Redstrata Knowledge Base

Fivetran + dbt Labs

7 min read

Fivetran + dbt Labs

Type: company (merged data-movement + transformation vendor; the dbt framework’s steward) Sector: Data pipeline, transformation, semantic-layer and agent-context infrastructure — the “pipeline/transform with governance relevance” line of the DG/DM vendor watchlist First seen: 2026-09-09 (as dbt, an Apache Ossie contributor) Last updated: 2026-09-18

Created 2026-09-17 from S-2026-09-16-dbt-summit-2026-announcements (daily data-governance vendor-intelligence scan). dbt was already cited in three vault sources — as an Apache Ossie contributor and Semantic Layer consumer S-2026-09-09-ataccama-apache-ossie-converter, as a lineage source Qlik’s metadata bridges read S-2026-08-24-qlik-idc-marketscape, and in the Alation IDC note’s Ossie cross-reference S-2026-09-04-alation-idc-marketscape — which triggers a company page per schema §2.3. Fivetran and dbt Labs completed their merger on 1 June 2026 (release seen in the 15 Sep scan; not separately ingested — see Open Questions). Capability and performance claims are vendor-reported and not independently verified. Updated 2026-09-18 based on S-2026-09-18-weekly-vendor-synthesis (weekly vendor-synthesis): cross-week DG/DM read — dbt Summit’s five-product wave is this week’s single largest DG/DM release and, with Informatica’s MCP expansion the same week, reinforces MCP/open-context-standard exposure as the dominant theme of the week. dbt State’s per-model lag_tolerance and dbt Wizard’s pre-ship impact validation are the most concrete BCBS 239-style timeliness/change-control primitives seen from any DG/DM vendor this month — but, as with every MCP-context vendor tracked, whether any of it produces a retained, citable record is unaddressed by the source. No contradiction; no EU/UK FS customer named.

Snapshot

Fivetran + dbt Labs is the merged owner of Fivetran (data movement, managed Iceberg data lake) and dbt (the SQL transformation framework used, by its own count, by “more than 100,000 data teams”). It matters to this wiki because the transformation layer is where reported numbers are actually computed, and at dbt Summit 2026 the company positioned that layer as “the data infrastructure layer that makes agents trustworthy”: pre-run lineage and column validation in the new dbt v2 engine, freshness rules codified per model in dbt State, project-grounded agents (dbt Wizard) that validate impact before changes ship, dashboards defined as version-controlled code (dbt Charts), and dbt metadata served to AI agents through an MCP Server, the open-source Agents Schema and the new Fivetran Context Layer [S-2026-09-16-dbt-summit-2026-announcements]. Nothing in the announcement names a regulator, regulation or EU/UK financial-services customer; the regulatory read-across below is this wiki’s inference.

Positions / Claims they advance

  • “Open Data Infrastructure” as vendor-neutrality: storage, compute, movement, transformation and visualisation should be independently chosen, with data in customer-owned Apache Iceberg tables, so that “AI systems … work across platforms using data and context the enterprise owns, not a single vendor” [vendor positioning — S-2026-09-16-dbt-summit-2026-announcements].
  • One engine (dbt v2, GA 16 Sep 2026): the Rust “Fusion” engine becomes dbt v2; dbt Core (Python) becomes dbt v1 and remains Apache 2.0; v2 ships as a free superset dbt and an Apache-2-only dbt-oss. Claimed to parse a 10,000-model project up to 10x faster and to surface “errors, column checks, and lineage before anything runs”. Adapters GA for BigQuery, Databricks, DuckDB, Redshift and Snowflake; Fabric, Athena and Postgres “coming soon” [vendor claim — S-2026-09-16-dbt-summit-2026-announcements].
  • dbt State (GA, paid add-on): change-aware builds (build/skip/clone/defer per model) with freshness moved “from the job to the model” via a per-model lag_tolerance in code; the company’s own framing is that “the guardrails now sit in the infrastructure instead of with whoever, or whichever agent, issues the command”. Claimed 15–30%+ average compute reduction; Virgin Media O2 (UK telco) cites 25% [vendor and testimonial claims — S-2026-09-16-dbt-summit-2026-announcements].
  • Context for agents: dbt Semantic Layer (GA), dbt MCP Server (GA — exposes models, metrics, lineage and test results to agents), Agents Schema (open-source context standard) and GA Anthropic/ChatGPT integrations, extended by Fivetran Context Layer (private beta) to unstructured sources [S-2026-09-16-dbt-summit-2026-announcements]. The dbt Semantic Layer is a named consumer of Apache Ossie YAML and dbt a named Ossie contributor, so Ataccama’s announced DQ-signal converter would land in dbt-consumed context [S-2026-09-09-ataccama-apache-ossie-converter].
  • Agents for the data team: dbt Wizard (public preview; CLI beta; Desktop private beta) is grounded in the project’s lineage, tests and contracts and “validates proactively: checking upstream and downstream impact, compiling and building the change before anyone sees the diff”; Wizard Explore Mode (public preview) answers business users’ plain-language questions from the same governed project [vendor claim — S-2026-09-16-dbt-summit-2026-announcements].
  • Dashboards as code — dbt Charts (public beta): dashboards defined as YAML “version-controlled right next to the models they depend on … the same pull request, the same CI as the SQL underneath it”, explicitly pitched against governance “living in a separate, closed tool” [vendor claim — S-2026-09-16-dbt-summit-2026-announcements].
  • Multi-engine transformation — Lake Compute (private beta): a DuckDB-based engine running tagged dbt models directly on Iceberg tables while other models stay on the warehouse, with refs working across engines [vendor claim — S-2026-09-16-dbt-summit-2026-announcements].
  • Governance read-across (inference, not in the source): for BCBS 239-facing firms the evidence primitives here are (a) pre-run lineage/column checks and Wizard’s pre-ship impact validation (Principle 3 accuracy, Principle 7 adaptability/change control), (b) codified lag_tolerance as a machine-checkable timeliness rule (Principle 5), and (c) Charts-as-code putting the executive-facing number and its logic under one change-control trail; the MCP/Context Layer route makes dbt metadata an input to AI systems (EU AI Act Art. 10 data governance and Art. 12 record-keeping angle). Whether any of these produce retained records a firm could cite is not stated [inference — S-2026-09-16-dbt-summit-2026-announcements].

Relationships

  • relates-to → Agentic Data Access Governance — transformation-layer entrant to the governed-context-for-agents category (MCP Server, Agents Schema, Context Layer) [S-2026-09-16-dbt-summit-2026-announcements].
  • relates-to → BCBS 239 and Data Lineage — dbt lineage, tests and contracts are the transformation-layer lineage source that catalogue and lineage tools (e.g. Qlik’s metadata bridges) read [S-2026-08-24-qlik-idc-marketscape]; evidence read-across is inference.
  • partners-with → Ataccama — Ataccama’s announced Ossie converter targets the dbt Semantic Layer as a consumer; both are Apache Ossie contributors [S-2026-09-09-ataccama-apache-ossie-converter].
  • partners-with → Snowflake — GA v2 adapter; dbt State runs on Snowflake; co-contributor to Apache Ossie [S-2026-09-16-dbt-summit-2026-announcements][S-2026-09-09-ataccama-apache-ossie-converter].
  • partners-with → Databricks — GA v2 adapter; dbt State runs on Databricks; co-contributor to Apache Ossie [S-2026-09-16-dbt-summit-2026-announcements][S-2026-09-09-ataccama-apache-ossie-converter].
  • partners-with → Anthropic — GA integration delivering dbt context to Claude via AI marketplaces [S-2026-09-16-dbt-summit-2026-announcements].
  • competes-with → Qlik — overlapping pipeline/transformation, lineage and governed-KPI-to-agent (MCP) positioning [inference — S-2026-08-24-qlik-idc-marketscape].

Tracked changes

  • 2026-06-01 — Fivetran and dbt Labs merger completed (release seen by the 15 Sep scan; not ingested as a Source page) [S-2026-09-16-dbt-summit-2026-announcements — banner reference only].
  • 2026-09-16 — dbt Summit 2026: dbt v2 and dbt State GA; Fivetran Context Layer (private beta), dbt Wizard (public preview), Lake Compute (private beta), dbt Charts (public beta) introduced; Anthropic and ChatGPT context integrations GA [S-2026-09-16-dbt-summit-2026-announcements].

Open Questions

  • What policy, masking or certification state travels with a dbt MCP Server / Context Layer call, and is each agent call logged with retention? Not stated — the same unanswered question recorded for every MCP-context vendor on Agentic Data Access Governance [S-2026-09-16-dbt-summit-2026-announcements].
  • Does dbt Wizard’s pre-ship validation leave a retained artefact (impact analysis, test results, approver) usable as change-control evidence, or only an in-session diff? Not stated.
  • Can lag_tolerance breaches be reported historically per model — i.e. is freshness a reportable timeliness control or only an execution optimisation? Not stated [inference].
  • No EU/UK regulated-FS customer named in any vault source; the FS reference base for dbt v2 / dbt State is unknown.
  • Fivetran’s stewardship of Great Expectations and the reported GX Cloud shutdown (June 2026) surfaced only in a search summary during the 17 Sep run and are unverified — a candidate future Source if a primary can be fetched [not sourced — lead only].
  • The 1 June 2026 merger-completion release has been seen but not ingested; a Source page would give this page its corporate-structure baseline.

Sources

Graph View

Backlinks

Recently updated

See 113 more →