dbt Model Index

When to Use

• Before writing any BigQuery SQL against production data
• When the task has not already explicitly stated which models/tables to query
• When resolving a vague or ambiguous data question into the right BigQuery tables

How to Set Up This Skill

• Organize models into logical domain sections (e.g., Users, Activity, Revenue, Events)
• For each model, include: the table name, a 1–2 sentence description of its grain and content, and "Useful for:" bullets covering common query patterns
• Note key join keys, standard filters, and partition fields where relevant

[Domain: e.g., Users & Identity]

your_model_name

• [Type of question this model answers, e.g., user counts, cohort sizes]
• [Another use case, e.g., filtering to a specific user segment]
• [Common join pattern, e.g., joining to other tables as the canonical user dimension]

another_model_name

[Domain: e.g., Activity & Engagement]

your_activity_model

• [Use case 1, e.g., daily/weekly active user metrics]
• [Use case 2, e.g., retention analysis]

your_engagement_model

• [Use case 1]
• [Use case 2]

[Domain: e.g., Revenue & Subscriptions]

your_revenue_model

• [Use case 1, e.g., MRR/ARR reporting]
• [Use case 2, e.g., churn analysis]

your_subscription_model

• [Use case 1]
• [Use case 2]

[Domain: e.g., Events & Telemetry]

your_events_model

• [Use case 1, e.g., raw event-level analysis]
• [Use case 2, e.g., building domain-specific funnels]

Important Notes

• Standard filters: Document any filters that should always be applied in user-facing queries (e.g., excluding test accounts, soft-deleted records, internal users, or flagged/fraudulent users). Example:

  where not is_internal_user

• Production data: Specify your default project/dataset path. Example:

  your-gcp-project.prod.<model_name>

• Cost control: For large partitioned tables, always filter on the partition field and constrain the date range to avoid full-table scans
• Model grain: Always note the grain (one row per what?) for each model to avoid accidental fan-outs in joins
• Plan/tier types: If your product has subscription tiers or plan types, document the valid values here so queries filter correctly
• Sensitive datasets: If any models live in a separate dataset, call that out explicitly so queries use the right fully-qualified table reference