following walkthrough is hypothetical: no real agency, client or reported result, and every number invented.
Picture a hypothetical agency that has run comparable campaigns for four clients in one broad category. The engagement lift came out differently every time: a hypothetical 18 percent, 40 percent, 9 percent and 61 percent. Four numbers, no obvious way to tell them as one story.
So the deck averages them. Those four make 128, which divides by four to give 32. The slide says the typical client sees a 32 percent lift, without saying which client, or that there were four, or that the figure is a mean. Not one of the four saw 32 percent. The number is arithmetically correct and belongs to nobody. That is the mechanical point of a composite: every input genuine, the calculation sound, the result describing an account that does not exist.
Composite is not the same claim as fabricated
Nothing in a composite has to be invented for the case study to mislead. Every number can come straight from platform reporting and the arithmetic can check out. The false part is not any figure in the blend. It is the implied claim wrapped around it, that this happened, once, to one identifiable client.
That changes what verification means. Verifying a named client’s result verifies a company: find the account, match the window, ask someone who was there. Verifying a composite means verifying a method. Which accounts went in, weighted how, which excluded, mean or median? Almost no case study says, because saying so would announce that a method was needed. Before asking what backs a number, which is where a case study number is only as good as what backs it up begins, you have to know whose number it is.
Why a composite resists verification by design
Each route is closed off by the structure of the claim:
- No company name to search. Every check starts with an entity, and there isn’t one.
- No single dated campaign window. Several engagements may span different quarters, so there is no period to check against.
- Nobody at the client to ask. A confirmation request has no recipient: the client is a category, not a contact.
- The stated “before” is nobody’s real starting point. A blended baseline is as averaged as the blended result.
- Cross-referencing has no target. Public profiles and review sites need one account to check.
What an honestly disclosed composite looks like
The fix is one sentence. An honest version reads: “This example is a composite drawn from results across several client accounts in the same category,” or “Illustrative, based on aggregated performance.” The undisclosed version instead swaps a company name for a category descriptor and keeps the grammar of a single client story, so the reader prices it as one company, one campaign, one outcome. Disclosure need not name which clients were blended. It only has to say that a blend happened. Confidentiality justifies withholding identities, not withholding the fact that a number is an average.
Advertising guidance is instructive here. The FTC’s endorsement guidance FAQ says that “Statements like ‘Results not typical’ or ‘Individual results may vary’ won’t change that interpretation,” the interpretation being that a claimed specific result reflects what others can expect. Two options remain: proof the results are typical, or clear disclosure of generally expected performance. Whether an equivalent expectation is written down specifically for blended client results is not something that FAQ settles; it does not address composites either way.
Questions to ask when the client is described, not named
Four questions settle it. Listen for whether the answers stay singular.
- “Is this one engagement or a blend of several?” Hard to answer evasively.
- “Which single account produced this number?” A real anonymised client has one; a composite answers with a set.
- “What specific time period does this cover?” One engagement has one window; a blend gives several or a vague general period.
- “Can I speak with this client directly?” Sometimes the no is legitimate under an NDA. It should still be a no about one company, not four.
Read a lot of them side by side
Volume builds this instinct faster than any checklist, sharpening the skill of how to read a case study without being misled and the habit of remembering that every case study library is a selected sample. Browse the case study library, each credited to the agency that published it, or submit your own.
FAQ
Is a composite case study always dishonest?
No. A composite that says so is a legitimate illustrative device; blending real results is not misconduct. The dishonesty lies in presenting a blend as a single checkable account, so the reader treats an average of an unstated sample as one company’s outcome. Disclosed, it is an example. Undisclosed, it is a claim about something that happened to nobody.
How can I tell a composite from a real client whose name was withheld?
Look for detail consistent with a single source. An anonymised real client still has one company behind it, so the write-up carries one campaign window, one specifically described business situation, and quotes that sound like one person. A composite averages those away: the timeframe goes approximate, the niche broadens, quotes turn generic. Ask about a particular month: single-source detail holds, blended detail gets vaguer.
Sources
- FTC, The FTC’s Endorsement Guides: What People Are Asking, fetched 6 September 2026.
