What Is Synthetic Data?

Synthetic data is artificially created information designed to resemble real-world data. Instead of collecting every example from the real world, organisations can create additional data that has similar characteristics and use it to train or test AI systems.

Imagine training an AI to spot fraudulent transactions but not having enough examples of fraud to work with. Synthetic data could be used to create realistic examples of fraudulent transactions, giving the AI more situations to learn from without those transactions actually having taken place.

Why Use Synthetic Data?

Getting enough useful real-world data can be difficult, expensive or sometimes impossible. There may also be privacy concerns around using information such as medical records, financial details or customer data. Synthetic data can help fill some of those gaps, create examples of unusual situations or allow organisations to work with realistic information without directly using someone’s personal data.

That doesn’t mean synthetic data is automatically accurate or unbiased. If the information used to create it doesn’t properly represent the real world, the artificial version can reproduce those problems too.

The important distinction is that synthetic data is designed to behave like real data without simply being another collection of real-world records.