Metrics

While measures are simple aggregations on a single model, metrics encode complex business calculations that might:

• Span multiple models

• Include complex business rules

• Combine multiple measures

• Apply specific filters by default

If measures are single building blocks, measures are a combination of them.

Additionally, metrics can specify more parameters like default filters (like exclude_deleted) and allowed dimensions (e.g. customer_segment, region, order_date), allowing the agent to reliably answer questions like “How does our average revenue per customer compare between regions over time?”. This prevents nonsensical breakdowns (like trying to show "revenue per customer" grouped by individual product SKU).

Examples

• Simple column relations: SUM(assigned_quotes) / SUM(total_quotes) * 100 → quote_acceptance_rate

• Aggregations: SUM(revenue) / COUNT(DISTINCT customer) → average_revenue_per_customer

• Complex metrics: retention_rate → COUNT(CASE WHEN last_order_date >= CURRENT_DATE - INTERVAL '90 days' THEN 1 END) / COUNT(*) * 100

• Cross-model: SUM(orders.amount) / COUNT(DISTINCT customers.customer_id) → revenue_per_customer