Stop Rebuilding Metrics That Already Exist
Teams reuse governed definitions instead of paying to recreate them, and valuable definitions surface automatically instead of dying in one repo.
AI Agent · Decision Intelligence
Canonical
One Metric. One Truth. Every Platform.
Enterprise Semantic Layer & Metric Governance
Canonical turns scattered metric logic into one governed, reusable source of truth. Every metric is defined once, approved by its owner, and used the same way by every BI tool, notebook, pipeline and AI agent.
10Source types scanned, from dbt and Power BI to Snowflake and Excel
3Trust tiers, matched to the business impact of each metric
1Governed definition per metric, resolved the same way by every tool and AI agent
The Challenge
Business Metrics Live in Silos, Not Systems of Record
Critical business metrics live as hand-written YAML, spreadsheets and code buried inside individual teams, repos and regions. Every BI tool, notebook and pipeline re-implements the same logic its own way, so "Revenue" can mean five different things across five dashboards, and no one can prove which is right. Teams rebuild metrics that already exist elsewhere, definitions stay locked inside one business unit, and leadership loses confidence in the numbers. Now AI agents are querying data on their own, and they inherit every one of those inconsistencies. The result is duplicated effort and no enterprise-wide source of truth. A semantic layer closes that gap: one place where each metric is defined once, governed centrally, and resolved the same way by every tool that queries it.
What Sets This Agent Apart
Scattered Metric Logic vs. Canonical
Without CanonicalWith Canonical
- Every team hand-writes its own version of "Revenue" or "Churn"Existing definitions are discovered and reused across teams
- Definitions stay locked in one team's repo, spreadsheet or headOne catalog every business unit and geography can search
- The same metric means different things on different dashboardsOne governed definition resolved the same way by every tool
- New certified metrics need manual discovery, YAML and sign-offAgentic scans draft candidates; SMEs review and approve
- No record of who decided what, or whyEvery approval logged with a full audit trail, version-controlled like code
- AI agents invent their own business logic uncheckedAgent queries are checked against approved definitions in real time
How It Works
Human Judgment Where It Matters. AI Scale Everywhere Else.
Canonical runs two paths that meet in one certified library. Strategic, board-level metrics are authored directly by their owners. Long-tail metrics are discovered at scale across code, tools and data, then drafted for review and never auto-published. Authors define metrics in a drag-and-drop, YAML-backed builder with full SQL editing, while SMEs run scans, review ranked candidates and manage the approval queue in a separate discovery console.
Live agent flow
Step 01
Scan & Draft
Agentic scans search everywhere metric logic already lives and surface candidate definitions.
What this step does
- Scans dbt models and YAML, Power BI DAX measures, Tableau calculated fields and Looker LookML
- Reads SQL stored procedures, Snowflake views, Databricks and Jupyter notebooks, Python scripts and Excel workbooks
- Drafts candidate definitions with source attribution
- Runs recurring, configurable scans (daily or on your schedule), so discovery stays continuous
Step 02
Rank
Candidates are prioritized so subject-matter experts review the highest-impact metrics first.
What this step does
- Which metric first: ranked by usage frequency, occurrences and senior or tenured user activity
- Which definition wins: every variant ranked by prevalence, with source attribution
- For example, "Revenue" found in 14 Power BI reports, 6 dbt models and 2 notebooks
- Analysts pick the winner from what the organization already uses, not from memory
Step 03
Adjudicate
In the adjudication workspace, SMEs compare variants side by side, like a pull request for metrics.
What this step does
- Side-by-side comparison of every variant
- Edit YAML or SQL, merge or rewrite, and add business context
- Approval routed by trust tier, from human-only to light-touch
- Every decision logged: who approved it, what alternatives existed, what was rejected and why
Step 04
Certify
Approved metrics are promoted to the canonical library, instantly discoverable and reusable by every team, tool, geography and AI agent.
What this step does
- Publishes certified definitions to one discoverable catalog
- Resolves queries from Power BI, Tableau, Looker, notebooks and NL2SQL agents through the semantic layer
- Notifies metric owners automatically when new metrics are found
- Keeps every metric versioned with a named owner and complete history
Built for AI Agents
Keeps AI Agents on Approved Definitions
As AI agents query data on their own, Canonical checks their logic against approved definitions in real time, so AI-generated business logic cannot drift away from what the business has certified.
Example alert
“An agent queried something that looks like a ‘blended margin’ metric 47 times today. It doesn’t exist in your semantic layer. Want to review and formalize it?”
Certified match
The agent used an approved definition correctly. Logged automatically, no action needed.
Variant detected
The agent used a known metric with a slightly different formula. Flagged as a candidate for the next review cycle.
New metric found
The agent created logic that doesn't exist in the library. A discovery notification goes to the owning SME.
Business Impact
Trusted Metrics.Faster Decisions.Lower Cost.Less Risk.
End Conflicting Numbers
One governed definition replaces conflicting spreadsheet math across dashboards.
Decide Faster, on the Right Number
Trusted metrics are available on demand, not weeks away, so leaders act on the right number the first time, not the third.
Preserve Institutional Knowledge
Metric logic outlives individual authors, teams and turnover.
Own Every Number
Every published metric carries a named SME or data steward and a complete history.
Align the Whole Enterprise
Globally distributed teams build on the same catalog and one enterprise standard, and respond to market shifts without waiting on metric plumbing.
Industry Applications
One Source of Metric Truth Across Industries
01
Retail & E-commerce
Align sales, margin and customer metrics across merchandising, digital and store reporting.
02
CPG
Standardize volume, share and trade-spend metrics across brands, regions and retail partners.
03
Manufacturing
Give plants and leadership one definition of yield, downtime and quality across sites.
04
High-Tech & Digital Platforms
Govern product, usage and revenue metrics across engineering, finance and go-to-market teams.
05
Gaming, Media & Entertainment
Reconcile audience, engagement and advertising metrics across platforms and partners.
06
BFSI
Keep risk, revenue and customer metrics consistent across business units and regulatory reporting.
Trust & Governance
A Tiered Trust Model, Proportionate to Business Impact
No metric is ever published without a human in the loop. Canonical applies three trust tiers, proportionate to business impact. Tier 1, Human-Only: board-level and earnings-call metrics such as net revenue, gross margin and active customers need a named approver, and the system blocks any agent or scan from modifying them. Tier 2, Standard Review: most business metrics go through full SME or data steward review before certification. Tier 3, Light-Touch: operational, low-stakes metrics can be suggested automatically and go live with lighter review, for speed where the risk is low. Every approval is committed with a full audit trail and version history. Queries resolve through the semantic layer, not around it, and Canonical is built on open, interoperable components with a MetricFlow-style core, so it plugs into the stack you already run (Databricks, Snowflake, SQL and dbt) instead of becoming a walled garden.
Get Started
Every Team. Every Tool. Every Agent. One Number.
Turn scattered tribal knowledge into a trusted, enterprise-wide metrics catalog: discoverable, governed and ready to reuse.