> ## Documentation Index
> Fetch the complete documentation index at: https://docs.bondata.ai/llms.txt
> Use this file to discover all available pages before exploring further.

# Data Normalization

> Use an LLM to clean, standardize, and transform field values

Uses an LLM to normalize and transform field values. Ideal for cleaning messy data - standardizing addresses, formatting names, converting units, or restructuring free-text fields.

<Frame caption="Data Normalization node with input-to-output field mappings">
  <img src="https://mintcdn.com/bondata/DlMd45I1zc1MvbNc/images/guide/nodes/transform/data-normalization.png?fit=max&auto=format&n=DlMd45I1zc1MvbNc&q=85&s=88e2088b9f7960341a9516aceda40e0d" width="720" height="598" data-path="images/guide/nodes/transform/data-normalization.png" />
</Frame>

## Configuration

| Setting             | Description                                                                                        |
| ------------------- | -------------------------------------------------------------------------------------------------- |
| **LLM Integration** | The AI provider to use (e.g., Gemini, OpenAI, Anthropic)                                           |
| **Field Mappings**  | Connect input fields on the left to output fields on the right. Draw lines to define relationships |

## How It Works

<Steps>
  <Step title="Define input fields">
    Select the fields from your data that need normalization (e.g., `BillingStreet`, `BillingCity`).
  </Step>

  <Step title="Map to output fields">
    Map each input to a normalized output field (e.g., `BillingStreet_Normalized`).
  </Step>

  <Step title="LLM processes records">
    The LLM processes each record and writes cleaned values to the output fields.
  </Step>
</Steps>

## Output

Normalized output fields become available as **Mentions** in all downstream nodes.

## Example

Standardize addresses across a Salesforce Account dataset:

1. Map `BillingStreet`, `BillingCity`, `BillingState` as inputs
2. Create normalized output fields for each
3. The LLM cleans formatting inconsistencies (e.g., "123 Main St." → "123 Main Street")

## Best Practices

* Use specific, descriptive output field names to avoid confusion downstream
* Test with a small batch first to verify the LLM produces the expected format
* Choose the right LLM - faster models for simple formatting, stronger models for complex transformations

## Related Nodes

* **[AI Enrichment](/guide/nodes/enrichment/ai-enrichment-node)** - uses an LLM prompt to generate new data rather than cleaning existing fields
* **[Transform](/guide/nodes/transform/transform-node)** - applies rule-based transformations without an LLM
* **[Regex Pattern](/guide/nodes/transform/regex-pattern-node)** - extracts or replaces text using regular expressions