# Data Graph


{{< customreadfile "/includes/audiences/private-beta-callout.md" >}}

The **Data Graph** is an entity mapping layer that maps your warehouse tables to entities and events and defines the relationships between them. This gives marketers a business-friendly interface to build audiences from. Your data team configures it once, pointing at warehouse tables or your own pipelines, and marketers self-serve from there.

{{< image src="images/audiences/data-graph/audiences-workflow-data-graph.svg" alt="Data Graph Overview" width="70%" class="mx-auto d-block" >}}

## Core concepts

A data graph has three key components: **entities**, **events**, and **relationships**.

| Term | <div style="width: 350px;">Description</div> |
|---|---|
| Entity | A warehouse table that represents a business object, for example, **Customers**, **Accounts**, or **Products**. Entities are the "nouns" marketers build audiences around. |
| Event | A warehouse table with a timestamp column that captures something that happened, for example, `product_viewed`, `order_placed`, or `cart_abandoned`. You can filter events with a time window in the Audience Builder. |
| Relationship | A `1:many`, `many:1`, or `1:1` link between two entities, or between an entity and an event. Relationships let marketers filter one entity using conditions on related records, for example, "customers with 3 or more orders". |

## Create a data graph

You can create a data graph in two ways:

{{< success >}}
Both methods produce the same data graph. Pick whichever best fits your team's workflow.
{{< /success >}}

### 1. Visual Builder

Configure the data graph directly in the RudderStack dashboard using the visual builder. Pick warehouse tables, mark them as entities or events, and draw relationships between them. **This is the recommended way to build a data graph**.

See [How to Create a Data Graph using Visual Builder]({{< ref "audiences/data-graph/create-data-graph.md" >}}) for a step-by-step guide.

### 2. Rudder CLI

Define your data graph as a YAML file and sync it to your workspace using the [Rudder CLI]({{< ref "/dev-tools/rudder-cli/" >}}). This is the right choice for teams that want to version-control the data graph in Git, review changes via pull requests, or manage multiple environments (dev, staging, prod) as code.

See the [Data Graph YAML reference]({{< ref "/dev-tools/rudder-cli/data-graph.md" >}}) for the full schema and a complete example.

### Prerequisites

You need a connected **warehouse account** before you can build a data graph.

The following warehouses are supported:

- Snowflake
- BigQuery
- Redshift
- Databricks

See [Configure a warehouse connection]({{< ref "/audiences/data-graph/create-data-graph.md#prerequisites" >}}) for detailed steps on configuring a warehouse connection.

## Current limitations

The data graph has the following limitations in the current release:

- Direct many-to-many relationships between entities is not supported
- Event-to-event relationships are not supported

## See also

- [How to Create an Audience from a Data Graph]({{< ref "audiences/audience-builder/create-audience.md" >}}): Build and activate audiences by leveraging the data graph