Home Semantic Layer

One definition of "conversion." Everywhere. Always.

Your pipeline works. Your warehouse has data. But Google says ROAS 4.2, Meta says 3.8, and your CRM shows neither in pipeline. That's not a data problem. That's a semantic layer problem.

Book a Semantic Layer Audit →

Three reasons your metrics never agree.

The data is fine. The problem is that everyone has a different copy of the truth.

Metric sprawl

Marketing uses one definition of MQL. Sales uses another. Finance uses a third. Every dashboard is right and none of them agree.

Logic embedded in dashboards

Business rules live inside Looker calculated fields, Tableau LODs, and spreadsheet formulas. Nobody knows which one is canonical.

No single source of truth

When the CFO asks "what was our CPA last quarter?" — who answers? Which tool? Which definition?


One source of truth for every metric.

Built in the warehouse, not in the dashboard.

01
dbt transformation layer
All business logic defined once, in version-controlled SQL. Metrics computed upstream of every dashboard.
02
Metric catalog
Every KPI defined: what it measures, how it's computed, which source it draws from, when it was last updated. Accessible to analysts and executives alike.
03
Semantic layer integration
Connect your warehouse to a semantic layer (dbt Semantic Layer, Cube, or Metriql) so every tool — Looker, Tableau, Claude, your internal tools — queries the same definitions.
04
Cross-platform entity resolution
One customer ID. One campaign ID. One channel taxonomy. Regardless of how many platforms define them differently.

Before and after a semantic layer.

The same data team, the same analysts — with one structural change.

✕ Before
6 dashboards, 6 definitions of ROAS
Business logic in Excel
New analyst spends 3 weeks "learning the numbers"
CFO meeting requires 2 hours of prep per slide
✓ After
One metric catalog, one definition per KPI
Logic version-controlled in dbt
New analyst productive in 2 days
CFO meeting prep is a dashboard refresh

The tools we build on.

dbtSnowflakeBigQueryDatabricksCubeLookerMetabaseSigmaPython

I built the normalization and aggregation layer at TapClicks that unified campaign data from 200+ platforms into a single coherent model — so that "impressions" meant the same thing whether it came from Google, Meta, TikTok, or a DSP nobody's heard of.

AR
Angshuman Rudra
Founding PM, TapClicks · 10 years

Start with a Semantic Layer Audit.

We map your current metric definitions, find where logic is duplicated or conflicting, and give you a build plan. $2,000 flat.

Book an audit →

Looking for measurement and analytics consulting? See all services →