Metric Definitions
Finance Says $487K. Sales Says $512K. Marketing Says $503K.
Before you can have a real business conversation, you spend 20 minutes arguing about whose number is right. Fig ends that. Define a metric once, get it approved, and every person and every AI agent in your company uses the same number — forever.
How It Works
Three steps from raw metric to enforced definition — with a human approval gate in the middle.
Define
Write the metric in plain language and SQL. Name it, describe it, choose the data source. Example: "MRR = SUM of active subscription amounts, excluding trials."
Approve
Assign an owner and require sign-off. The metric is marked provisional until approved. Only approved metrics can be used in production decisions or agent queries.
Enforce
Once approved, Fig enforces the definition on every query. Every agent call, every Flow, every analysis uses the blessed version. If a conflict is detected between sources, Fig flags it before surfacing the number.
Why This Matters
The problems that happen when metric definitions live in spreadsheets, Slack threads, and institutional memory.
“Finance says $487K. Sales says $512K. Marketing says $503K.”
With Fig, there's one version — the one Finance approved. Everyone else references the same definition.
“The definition changed last quarter and nobody told the agents.”
Every change creates a new version. Agents always reference the current approved version. Rollback is one click.
“Our AI gave us a confident number based on the wrong table.”
Fig's agents can only use approved metric definitions. No guessing what your data means. No wrong data connections.
What You Can Do
Four capabilities that make metric definitions a foundation — not a formality.
Version history
See every change to a metric definition and who approved it.
Conflict detection
When two sources disagree on the same metric, Fig surfaces the conflict and blocks decisions until resolved.
Approval chain
Require sign-off from named owners before a metric goes to production.
Agent enforcement
Fig's AI agent skill gives AI agents access only to approved definitions — not your raw data structure.
The result: When an executive asks “what was our MRR last quarter?” — whether they ask a human analyst, a BI tool, or an AI agent — they get the same number. Because there is only one approved definition, and Fig enforces it everywhere.
Frequently Asked Questions
What is a metric definition in Fig?+
A metric definition is the approved, versioned formula for how a business metric is calculated — who defined it, who approved it, and what data it comes from. Once approved, Fig enforces it on every query made by any human or AI agent.
Why do metric definitions matter for AI agents?+
AI agents reason from the data they can access. Without approved metric definitions, they guess — and different agents produce different numbers for the same question. Fig's metric definitions give every agent the same approved answer every time.
What happens when two data sources disagree on the same metric?+
Fig detects the conflict, surfaces it, and blocks decisions that depend on that metric until it's resolved. No analysis runs on conflicting data without you knowing about it.
Can metric definitions be changed?+
Yes. Any change creates a new version. The previous version is preserved, and every change shows who made it and when. Agents always use the current approved version.
Explore the Platform
Metric Relationship Tree
Causal map of what drives what in your business
Learn moreBusiness Policies & Rules
Your business logic enforced on every query
Learn moreBusiness Entities
Accounts, customers, products resolved across sources
Learn moreFlows
Signal → Decide → Act automated end-to-end
Learn moreMonitoring
24/7 anomaly detection with automatic root cause
Learn moreDecision Algorithms
12 built-in algorithms for governed analysis
Learn moreRoot Cause Analysis
Automated causal chain tracing
Learn moreIn-Memory Cache
Reduce warehouse costs, sub-second answers
Learn moreOne Metric. One Definition. Trusted by Everyone.
Define your metrics once, get them approved, and let Fig enforce them on every query — for every person and every AI agent in your company.