Models¶
Models represent the core business entities in your organization, such as customers, orders, support tickets or transactions. Each Model is built from a warehouse table or a SQL query on one data connection, and defines how AI Analysts can query and understand that data.
Availability
- Who can do this: members with the Edit capability in the active Workspace.
Models belong to a Workspace. Each Workspace has its own set of Models, and a Model in one Workspace is not visible from another.
To create a Model, see Creating Models.
The Model page¶
In Studio, click Models in the sidebar and select a Model from the list. The Model page has these tabs:
| Tab | What you do there |
|---|---|
| Overview | See the data source; edit the description, guidance and Tags |
| Dimensions | Define the attributes you can filter and group by |
| Measures | Define the numbers you can aggregate |
| Filters | Define reusable conditions that narrow the data |
| Relationships | Define how this Model joins to other Models |
The buttons at the top right of the page are Version history, Duplicate and Delete model. See Managing a Model.
When you change the name, description, guidance or Tags, an Unsaved changes bar appears. Click Save changes to save, or leave the page and choose Discard changes.
Model name¶
The identifier for this business entity. Use clear, descriptive names that match how your team talks about the data (for example customers, orders, products).
Model names must be lowercase snake_case: they start with a letter and use only lowercase letters, digits and underscores (for example orders, line_items). Each name must be unique among the Models in your Workspace. To rename a Model, click its name at the top of the page, type the new name and click Save changes.
Data source¶
Every Model is built from a warehouse table or a SQL query on one data connection. The Overview tab shows it under Data source:
- Table-based Models show the connection name (a link to the connection) followed by the table's full warehouse path, for example
analytics / public.orders. Depending on your warehouse, the path can include a catalog or project, such ascatalog.schema.tableorproject.dataset.table. - SQL-based Models show the connection name and a SQL button. Click SQL to see the Source SQL query.
The data source is fixed after the Model is created. To build on a different table or query, create a new Model.
Next to the data source:
- Preview opens a dialog with sample rows from the table or query.
- The modeled indicator (for example 2/3 modeled) shows how many source columns are referenced by at least one dimension or measure expression. Click it to open Column coverage, which lists each column with the dimensions and measures that use it, and marks the rest as Not modeled.
Description¶
A brief explanation of what this Model represents. It is shown to people browsing the semantic layer and helps AI Analysts understand what data is available.
Example: "Customer accounts and contact information for the organization"
Keep it concise: one or two sentences explaining what each row represents.
Guidance¶
Rules for the AI Analyst: when to use this Model and how to query it. Use this field to explain:
- Special business rules or constraints
- How the data should be interpreted
- When to use (or avoid) this Model
- Important context about the data
Example: "Use this model to analyze customer behavior and segmentation. Customers can be either B2B or B2C, with different pricing tiers for each type."
Tags¶
Labels that control which AI Analysts can use this Model. Type a Tag in Add tag… and press Enter (or type a comma), then click Save changes.
Tags matter for access: an AI Analyst can be configured to use every Model that carries certain Tags, or every Model except those that carry certain Tags. Tag-based access updates automatically, so adding or removing a Tag on a Model changes which of those AI Analysts can use it. See Tags and Models for an AI Analyst.
Model components¶
Once you've created a Model, you can define four types of components.
Dimensions¶
The attributes you can filter, group or analyze by. These are typically the columns in your table.
Measures¶
The numerical values you want to aggregate: counts, sums, averages. Measures always use aggregation functions.
Filters¶
Predefined conditions that narrow your data. Filters help AI Analysts query the right subset of records.
Relationships¶
Connections to other Models that let AI Analysts query across multiple business entities.
Learn more about Relationships →
Managing a Model¶
View and restore earlier versions¶
Click Version history at the top of the Model page to see earlier versions of the Model. Click View changes to compare a version with the current one, or Restore to bring it back. Restoring records a new version that matches the one you selected; your current version stays in the history. See Version history.
Duplicate a Model¶
Click Duplicate to copy the Model, including its dimensions, measures, filters, Tags and relationships to other Models. The copy gets an unused name based on the original and opens straight away, so you can rename it. Relationships from the Model to itself are not copied.
Availability
- Plan: Duplicate appears only when your organization's plan includes copying Models.
Delete a Model¶
- Click Delete model (the trash icon).
- In Delete model?, click Delete.
The Model disappears from the Models list. Its name becomes free, so you can create a new Model with the same name.
Warning
Before you delete a Model, review the Metrics, relationships and AI Analyst access that use it.
The confirmation dialog says you can restore the Model later. Studio doesn't currently provide a control for restoring a deleted Model, and restoring a version doesn't bring a deleted item back (see Version history). If you need the Model again, create it again.
Best practices¶
- Start with the essential dimensions and measures and add more later.
- Write descriptions for someone unfamiliar with your data warehouse, without internal jargon or abbreviations.
- In guidance, anticipate the questions AI Analysts will be asked and give the context they need to answer correctly.
- Use the modeled indicator to spot source columns that no dimension or measure uses yet.