Symbiosis Analytics sits on a code-defined semantic layer for metrics, dimensions, joins, and ownership. Ask in natural language, get a chart with the SQL always visible, plus lineage, freshness, and definition context. Verified dashboards, self-serve exploration with row-level security, metric certification, anomaly alerts, scheduled packs, and Agent Cloud-governed insight agents. If the layer can't answer, the product says so.
NL-to-chart tools fail in the dark. A wrong join produces a confident lie. Showing the SQL turns every answer into a reviewable artifact. Analysts verify in seconds; executives learn to trust in weeks.
BI sprawl dies when every team invents revenue differently. Analytics defines metrics once in code, versions them, certifies them, and serves the same definitions to humans, dashboards, and agents. The warehouse stays the engine; the semantic layer is the contract.
Honesty applies to numbers: cite the definition, show the query, surface freshness, abstain when the model can't compile safely. Agents that act on metrics inherit Agent Cloud policy. An insight can recommend; writes stay governed.
NL → SQL → chart
Natural-language questions compile against the semantic layer; SQL, tables, and metric defs always visible beside the chart.
Governed semantic layer
Metrics, dimensions, entities, and joins defined as code - owned, versioned, tested, reused by humans and AI alike.
Trust affordances
Lineage, freshness, grain, and definition surfacing on every answer - no black-box KPIs.
Honest abstention
If the layer can't support the question, say so - never improvise a join or invent a measure.
Metric certification & review
Draft → review → certified lifecycle; breaking changes require owners; consumers see status badges.
Verified dashboards
Pin certified answers into living boards with subscriptions, embeds, and access controls.
Self-serve exploration
Drag/NL explore within certified metrics; sandbox for analysts without polluting prod defs.
Row/column security
RLS/CLS from IdP groups and data contracts - NL answers never become a permissions bypass.
Anomaly & threshold alerts
Metric monitors with explain-the-spike contexts (SQL + contributing dimensions).
Scheduled packs & exports
Board, regulator, and ops packs assembled on cadence - live data, not quarter-end heroics.
Embedded analytics
Charts and NL ask inside ERP/CRM/HR/product UIs with the same semantic contract.
Warehouse connectors
Snowflake, BigQuery, Redshift, Databricks, Postgres, and lakehouse patterns - pushdown where possible.
Data Platform bridge
Inherits catalog ownership, quality tests, and freshness contracts from Symbiosis Data Platform.
Operational system handoffs
Certified metrics over ERP/CRM/HR/Cloud facts - one revenue, one headcount, one margin.
Insight agents (governed)
Agent Cloud agents that summarize, explain variance, and draft actions - reads via semantic layer; writes only under policy.
Knowledge grounding for defs
Metric documentation and playbooks retrievable with citations - 'what is net revenue?' answers from owned defs.
Eval harness for NL accuracy
Golden-question suites: SQL correctness, metric match, abstention quality - regressions block deploys.
Cost & performance OS
Query cost attribution, cache/materialization suggestions, warehouse spend guards.
Collaboration & comments
Threaded discussion on answers/boards with pinned SQL snapshots for audit.
Migration from BI sprawl
Import Looker/Tableau/Power BI logic into semantic code; coexistence until certified cutover.
SEMANTIC STUDIO
Code-defined metrics/dimensions, ownership, tests, certification workflow.
ASK PLANE
NL→SQL→chart with visible SQL, lineage, freshness, and abstention.
TRUST & ACCESS
RLS/CLS, IdP sync, audit of every question and query.
BOARDS & PACKS
Verified dashboards, subscriptions, scheduled board/regulator packs.
MONITOR OS
Anomaly/threshold alerts with explain-the-spike SQL context.
EMBED & API
Embedded charts/NL, semantic query API for products and agents.
PLATFORM BRIDGES
Warehouse + Data Platform + ERP/CRM/HR/Cloud metric sources.
AGENT & EVAL
Governed insight agents + NL accuracy regression suites.
Investment is bespoke to estate size, compliance needs, and deployment shape - book a demo and we'll map the right fit.
STARTER
- → Semantic layer (one domain)
- → NL→SQL→chart with visible SQL
- → Basic lineage/freshness
- → Pin to dashboards
- → One warehouse connector
BUSINESS
- → Everything in Starter
- → Metric certification workflow
- → RLS from IdP
- → Alerts & scheduled packs
- → Embedded analytics
- → Data Platform / ERP bridges
- → NL eval golden set
ENTERPRISE
- → Everything in Business
- → Multi-domain semantic mesh
- → Insight agents (Agent Cloud)
- → BI migration & coexistence
- → Query cost governance
- → Compliance evidence exports
- → Private deployment options
Analytics, answered
What stops it from inventing numbers?
It can only query the governed semantic layer, and it must show its SQL. If the layer can't answer, the product says so instead of improvising.
Do we replace our warehouse?
No - Analytics sits on your warehouse/lakehouse. The semantic layer is the contract; the warehouse remains the engine.
How is this different from ChatGPT-on-CSV?
Owned metric definitions, RLS, lineage, visible SQL, abstention, and eval gates. Fluency without a semantic contract is how confident lies ship.
Can agents act on insights?
Insight agents summarize and recommend via Agent Cloud; any write-back (tickets, ERP adjustments) stays policy-gated and audited.
What about Looker/Tableau/Power BI?
Often coexist during migration. We help port logic into semantic code so 'revenue' stops meaning five things. Sometimes the ecosystem tool stays as a consumer of the same layer.
Who owns metric definitions?
Named owners in Semantic Studio - certification and breaking-change review are product features, not wiki hope.
See it against your reality.
A 30-minute walkthrough with an engineer who builds it - your use case, not a script.