Semantic Layer
One definition of every metric. Across teams. Across tools.
The 'whose number is right' meeting happens because there isn't one number. Sales says MRR is X, finance says it's Y, the dashboard says Z. The semantic layer is the contract that makes them agree — once, codified, queried by every tool downstream.
What's inside
Definitions that hold up because they're queryable.
- 01
Metric catalog
Every metric the business uses, defined once. Owner, formula, source tables, edge cases, version history. Searchable, peer-reviewed.
- 02
dbt + semantic layer
Definitions live in code, version-controlled, tested. Built on dbt's semantic layer, Cube, or MetricFlow — whichever fits your stack.
- 03
Tool-wide consistency
BI, sheets, CRM, marketing platforms — all hitting the same definitions. The dashboard and the spreadsheet finally agree.
- 04
Migration from chaos
Most companies have the same metric defined 12 ways across tools. I consolidate, deprecate, and migrate without breaking what works.
How an engagement starts
Audit the definitions. Codify the canonical set. Migrate.
- 01
Definition audit
Pull every metric in use across the org. Most have 3–8 conflicting versions. We rank by criticality and disagreement spread.
- 02
Codify the canonical set
Top 20 metrics defined in dbt or your semantic layer. Reviewed by the metric owner. Tested. Documented.
- 03
Cut over the consumers
BI dashboards, CRM fields, scheduled exports — pointed at the new definitions one at a time. Old definitions deprecated, not deleted.
Half-day. Output is a ranked list of where your metrics disagree.
Common questions
- Which semantic layer?
- dbt semantic layer by default — best fit for most modern stacks. Cube for embedded analytics, MetricFlow when dbt isn't already in play.
- Will this slow us down?
- Short-term: a little — the discussion of what 'revenue' means is the slow part. Long-term: every report gets faster because there's no rebuild.
- What if teams disagree?
- They will. The semantic layer forces the conversation that should've happened years ago, and pins the answer. That's the value.
Also under Analytics
Business Intelligence
Dashboards that answer the question. And surface the next one.
Automated Reporting
The Monday brief, written by the data. Not the analyst.
Interactive BI
Self-service dashboards the team actually opens.
Agentic Data Intelligence
Agents that watch the metrics. And explain the spike before you ask.