The short answer
Your master services agreement decides what you can license. If the IP clause assigns only deliverables and the confidentiality clause lists named categories, your process records are probably yours. If either clause reaches for everything connected to the project, assume you need the client's consent.
On this page
Every agency founder who looks at this market has the same first thought: we have ten years of work on the server. The second thought should be: whose work is it? The answer sits in a document most founders signed once and never reread.
Why the MSA comes before the archive
A data buyer will ask you to warrant that you own what you license, or hold the right to license it. That warranty is yours. If a client later shows its material was in the dataset, the buyer points at the warranty and you are the one explaining.
So the order of work is fixed. Read the contracts, decide what is clearly yours, and only then look at what your project tools hold. Founders who start from the archive end up arguing themselves into records that were never theirs.
The three clauses that matter
IP assignment
This is the clause that moves ownership to the client, usually on payment. The question is what it moves. Compare two common drafts:
- “The Agency assigns to the Client all rights in the Deliverables.”
- “The Agency assigns to the Client all rights in all materials created in the course of performing the Services.”
The first leaves your timesheets, internal notes and working process with you. The second can sweep in drafts, internal briefs, even the revision log, because they were “created in the course of” the work. Many agencies signed the second without noticing, often on a client’s paper rather than their own.
Confidential information
Even where you own a record, you may have promised not to disclose what is in it. A definition that lists categories (pricing, product plans, customer data) is workable: strip those out and the rest is yours to use. A definition that covers “all information relating to the Client or the Project” is not. Under that wording, a brief with the client’s name removed may still be disclosure.
Pre-existing materials
Good agency contracts keep the agency’s own tools, templates and methods with the agency, often with a license to the client to use them inside the deliverables. This clause is your friend. It confirms that SOPs, estimating templates and onboarding guides are yours, and those are exactly the documents buyers such as micro1 list first.
A quick sort
| Record | Usually | Why |
|---|---|---|
| Final files and source files | Client | Assigned on payment |
| The client’s own brief and research | Client | Their material, and confidential |
| Your internal brief written from theirs | Depends | Yours only once nothing traces back |
| Timesheets and resourcing | Agency | Operating record; remove staff names |
| Revision log and QA notes | Depends | Check the assignment wording |
| SOPs and templates | Agency | Usually covered by pre-existing materials |
What the buyers say about ownership after you license
This matters as much as what you own going in. Going by their own pages on 7 November 2026: micro1 leaves ownership of the records with the company; Labelbox Prism leaves the seller in control, though its licenses start out exclusive; Nyne hands the question to the contract.
Exclusivity is the trap here. If you grant one buyer an exclusive license to your revision history, those records are off the market for everyone else until the grant ends. Read the grant: scope, field of use, term, exclusivity and what happens to derived models when the license ends.
When to ask the client
Ask when the confidentiality definition is broad, when the assignment covers “materials” rather than “deliverables”, or when the records are from a single large account whose work is recognizable even without the name. A short, specific request (which record types, how they will be de-identified, that no deliverables are included) gets a yes more often than founders expect. A vague one gets legal involved.
Where you cannot get consent, leave that client’s records out. The archive of a 60-person agency rarely depends on one account.
Personal details are a separate problem
Owning a timesheet does not make the names on it licensable. Staff names, client contacts in email threads and anyone’s personal information should come out before a buyer sees a sample, even though every buyer we read says it de-identifies. Their process is the second line; yours is the first.
Older contracts and statements of work
Agencies that have run for a decade usually have three or four generations of contract template, plus clients who insisted on their own paper. Do not assume the current template governs the old work. Pull the MSA and any statement of work for each account you plan to include, because a statement of work can override the master terms for a single project. Where you cannot find the signed contract at all, treat that account’s records as off limits until you can.
Where this leaves you
If you want the wider picture first (which records buyers want, their published terms and what deals have been reported), the front page on how to sell agency data covers it in one read. The clause-by-clause checklist is on client-owned vs agency-owned.
None of this replaces your own lawyer, who should read the license and your largest MSAs together. Bring them the three clauses above, already highlighted, and the meeting gets much shorter.