What we remove before a lab sees anything
Labs want to learn how work gets done, not who did it. Before a dataset is listed, we take out the details that identify people, and you review the result.
What comes out
- People’s namesEmployees, customers, vendors and anyone mentioned in passing.
- Contact detailsEmail addresses, phone numbers and physical addresses.
- IdentifiersAccount numbers, invoice numbers and other IDs that could tie a record to a person or company.
- Out of scope entirelyConsumer personal data, health records tied to people, and payment card data. We don’t buy these, so they never enter a dataset.
How it’s done
Removed details are replaced with consistent placeholders such as [PERSON_1] or [EMAIL_2], so a thread still reads naturally and the same person maps to the same token within a dataset. See a synthetic sample of the result.
Today every dataset is prepared and checked by a person on our team before it is listed. You then review the redacted preview yourself. If you spot anything we missed, we fix it before the listing goes live.
What buyers see, and when
Before a sale, a lab sees a redacted sample only. After a sale, the licensed lab receives the de-identified dataset. Your raw exports are never delivered to anyone.
The limits
De-identification reduces the risk that someone in your data can be identified; it is not a guarantee. Context can still hint at who a person is. That’s why we exclude the most sensitive categories outright, why we check rights with you first, and why you see the preview before any lab does. If a dataset can’t be made safe to a standard you’re comfortable with, it doesn’t get listed.