dbt Labs

Your AI is only as good as your data. Without a semantic layer, AI systems are left guessing. That leads to: 🚨 Misinterpreted metrics 🚨 Conflicting business definitions 🚨 Security risks and compliance gaps To succeed, AI needs: 𝗖𝗼𝗻𝘀𝗶𝘀𝘁𝗲𝗻𝗰𝘆 – One definition of “revenue,” not five. A semantic layer ensures AI always queries the right business logic. 𝗚𝗼𝘃𝗲𝗿𝗻𝗮𝗻𝗰𝗲 – Sensitive data stays protected, access is controlled, and changes don’t break everything downstream. 𝗖𝗼𝗻𝘁𝗲𝘅𝘁 – AI needs more than raw data; it needs relationships, metadata, and logic to understand the "why" behind the numbers. 𝗦𝗰𝗮𝗹𝗮𝗯𝗶𝗹𝗶𝘁𝘆 – Precomputed, validated metrics make AI-powered insights faster and more reliable—so teams actually trust them. Check out the blog (link in comments) to learn more.

This week on The Analytics Engineering Podcast, Tristan Handy sits down with Daniel Avancini, Chief Data Officer and Co-founder of Indicium—a fast-growing data consultancy started in Brazil. What sets Indicium apart is their HR model. Instead of focusing on experienced hires, they’ve built a talent pipeline through courses and an internal onboarding program, training new employees from zero to 60 in just a few months. The result? Exceptional client outcomes and hundreds of new data professionals gaining the skills to succeed in a field that’s notoriously hard to break into. Companies investing in scalable hiring and training for analytical talent are shaping the future of data. Indicium is one of them. 🎧 Listen to their full conversation here https://lnkd.in/g5MX2k6b or wherever you get your podcasts.

We’ve been named one of the top 100 midsized companies to work for 🎉 Thank you to Built In for the recognition—and to our amazing team at dbt Labs for building a culture that’s as impactful as the work we do. And we’re just getting started: our team is growing globally. If you’re ready to shape the future of data, check out our latest openings https://lnkd.in/gt2vdiB7

What’s new in dbt Cloud? (spoiler alert: A LOT) 🚀 We’re only a few weeks into 2025, and we’ve added a bunch of new dbt Cloud features to help you ship trusted data faster, optimize costs, and collaborate across teams. 👉 See what’s new and read the blog (link in comments) for the full breakdown.

🧪 Test smarter, not harder. We’re wrapping up our series on the four layers of data testing with the marts layer. At this layer, focus on net-new columns and high-priority business rules, following the same hygiene-or-anomaly approach as staging and intermediate layers. What to test in the marts layer: 𝗨𝗻𝗶𝘁 𝘁𝗲𝘀𝘁𝘀  Validate complex transformation logic, such as: 📅 Calculating dates for forecasting models. 🎯 Customer segmentation logic with multiple CASE-WHEN statements. 𝗣𝗿𝗶𝗺𝗮𝗿𝘆 𝗸𝗲𝘆 𝘁𝗲𝘀𝘁𝘀 🔑 Apply to models where your mart’s granularity has changed. 🧩 For enriched models (even without a grain change), primary key tests communicate your intent to future developers. 𝗕𝘂𝘀𝗶𝗻𝗲𝘀𝘀-𝗳𝗼𝗰𝘂𝘀𝗲𝗱 𝗮𝗻𝗼𝗺𝗮𝗹𝘆 𝘁𝗲𝘀𝘁𝘀 Focus on critical fields and specific business problems: 🧠 Detect fuzzy matches, such as a user with multiple emails extending free trials. 📊 Ensure calculated fields don’t vary beyond a set percentage in a week. 💲 Verify ledger tables follow rules, like today’s total always exceeding yesterday’s. Testing at the marts layer keeps your data aligned with business rules and ensures high integrity, even for complex scenarios.

Because at the end of the day, it’s all about those dashboards that drive business impact 🫶

Wow, we’ve got a busy month coming up with five in-person dbt Meetups scheduled. So if you’re looking for opportunities to learn with fellow members of the dbt Community, and have fun while doing so, join us at one of the sessions listed below: 🇪🇸 Barcelona | Thursday, February 13th, organized by Spaulding Ridge 🇹🇼 Taipei | Wednesday, February 19th, organized by community members Karen Hsieh, Laurence Chen, Allen Wang, and LI KUAN LIAO 🇧🇪 Belgium | Thursday, February 20th, organized by community members Sam Debruyn and Lise Kerckhove 🇺🇸 Chicago | Thursday, February 20th, organized by Analytics8 | Data & Analytics Consultancy 🇯🇵 Tokyo | Friday, February 21st, organized by community member Shinya Takimoto dbt Meetups are gatherings dedicated to helping you own your analytics engineering workflow. RSVP and tag a friend in the comments to join you 👯♂️ https://lnkd.in/ekknesFN

🧪 Test smarter, not harder. Our series on the four layers of testing continues with the intermediate layer. At this stage, focus on data hygiene and anomaly tests for new columns. Skip re-testing passthrough columns from sources or staging to keep things efficient. What to test in the intermediate layer: 𝗣𝗿𝗶𝗺𝗮𝗿𝘆 𝗸𝗲𝘆 𝘁𝗲𝘀𝘁𝘀 🔑 Add a primary key test to re-grained models. 🧩 Use primary key tests for enriched models (even if the grain stays the same) to help future developers understand your logic. 𝗔𝗻𝗼𝗺𝗮𝗹𝘆 𝘁𝗲𝘀𝘁𝘀 𝗳𝗼𝗿 𝗷𝗼𝗶𝗻𝘀 𝗮𝗻𝗱 𝗮𝗴𝗴𝗿𝗲𝗴𝗮𝘁𝗶𝗼𝗻𝘀 🎨 Categorical values: Use an accepted_values test for new categorical columns. 🔀 Range relations: Add a mutually_exclusive_ranges test for columns with related ranges (e.g., non-overlapping age ranges). 🔄 Dynamic values: Apply a not_constant test for columns expected to vary over time (e.g., page view counts in website analytics). 𝗖𝗼𝗺𝗽𝗹𝗲𝘅 𝗼𝗽𝗲𝗿𝗮𝘁𝗶𝗼𝗻𝘀  🛠️ For models isolating complex operations, the tests above are often sufficient. 🔍 Consider unit testing particularly complex SQL logic to ensure clarity and reliability. Testing smarter at the intermediate layer keeps your models clean, consistent, and ready for downstream marts.

“Work before dbt was a rollercoaster. We had multiple places where transformation was happening.” Now, with 1,400+ models seamlessly migrated and maintained in dbt Cloud, Enpal has transformed their workflows. They’ve achieved 70% monthly cost savings and a 30x speed increase for their heaviest queries—all while scaling to meet rapid growth. See how they built a modern data stack that scales: https://lnkd.in/gzwyEXd5

The Analytics Development Lifecycle (ADLC) simplifies the analytics workflow so you can focus on delivering trusted insights. If you’re constantly battling metric discrepancies across the tools in your data stack, the 𝗔𝗻𝗮𝗹𝘆𝘇𝗲 phase of the ADLC fixes that. With the dbt Semantic Layer, you can define metrics once and access them everywhere—from dashboards to spreadsheets to LLMs. Semantic models align engineers, analysts, and business stakeholders with a single source of truth. 👾 Lean on dbt Copilot to take the manual work out of creating semantic models. 🔍 Visualize models as part of your overall lineage to trace dependencies from data source to dashboard. ⚡ Built-in caching means less time waiting and more time delivering insights. Join the Data Leaders Panel tomorrow to hear dbt Labs CEO Tristan Handy and Head of Data Alex Welch discuss the ADLC with data leaders from Cox Automotive Inc. and Salesforce. Register now: https://lnkd.in/gBFaY2JG