Skip to content

Stockertim Dictionary · Artificial intelligence · Technology

Synthetic data

Also called artificial data, synthetic data set, synthetic dataset

In short

Artificial data generated to resemble real data in structure and statistics without describing real people, often used to test formulas, queries and tools safely.

  • Intermediate
  • Concept
  • Checked

In detail

Synthetic data is made-up data that looks like the real thing. It has the same columns, types and formats as a real data set, and may be built to match its broad statistics, such as averages, ranges and the mix of categories, but each row is invented rather than taken from a real customer, patient or transaction.

There are different ways to make it. The simplest is a few hand-written rows that follow the data dictionary, enough to test a formula. More advanced methods use statistical models or machine learning trained on the real data to generate many rows with similar patterns, which can support testing, training and sharing.

For everyday analysis with an AI assistant, synthetic data is a safe way to get help. Instead of pasting real records into a tool, an analyst can share synthetic rows with the same structure, get the formula or query working, and then run it on the real data inside the organisation's own systems.

Synthetic data has limits. Data generated from real records can still leak information about real people if it copies them too closely, so privacy is checked rather than assumed. And conclusions about the real world come from real data: synthetic rows show that a method works, not what the answer is.

Worked example

An analyst at Kite & Co Accountants wants help writing a formula for overdue client invoices without sharing client details.

  1. The analyst writes eight synthetic rows using the data dictionary: invented client names, regions, amounts, due dates and statuses.
  2. Two of the eight rows are deliberately overdue, so the right answer is known in advance.
  3. The assistant writes a formula that flags overdue invoices, and it flags exactly those two rows.
  4. The analyst runs the tested formula on the real invoice file inside the firm's own spreadsheet.

The formula is checked on data where the answer is known, and no real client record is shared with the tool.

In practice

Analysts use synthetic rows to test formulas and queries with AI tools, developers use synthetic data to test software, and some organisations share synthetic versions of data sets for research. Privacy teams assess generated data sets before release.

Watch out

  • Synthetic data generated from real records is not automatically anonymous; it can still reveal real people.
  • Results computed on synthetic data describe the method, not the real world.

Learn it properly

Official sources

Facts last checked on 3 November 2026.

Education, not advice: the Stockertim Dictionary explains what terms mean. It never recommends buying or selling anything.

English