Generate synthetic data

  • Release version: Australia
  • Updated May 7, 2026
  • 3 minutes to read

    • Create synthetic data using the Standard data generator in Now Assist Data Kit. Use synthetic data to imitate real-world records so you can run evaluations or create training for a test model without using production data.

    Before you begin

    Role required: sn_data_kit.admin

    Procedure

    1. Navigate to All > Now Assist Data Kit > Home.
    2. On the Synthetic datasets tab, select Generate dataset.
      The Select a generator type dialog opens.
    3. Select Standard data generator.
      Note:
      A Multi-table data generator (beta) option is also available. This generator creates test data across multiple related tables simultaneously, maintaining referential integrity. As a beta feature, it may behave unexpectedly.
    4. On the Define dataset page, fill in the fields and select Next.
      Table 1. Define dataset fields
      Field Description
      Name Name for the synthetic dataset.
      Choose a template (optional) A pre-built configuration for a common data type. Selecting a template pre-fills the industry, data description, and category fields. Available templates include Catalog item and Incident data.
      Describe department or industry The domain area, such as healthcare or finance that the data pertains to. This specification helps generate high-quality data. Find sample descriptions in the "Get help" section.
      Describe the data you want to generate Detailed description to help generate a wide range of relevant keywords, resulting in higher quality data.
      Add categories to generate a diverse dataset Keywords that guide the model to generate relevant data. For example, “software issue”, “laptop”, “password reset”, “payroll”, “benefits”, etc.
      Select language to generate data in The language for the generated records. Defaults to English.
    5. Optional: On the Select sample data page, add sample records to help the model understand the expected structure, format, and tone of the generated data, then select Next.

      You can select up to 3 records from an instance table or an existing data collection. For detailed steps, see Select the sample data.

      Note:
      Sample data is not replicated in the output. It is used only to guide generation.
    6. On the Define columns page, review and update the column definitions, then select Next.

      Each column requires a label, a data type, and a description of the data to generate. You can use Generate column descriptions to auto-populate descriptions, add new columns, or delete all columns and start fresh. For detailed steps, see Define columns to generate data.

      Note:
      If you selected sample data in the previous step, column definitions are pre-populated based on the columns you chose.
    7. On the Test and generate data page, configure the generation settings.
      Table 2. Test and generate data fields
      Field Description
      Number of records to generate The total number of records to generate. Defaults to 20.
      Additional rules (optional) Extra constraints or instructions to guide the data generation.
    8. Optional: Select Generate preview to review a 5-record sample before committing to full generation.
    9. Select Review summary.
      A summary of all dataset settings appears. Select Edit next to any section to make changes before generating.
    10. Select Start generation.
      The generation job starts and a progress page opens. You can leave this page and return later. The records appear on the Synthetic datasets tab when generation is complete.

    What to do next

    When generation completes, the dataset detail page shows three tabs:

    • Output data: The generated records. Review these for accuracy. A disclaimer reminds you that AI-generated content should be verified before use.
    • Data insights: A quality score (0–100) and a breakdown of metrics including data volume, similarity to sample data, data hygiene, missing or empty values, and temporal consistency. A score of 80 or higher indicates the dataset is ready to use. For more information, see View data insights.
    • Input settings: A record of the data definition, sample data, and column definitions used to generate the dataset.

    Select Save as Template to save the generator configuration for reuse, or Add to data asset to make the dataset available for use in data collections.