Five Layers of Understanding

When an agent asks "show me the money", OnlyMetrix resolves it to your revenue metric. Not through keyword matching, through five layers of semantic understanding.

Each layer builds on the last

Resolution isn't a single lookup. It's a pipeline of understanding that narrows millions of possible interpretations to the one metric your business actually means.

1

Glossary

Synonym resolution

"Money" maps to "revenue". "Churn" maps to "customer_attrition". Your team's vocabulary, formalized. Every synonym is an explicit link, not a guess.

2

Taxonomy

Tag-based organization

Metrics organized by domain, team, data source, and business function. When an agent searches "finance", it finds revenue, costs, margins, not a keyword match on SQL column names.

3

Ontology

Join graph and relationships

The system knows that orders joins to customers on customer_id, and that products relates through line_items. When an agent asks for "revenue by customer segment", the join path is already mapped.

4

Temporal

Time-aware queries and fiscal calendars

"Last quarter" means your fiscal quarter, not the calendar one. Time grains, retention windows, and comparison periods are built into the layer, so agents never guess at date math.

5

Context

Table and column descriptions, business rules

Every table, column, and metric carries human-written context. "This column contains gross revenue before returns." Agents read this context to disambiguate, and it's visible in the audit trail.

Weighted multi-signal scoring

When an agent searches for a metric, OnlyMetrix scores candidates across six signals. No single signal dominates, and the weights are tuned on real-world queries.

Embedding similarity
35%
Keyword match
20%
Canonical name
15%
Importance weight
10%
Tag match
10%
Intent classification
10%

See it resolve your metrics

Type a natural language query. Watch five layers of understanding find the right metric.