Metrics glossary

Short definitions of the terms used across Duodata and the metrics-layer space.

Business-Approved Metric

A metric whose meaning has been defined, reviewed and approved by the business owner who is accountable for it, not just by the data team. Duodata is the system of record for business-approved metrics.

Metrics Ontology

A governed model of an organization's metrics: their approved definitions, valid variants (slices), owners, value drivers and where each metric is implemented. Duodata builds and maintains the Metrics Ontology and projects it into data platforms, BI tools and AI assistants.

Metrics layer

The layer that defines what a company's metrics mean, once, so every tool calculates them the same way. Duodata is a business-first, conceptual metrics layer above the physical semantic layers in Snowflake, Databricks and dbt.

Semantic layer

A technical layer that maps business terms to tables, columns, joins and aggregations so tools can query data consistently. Examples: Snowflake Semantic Views, Databricks Metric Views, the dbt Semantic Layer (MetricFlow) and Microsoft Fabric semantic models.

Context layer (AI context layer)

The business context an AI needs to answer data questions correctly: approved definitions, owners, valid variants, value drivers and instructions for retrieving the number. Duodata's Metrics Ontology is the context layer for metrics.

Talk to your data

Asking questions of company data in plain language through an AI assistant such as Copilot, ChatGPT, Claude, Snowflake Cortex or Databricks Genie, instead of writing SQL or building a dashboard. The answers are only reliable when the assistant uses approved metric definitions.

Self-service analytics

Business users answering their own data questions without waiting for the data team, increasingly through AI assistants. It only scales when everyone draws on the same governed metric definitions.

Agentic analytics

AI agents that query data, explain results and trigger actions on their own. Without approved semantics, agents invent metric logic and contradict the dashboards.

MCP (Model Context Protocol)

An open protocol that lets AI assistants connect to external tools and data sources. Duodata's AI Connector serves approved metric context over MCP, API and CLI.

Metric governance (KPI governance)

Assigning owners to metrics and reviewing, approving and versioning their definitions, so every team and tool uses the same meaning.

Agent governance (business meaning)

Governing what AI agents mean when they answer: which approved metric, variant and implementation they use, and stopping when a requested metric is not approved. It complements data governance (access, quality, lineage) and runtime agent controls (permissions, identity, guardrails). Duodata governs agents at this business-meaning level.

Single source of truth for metrics

One approved definition per metric that every dashboard, data platform and AI assistant uses, even when the implementations differ.

KPI dictionary

A list of KPI definitions, usually in a spreadsheet or wiki. Unlike a metrics layer, it has no approval workflow and is not connected to the platforms that calculate the numbers.

Data catalog

An inventory of an organization's datasets, owners and lineage. It shows which tables exist, not which number is the approved one.

Metric Steward

A named Duodata user who creates, edits, approves and publishes metric definitions. Viewers are unlimited and free.

Metric owner

The business person accountable for what a metric means, for example a Head of Controlling for gross margin. Owners can be assigned without logging in to Duodata.

Slice

A valid variant of a metric, such as revenue by region or by product line, defined once and governed alongside the metric.

Value driver tree

A map of which metrics influence which others, for example activation influencing churn and churn influencing net revenue retention. It explains why a number moved, which data lineage cannot.

Implementation Mapping

A link between an approved metric and where it physically lives in a data platform, such as a Snowflake Semantic View or a Power BI measure.

Definition drift

When the same KPI is calculated differently across dashboards, platforms and AI tools, so teams get conflicting numbers.

AI Connector

Duodata's interface (MCP, API, CLI) that serves approved metric context to Copilot, Claude, ChatGPT and custom agents, so they answer with approved logic instead of inventing it.

Apache Ossie

An open, vendor-neutral standard for semantic and metric definitions, backed by Snowflake, Databricks, dbt Labs and others. Duodata stores metric definitions as Apache Ossie / MetricFlow YAML in Git.