The semantic layer
Metrics that mean what you mean.
"Active customer." "Churn." "Net revenue." Every company calculates them a little differently — and most tools just guess. Synopsis learns your definitions from your own documents and enforces them in every answer.
# learned from: FY budget, Q3 board deckactive_customer: definition: paid plan + order in last 90 days excludes: internal and trial accounts source_of_truth: FY budget, revenue tab tests: 6 passing# every surface — chat, dashboards, workflows — uses thisVerified against: budget · board deck · ops report
Learned, not configured
Reverse-engineered from your documents.
Nobody sits down and writes out metric definitions — they live implicitly in budgets, board decks, and the reports your team already trusts. Synopsis reads those documents, works out how each number was actually calculated, and keeps refining until its results match your authoritative figures.
- Your documents are the spec. Budgets, decks, and reports define the target — not a config file someone has to write.
- Matched to the number. A definition isn't accepted until it reproduces the figures your company already reports.
- No semantic-modeling project. You don't hire consultants to encode your metrics. Synopsis derives them.
Tested, and honest
It builds test cases — and asks when it's unsure.
Every definition gets test cases built from your authoritative figures, so a metric can't silently drift from what your company reports. And when two plausible definitions produce different numbers, Synopsis doesn't pick one quietly — it asks you which one you mean.
- Tests from real figures. Each definition must reproduce the numbers in your budget and board deck, quarter by quarter.
- Ambiguity surfaces as a question. A clarifying question beats a confident wrong answer.
- Your call is recorded. Answer once, and the definition is settled for every future question.
Two candidate definitions produce different numbers: monthly logo churn is 2.4%, annualized revenue churn is 3.1% — the figure in your board deck. Which should "churn rate" mean by default?
Test cases built from: FY budget · Q3 board deck
One definition, everywhere
Every surface agrees, because there's one source.
The reason numbers disagree between tools is that each tool keeps its own definition. In Synopsis there's one semantic layer over one warehouse — so the chat answer, the dashboard widget, and the workflow trigger are all computing the same thing.
- Chat and dashboards match. Same definition, same warehouse, same number — the meeting stops being about whose figure is right.
- Workflows fire on the real metric. An alert on "active customers" uses the definition your board sees, not an approximation.
- Change it once. When the business redefines a metric, update the definition and every surface follows.
What your team gets
A single vocabulary for the whole company.
Learned from your documents
Budgets, board decks, and reports are read directly — the definitions come from how your company already calculates.
Tested before trusted
Every definition carries test cases built from your authoritative figures, so drift gets caught, not shipped.
Asks, doesn't guess
When a definition is genuinely ambiguous, you get a clarifying question — not a silently chosen assumption.
Your words, mapped
Products, regions, segments, fiscal calendars — the vocabulary your team uses is the vocabulary that works.
Consistent on every surface
Chat, dashboards, and workflows share one definition, so the same question gets the same answer everywhere.
Changes with the business
Redefine a metric once and every answer, widget, and alert picks it up — no hunt through saved reports.
Keep exploring
The semantic layer runs under every surface.
Stop debating whose number is right.
Bring the metric your teams can never agree on. In days, Synopsis will have learned how your company actually calculates it — and every surface will say the same thing.
