Documentation

Metric Schema Reference

Complete field reference for the MetricRegistry definition format.

Field Reference

Field	Type	Required	Description
id	string		Fully qualified metric ID (e.g. uk.banking.capital.cet1_ratio)
version	semver		Semantic version of this metric definition
schema_version	integer		Schema format version (currently 1)
title	string		Human-readable metric name
description	string		One-line summary of what this metric measures
certification	enum		Lifecycle status: draft, pending_review, certified, deprecated
formula.dialect	string		SQL dialect (ansi, postgres, bigquery, etc.)
formula.sql	string		The canonical SQL formula with template variables
inputs[]	array		List of input datasets with columns required
inputs[].dataset	string		Template variable name for the source table
inputs[].columns[]	array		Template variable names for required columns
output.column	string		Name of the output column produced
output.type	string		Data type of output (numeric, integer, text, etc.)
output.unit	string		Unit of measure (ratio, percentage, gbp, tonnes_co2e, etc.)
frequency	string	--	How often this metric is typically reported
thresholds[]	array	--	Regulatory or internal thresholds (type, value, source)
regulatory_source	object	--	Source framework, URL, excerpt, section reference, caveats
provenance.source_excerpt	string	--	Verbatim text from the source document
provenance.calculation_notes	string	--	Notes on implementation details or edge cases
provenance.caveats	string	--	Known limitations or jurisdictional variations

Certification Lifecycle

draft

Initial definition, may be incomplete

pending_review

Submitted for expert validation

certified

Validated by domain expert, production-ready

deprecated

Superseded or no longer applicable

Template Variable Syntax

All formulas use ${variable_name} placeholders instead of hardcoded table or column names. This ensures metric definitions are schema-agnostic and portable across different data warehouses.

Each variable in the formula maps to an entry in the inputs array. When implementing the metric, substitute each variable with your actual table and column names.

Zero Fabrication Policy

Every metric in the registry must trace to a verifiable primary source. No metric definition may contain fabricated regulatory references, invented thresholds, or unverifiable source excerpts. If a value cannot be verified, it must be marked as null with a caveat explaining why.

Complete Example

uk.banking.capital.cet1_ratio.yaml

id: uk.banking.capital.cet1_ratio
version: 0.1.0
schema_version: 1
title: CET1 Ratio
description: >
  Common Equity Tier 1 capital as percentage
  of risk-weighted assets
certification: draft

formula:
  dialect: ansi
  sql: >
    SELECT ${cet1_capital} / nullif(${rwa}, 0)
    AS cet1_ratio
    FROM ${capital}

inputs:
  - dataset: capital
    columns:
      - cet1_capital
      - rwa

output:
  column: cet1_ratio
  type: numeric
  unit: ratio

frequency: quarterly

thresholds:
  - type: regulatory_minimum
    value: 0.045
    source: CRR Article 92(1)(a)
  - type: buffer_requirement
    value: 0.025
    source: CRR Article 129

regulatory_source:
  framework: CRR Article 92(2)(a)
  url: https://legislation.gov.uk/eur/2013/575
  excerpt: >
    The CET1 capital ratio shall be the CET1
    capital expressed as a percentage of the
    total risk exposure amount.
  caveats: >
    Transitional provisions may apply.
    Check PRA PS1/26 for UK-specific rules.

provenance:
  source_excerpt: >
    "The CET1 capital ratio shall be the CET1
    capital expressed as a percentage of the
    total risk exposure amount."
  calculation_notes: >
    Uses nullif to guard against division by
    zero when RWA is reported as zero.
  caveats: >
    Different jurisdictions may have different
    minimum requirements.