DBT Semantic Layer | Workshop

Surfalytics community workshop on the dbt Semantic Layer: what a semantic layer is, how it differs from metric logic inside each BI tool, and how to build and consume metrics in dbt Cloud (YAML, Metric Flow, time spine, Google Sheets). 📚 What you'll learn Theory • What a semantic layer is and why teams get different numbers for the same KPI • BI-native layers (Tableau, Power BI) vs code (LookML) vs external layers (dbt, Cube, AtScale) • dbt history: dbt Metrics → Transform / Metric Flow; semantic models, metrics, saved queries, time spine • dbt Core vs dbt Cloud: full Semantic Layer and integrations in Cloud; Metric Flow on Core is engine-only without native consumer integrations • Governance: metric owners, review process, warehouse cost of generated SQL • When to adopt a semantic layer vs staying in models or BI only (hybrid approaches) Hands-on (dbt Cloud) • Project layout (staging → intermediate → marts) and enabling the semantic layer • Orders semantic model: entities (order_id, location_id, customer_id), time dimension, categorical dimensions • Metrics: revenue (sum), orders count; `dbt parse` and `dbt sl list metrics` • Time spine daily model and YAML config (BigQuery syntax) • Locations semantic model to slice revenue and order count by location name • Google Sheets add-on: connect, query builder, time grain and filters • Troubleshooting: production job with new SL syntax vs standard dbt run (dbt Fusion / future dbt 1.12) • Mention of Open Semantic Interchange and Q&A (Power BI, Snowflake semantic views for agents) ⏱️ Timestamps 0:00 – Intro and agenda 0:33 – Title slide 0:40 – What is a semantic layer? 2:08 – Benefits: one definition, BI connectivity 2:35 – LookML, BI UI layers, external services 4:23 – dbt SL and multi-tool integration 5:04 – Architecture: with vs without semantic layer 6:28 – dbt, Transform, Metric Flow 8:33 – Semantic models, metrics, saved queries, time spine 9:31 – dbt Core vs Cloud for SL 11:32 – Entities, dimensions, metrics 13:10 – Integrations and API 13:45 – Tradeoffs and cost 15:46 – When to use SL / hybrid 17:42 – Demo start 18:06 – Project tour 20:41 – Orders YAML and semantic model 22:10 – Time dimension and entities 26:27 – Dimensions 29:51 – Revenue and orders_count metrics 34:48 – Time spine concept 36:28 – timespine_daily (BigQuery) 41:08 – dbt sl list metrics 43:24 – Locations model for joins 47:53 – Google Sheets integration 48:40 – PR, merge, production job 56:00 – Fusion / syntax and failed job fix 59:38 – Open Semantic Interchange 1:01:40 – Sheets query demo 1:04:25 – Recap 1:06:13 – Q&A 📌 Resources mentioned • dbt Semantic Layer / Metric Flow (dbt Cloud) • dbt documentation and quick start (semantic models, time spine) • Google Sheets Semantic Layer add-on • Open Semantic Interchange (industry initiative) • Snowflake semantic views (webinar / events) ✅ Key takeaways • A semantic layer centralizes metric definitions so dashboards agree on logic. • dbt Cloud exposes the Semantic Layer to many tools; Core + Metric Flow alone does not replace hosted integrations without custom work. • You need a time spine for time-based metrics in the dbt SL model described in the demo. • New YAML/spec may require dbt Fusion or a new dbt version until core catches up—validate jobs after upgrades. • Many teams use a hybrid: critical company metrics in the SL, other metrics in BI tools. --- Surfalytics – Community workshop. dbt Semantic Layer: Concepts and Hands-On in dbt Cloud. #Surfalytics #dbt #dbtSemanticLayer #MetricFlow #dbtCloud #DataEngineering #AnalyticsEngineering #SemanticLayer #MetricsLayer #BI #GoogleSheets #DataModeling #Workshop