Definition
Conversational analytics is a data interaction paradigm where users ask questions about their data in natural language and refine answers through follow-up questions — like a conversation with an analyst who has instant access to every table.
How multi-turn context works in Chion
When you ask "revenue by region last quarter," Chion generates verified SQL and renders a chart. When you follow up with "break that down by month," the pipeline maintains context from the previous turn:
- 1The prior SQL contract is preserved as context
- 2The follow-up intent is extracted relative to the previous answer
- 3A new SQL contract is built that extends (not replaces) the prior state
- 4The new query is validated and executed
What a session stores and discards
Each conversation turn stores: the question, the SQL contract, the generated SQL, the chart configuration, and the result summary. Raw query results are not persisted — they're re-fetched if needed. Session state is user-scoped and template-fenced to prevent cross-conversation data mixing.
Examples
- "Revenue by region" → follow-up: "just EMEA" → follow-up: "as a line chart by month"
- "Top 10 customers by churn risk" → follow-up: "show their last order date"
- "MRR by plan tier" → follow-up: "compare to last quarter"
How it's different from a chatbot
A chatbot wraps your question in a prompt and returns text. Conversational analytics generates verified SQL, executes it against your real database, and renders an interactive chart with the SQL visible underneath. Every answer is traceable to a query you can verify.