How can we check that a sample does not resemble real people or organisations?

**How can we check that a sample does not resemble real people or organisations?**

You cannot prove this from the outside, and nobody should promise it. Generated data is not guaranteed to be anonymous or free of disclosure risk. What a reseller can do is run inspections and ask the supplier to explain the process.

Practical inspections:

- search names, addresses, account numbers and free-text notes for patterns that look like real identifiers - look for oddly specific narrative details in hardship or dispute notes - check that amounts, dates and rare events are not repeated in a way that suggests copying - confirm identifiers follow a documented synthetic style, such as `LOAN-3391` - compare rare combinations of attributes, since uncommon pairings can stand out

Questions to put to the supplier:

- How was the generated world informed by observed practice, and what was kept at the level of workflow rather than record? - Which fields are free text, and how were they produced? - What disclosure checks do you recommend a buyer run?

MirrorFirm's output is generated and informed by direct observation of how real businesses operate. It is never real business records. Record the answers you receive, keep the sample files you tested, and pass an accurate description to your own customers instead of an assurance.