Define each metric once. Trust every number after.

The semantic layer is the single, governed definition of your business — what each metric means, how it's filtered, what it can be sliced by. Every dashboard, embed and AI answer reads from it, so a figure means the same thing everywhere and is never invented on the spot.

Your sources
Warehouse HubSpot Product DB
One definition
semantic layer
active_customer
= paying AND not trial
grain: customer · monthly
defined once · signed by your lead
Every surface
Dashboards AI answers Embeds Reports

written once, read everywhere — the same number in every dashboard, embed and answer.

The problem it solves

One question. Four different answers.

Without a shared definition, every tool counts your business its own way — and nobody can say which number is the right one.

“How many active customers?”
asked in four places, one afternoon
?
Spreadsheet
1,240
?
Legacy BI
1,187
?
CRM report
1,204
?
SQL export
1,150

— four counts, four owners, no agreement —

With a semantic layer,
one definition, one answer:
1,204

Anatomy of a definition

A metric, pinned down.

One definition holds everything a number needs to stay honest — its formula, its filters, the grain, the dimensions it's allowed, the words people use for it, and a human who owns it.

definition · active_customerv4 · signed
metricactive_customer
formulacount(customers) where status = 'paying' and plan != 'trial'
filtersexcludes churned · excludes internal accounts
graincustomer · monthly
dimensionsregion · plan · segment
synonyms“active client” · “paying customer”
ownerRevOps · signed by your Toucan lead
formula
One agreed calculation, written down. Every surface reads this exact formula — no more copy-pasted logic drifting between tools.
filters & grain
The exclusions and the level of detail travel with the metric, so a filter can't quietly change what the number counts.
dimensions
The slices a metric is allowed to be broken by — so a chart can't cut the data in a way that makes it meaningless.
synonyms
The words your teams actually use, mapped to the metric — so the AI understands the question however it's phrased.
owner
A named human accountable for the definition. When it changes, it changes once, on purpose, and everyone inherits it.

AI runs through it

The AI can't make a number up. It has to resolve one.

Ask in plain language, and Studio resolves your question to a defined metric, applies the governed filters, and returns an answer you can trace to its definition.

resolve · semantic layer
ask “active customers in EMEA this quarter?”
├─ resolved metric  → active_customer
├─ applied filter    → region = EMEA
├─ applied filter    → period = Q3 2026
├─ read definition  → paying, non-trial (v4)
└─ answer         → 1,204   ✓ traceable to definition

No number is invented. Every answer is a defined metric plus governed filters — which is exactly why Studio answers with zero hallucinations.

What you get

Four things a definition buys you.

01
One definition of truth

A metric means the same thing in every dashboard, embed and answer. No more reconciling four numbers before a meeting.

02
Zero hallucinations

The AI resolves questions to defined metrics instead of guessing. Figures are governed, never free-styled.

03
Governed & contextual

Allowed dimensions, synonyms and access rules travel with each metric, so answers stay in-context and in-bounds.

04
Traceable to the source

Every number links back to its formula, its filters and its owner — defensible in a board meeting, auditable after one.

Govern your numbers — define them once.

Start free →

Want the definitions built with you? Hire a Crew →