Can I leave out certain customers, people or folders?

Yes. Scope is set in the license agreement: you can leave out systems, date ranges, mailboxes, channels, customers or record types. Write each exclusion into that scope before anything is exported; cutting whole systems or years can lower the price.

What do providers say about leaving things out?

Every program below describes scope as something agreed before records move. They differ in how much they say about exclusions. As of October 2026:

Provider	What it says about scope and exclusions
Appen	“You name the systems, teams, workflows and time periods in scope. Nothing outside that list is read.”
Handshake AI	“You decide what data you share. We work with your team to scope what’s relevant and what protections apply.”
License My Data	Scope goes in “a signed exhibit — systems, date ranges and exclusions”, and “Board, legal, finance, HR and M&A can all be excluded.”
micro1	The company and micro1 define the approved scope and boundaries, and the anonymization and redaction requirements, together before the company takes part (micro1)
Miro Advisory	Owners decide which systems and fields are in scope, and “Client information should not be included simply because it can be anonymized.”
Polyshares	Its intake says “Data you hold for your own clients is out of scope.”
Scale AI	“Partnerships can begin with a very limited scope”, with your sign-off on every data package
SimpleClosure (AssetHub)	Its process excludes sensitive employee and health records and anything the seller marks out of scope (SimpleClosure release)
Telegraph Lab (affiliated with this site)	“Your team approves the scope before any work begins”; regulated, privileged or third-party information “requires additional review and may need to be excluded”
Troveo	“Decide what’s in, what’s out.”

None of these statements is a contract. The exclusions you get are the ones written into the agreement you sign.

What exactly can I leave out, and how do I describe it?

Anything you can describe precisely enough for someone else to apply without judgment. “Nothing embarrassing” cannot be applied; “no mail to or from our law firm’s domain” can.

To leave out	Write it as	Watch for
A whole system	“Accounting software: excluded”	Records from it that also travel elsewhere, such as invoices attached to emails
A date range	Start and end dates for each system	Older threads quoted inside newer emails
A person’s mailbox	The mailbox address	That person’s messages still sit in everyone else’s mailboxes
A channel or mail folder	The exact channel, label or folder names	Direct messages and private channels on the same topic
A customer or GC	Company names, email domains, project names or numbers, and whether matching threads come out whole or have the names replaced	Nicknames and project names that identify them without the company name
A record type	The category and where it lives: HR, payroll, benefits, legal, insurance claims	The same records forwarded into ordinary email
A document folder	The folder path	Copies saved in other folders

The last column is where an exclusion can fail, because the record you meant to leave out also exists in a place you did not name. A sample is how you find those places before full delivery.

Does leaving things out lower the price?

It can. Removing a whole system or years of history takes out what providers say they price on. Appen prices by “volume, the number and connectedness of systems, workflow complexity, data quality, and how rare the operational knowledge is”; Troveo says “Value scales with your systems, team size, and years of history”; Miro Advisory lists licensing scope among the factors that set value (Miro Advisory).

Leaving out categories that programs already expect to drop is a different matter. HR files, privileged material and client-held data appear on the providers’ own lists above, and on Frankfurt Kurnit’s October 1, 2026 list of what to consider keeping out: “privileged material, HR and personnel files, payroll and tax records, health and benefits information, union-related communications, customer and partner confidential information, licensed data, and data from regulated systems”.

• If you remove a system that links the others, such as the job software that ties estimates to invoices and email, expect the offer to be re-priced, and ask for the new number in writing.
• If you remove a few mailboxes, channels or one sensitive category, ask whether the price assumed they were in.
• If you add exclusions after a firm offer, the buyer may re-price, because a firm offer rests on a sample of the scope it was given. What determines the price covers the other factors.

My estimator’s emails trash GCs by name. Does that go out?

Only if his mailbox, or threads he is on, are in scope. If the GCs’ names are on the replacement list, de-identification turns them into labels such as Company 01 but keeps what he wrote; if they are not, they stay. Either way, in a local market the project, the date and the dollar figure can still say which GC he meant (how context identifies people and companies). As Frankfurt Kurnit puts it, de-identification “addresses whether a record can be linked to an individual, not whether its content is confidential”.

You have three ways to handle it:

• Leave his mailbox out. His messages to others still sit in their mailboxes, so add a rule that removes threads he wrote, by his address, wherever they are.
• Replace the GCs’ names everywhere. Give the provider the company names and email domains to replace. His opinions stay; the names go.
• Remove the threads that name them. A list of GC names and project names, applied before export, takes those threads out whole.

The second and third options are only as good as the list. Test it first: in Data Licensing Report’s sample tool, add each GC to your own terms with “= company” and read how his threads come out. Separately, documents a GC gave you may be covered by a confidentiality clause in your contract; whether you can license them is a contract question.

What about the lawsuit correspondence and anything our lawyers wrote?

Leave it out entirely. Frankfurt Kurnit lists privileged material first among the categories to consider excluding and warns that a sale “may waive privilege over legal communications”. Telegraph Lab says privileged information “requires additional review and may need to be excluded”, and License My Data names legal among the departments that can be excluded (License My Data). Which old documents are privileged, and what waiver means, is covered in is it legal to sell company emails.

Case material sits in several places, so describe it four ways:

1. The case name or number and the dates it ran.
2. Your outside lawyers’ email domains, so any thread that includes counsel comes out whole.
3. The folders and mailboxes where case files were kept.
4. The settlement agreement and anything sent under it.

How do I write exclusions into the agreement?

In writing, in the license agreement or an exhibit to it, before any export. License My Data calls this “a signed exhibit”; Telegraph Lab’s terms say its separate written agreement “controls scope, rights, authorized recipients, preparation, confidentiality, pricing, delivery, and payment”.

1. Start from your inventory. Frankfurt Kurnit suggests “listing each data set by system, record count, and start date, and marking each as included or excluded”. A data inventory already has those columns.
2. Use exact names. Mailbox addresses, channel names, folder paths, email domains, date ranges and name lists.
3. Say who applies each rule, and when. If you apply it before export, the excluded material never leaves. If the provider applies it, the provider sees that material first.
4. Check a sample against the rules. Corpus says sellers get a redaction report with samples to sign off before transfer (Corpus); ask any provider for the equivalent before full delivery.
5. Cover mistakes. The agreement should say that anything outside scope delivered by error is not licensed and is deleted when you ask.

For example, a hypothetical commercial contractor’s exhibit might read:

System	In scope	Excluded
Microsoft 365 email	All mailboxes, January 2015 to December 2025	HR and payroll mailboxes; the controller’s mailbox; any thread with outside counsel’s domain; threads naming the GCs on Schedule A
Job-management software	Jobs, estimates, change orders, schedules	Customer contact columns; jobs tied to the 2019 lawsuit
Shared drive	Project folders	HR, Legal and Insurance Claims folders
Slack	Dispatch, field and safety channels	Direct messages; the management channel

Run the same rules on your own exports before anyone else applies them. Data Licensing Report’s sample tool shows every item that remains, so you can see whether the rules caught what you meant.