Judging a talent agency case study: what it should actually prove before you sign

The four situations below are constructed, carry no figures, and report nobody's results. A result from somebody else's campaign is not evidence about yours: you cannot see their brief, product, timing, or what else they were running that month.

What transfers is the anatomy of the decision. Each one is written as what was asked, what was actually being decided, and what changed afterwards. Read each for the clause it turned on, not for the outcome.

What to take away

What a case study has to prove before you sign

One: the ambassador deal that turned on a definition

Asked: a year-long partnership with one creator, exclusive in the brand's category. The deal turned on one word inside that category definition, and terms like indemnity carry a settled legal meaning that a summary in a deck will not convey.

Actually being decided: who writes the category definition. The brand meant its own narrow product type. The agency read the drafted wording as covering everything adjacent, and priced the year accordingly, since it removed most of the creator's likely income.

What changed: the definition moved from a category to a named list of competitors, reviewed each quarter. The fee came down because the block was smaller, and both sides could tell whether a given approach was allowed without asking a lawyer. The lesson generalises: an exclusivity clause is a description of what the creator may not earn from, so vagueness in it always costs the buyer money.

Two: the roster package and the right of substitution

Asked: a package deal across several creators from one agency, at a better rate than booking each individually.

Actually being decided: whether the brand was buying named people or a quantity of reach. The draft allowed the agency to substitute a creator of comparable size, which is what made the discount possible. Why an agency defends that flexibility so hard is set out in the pillar guide to how talent representation works.

What changed: the brand split the roster into two names it would not accept substitutes for and a group where substitution was fine, and paid different rates for the two halves. Package pricing is not a discount on the same thing. It is a different product with flexibility built in for the seller, and it is worth having when the campaign is about volume and worth refusing when it is about a particular person.

Three: the conflict discovered late

Asked: nothing, at first. The campaign was signed and in production when someone noticed the creator had an existing arrangement in an adjacent category.

Actually being decided: whether the brand had ever asked. It had not, and the contract contained no warranty covering existing commitments, so there was no breach to point at.

What changed: a single question was added to the standard enquiry, asking the agency to confirm existing commitments that would sit alongside the work, and a warranty went into the template. The wider pattern is that conflicts are cheap to ask about and expensive to discover. Asking what a creator is already committed to belongs with the same pre-contract diligence as checking that an account is what it appears to be, which is covered in the guide to verification tooling.

Four: the renewal that cost more than the original

Asked: permission to keep running a piece of content that was performing, some months after the campaign ended.

Actually being decided: whether the brand was negotiating from a position of choice. It was not. The content was already running, the license had lapsed, and the agency knew both facts. A renewal negotiated under those conditions is not a negotiation.

What changed: license end dates were attached to the asset files rather than left in the contracts, with a reminder before expiry, and the option to extend was priced at the time of the original deal instead.

Where the intention from the start is to re-cut and re-run material as your own advertising, that is a different license again. The seam between a contract system and an asset library that causes this failure is set out in the guide to campaign workflow tools.

Why published case studies rarely help

Four problems, and they apply to every case study anyone shows you.

• Survivorship. Nobody publishes the campaign that was canceled in week two, so the sample you see is the successful end of a distribution whose shape you cannot see. Survivorship bias is the whole of the problem with this genre.

Selection by the vendor. The agency decides which campaigns appear in its own deck, and it publishes the ones that make the methods it wants to sell look sound. No baseline. A published case rarely shows what the same budget did beforehand, or what a comparable campaign looked like, so the result has nothing to be measured against. No date and no scope. A result without the month, the market, the spend and the deliverables attached cannot be checked and cannot be compared with your campaign.

That does not make them worthless. It makes the results section worthless. Read the mechanics instead: deal structure, who approved what, rights coverage, how long it took.

Those parts are usually accurate because nobody thought to embellish them. Your own version of that record, including the memo on what was agreed, sits in our worked examples of talent agency deals.

Three questions to ask about any case shown to you

An agency that answers all three has told you more about how it works than any results slide. One that cannot describe a single change of plan is describing a campaign it did not run closely.

Common questions

Should we write our own case studies internally?

Yes, and keep them short and unflattering. One page per campaign: what we asked for, what we agreed, what we changed mid-flight, what we would write differently next time. Nobody outside needs to see it.

Can we ask an agency for a reference from a former client?

You can, and you will be given a happy one. The more useful request is for a client who stopped working with them, which some agencies will actually provide, and the willingness to discuss it is itself informative.

Do numbers ever belong in a case study?

Inside your own business, yes, with the definitions attached and the caveats intact. Published outside, they invite a comparison nobody can make honestly, since the conditions never match.

How many of these should we keep?

Enough to see a pattern, which is usually four or five. The point is to notice that the same clause keeps causing the same argument, and one campaign cannot show you that.