Dashboards & BI
One prompt. A board-ready dashboard.
Describe the dashboard you need the way you'd brief an analyst. Synopsis builds it from your warehouse, keeps every number live, and can show you exactly where each one came from.
Built by asking
Describe it. Don't drag it.
BI tools hand you a blank canvas and a widget palette, and the dashboard becomes a project. In Synopsis, the prompt is the project: say what you need to see, and the queries, charts, and layout are built for you — against data that's already joined.
- The prompt is the spec. "Revenue by segment, churn trend, top accounts at risk" is a finished dashboard, not a backlog ticket.
- No widget wiring. Every chart already knows its query, because the AI wrote both.
- Change it by asking. "Add a regional split" edits the dashboard the same way it was built.
Done — three sections, live from your warehouse. Churn is concentrated in one segment: mid-market accounts for most of the movement.
Sources: CRM · billing · product events
Live data, your definitions
The numbers agree — everywhere.
Every widget runs on the warehouse, not a saved extract, so the dashboard is as current as your data. And because the semantic layer holds your company's metric definitions, the figure on the dashboard is the same figure chat gives you — one warehouse, one definition, one answer.
- Live, not a snapshot. Widgets re-run on fresh data; nobody re-exports before the meeting.
- Your definitions, enforced. "Net revenue" is computed the way your finance team defines it, in every widget.
- No dueling numbers. Chat, dashboards, and workflows all read the same semantic layer, so they can't disagree.
Net revenue last quarter was $2.41M — the same figure on your revenue dashboard, computed from the same definition.
Definition: metrics.yml › net_revenue · Sources: billing · GL
Trust every widget
Click a number, see the proof.
A dashboard is only as good as your team's trust in it. Every widget in Synopsis traces to source: open it and see the SQL it runs, the tables it reads, and the systems those tables came from. Debates about "whose number is right" end at the trace.
- The query, visible. Anyone can open the SQL behind a widget — analysts can edit it.
- Sources, cited. Each widget names the systems it draws from.
- Definitions, linked. When a widget uses a defined metric, the trace shows which one.
-- widget: "Churned ARR by segment"SELECT segment, sum(arr)AS churned_arrFROM metrics.churn_eventsWHERE quarter ='this'GROUP BY segmentORDER BY churned_arrDESC-- definition: "churn" = metrics.yml › churn_eventsTraced to: CRM · billing
What your team gets
Dashboards people actually open.
Board-ready by default
Clean layout, sensible charts, the right level of summary — built to be presented, not just built.
Always current
Every widget runs on live warehouse data. The Monday meeting starts with Monday's numbers.
Cross-system views
Revenue next to support load next to product usage — joined on shared keys, on one screen.
Edited in English
Add a chart, change a grouping, or swap a metric by asking — the same way the dashboard was made.
Drill in with chat
Any widget can become a conversation. Ask why the line bent and get an answer, not a filter panel.
One version of the truth
Definitions come from the semantic layer, so two dashboards can't quietly disagree about the same metric.
Keep exploring
Dashboards are one surface of the platform.
Bring the dashboard you've been meaning to build.
Describe it in a sentence and watch it come back live, on your data, with every number traceable. That's the whole setup.
