Context and accuracy
How shared definitions and vetted knowledge help your team get more accurate, consistent answers.
Context is the shared knowledge Querio uses to understand your data: what tables mean, how they relate, how your team defines metrics and how you approach an analysis.
Each analysis can improve the next
Working with Querio produces more than answers. Your team uncovers business rules, validates queries and develops ways to investigate questions.
When you capture that vetted knowledge as context, it becomes useful beyond the original analysis:
- Analyze: ask questions and explore your data.
- Validate: check the logic and results, resolve mistakes and confirm definitions.
- Build context: save useful definitions, reviewed queries and repeatable procedures.
- Reuse: give the agent that knowledge for the next question.
As your team builds this shared understanding, Querio has better guidance for future analyses. You spend less time explaining familiar rules, and more of your team's experience contributes to accurate answers.
Using Querio helps improve Querio. The value comes from the knowledge your team validates and makes reusable.
Four kinds of context
| Context | How it helps |
|---|---|
| Catalog and joins | Describes tables, columns and relationships so the agent can choose appropriate data and combine it correctly. |
| Knowledge | Supplies business definitions, reference queries and known caveats for a datasource. |
| Memory | Sets conventions that should apply across the workspace, such as reporting periods, currencies and timezones. |
| Skills | Gives the agent procedures to follow for recurring analyses and checks. |
Start with the SQL, dbt documentation and business definitions your team already maintains. Context lives in files, so you can keep it alongside your data project and review changes through Git.
From a rule to a question
Suppose your team defines revenue as:
Revenue is the sum of
amountfor paid, non-refunded orders, assigned topaid_atin UTC.
You ask: What was revenue in September 2026?
That definition guides the agent toward a query that:
- Sums
amountfor paid orders. - Excludes refunded orders.
- Uses
paid_atfrom September 1 inclusive to October 1 exclusive, in UTC.
Inspect the query in the notebook and compare the result with a trusted report. Once you've reviewed the definition and query, they can guide future questions about revenue.
Give useful knowledge a clear home
Querio uses catalog and knowledge from the datasources selected for the analysis, together with workspace memory and relevant skills.
Essential datasource rules can be marked Always. Longer explanations and examples can be Only when relevant, so the agent can find and read them when needed. Clear names and descriptions help it choose the right knowledge and skills.
Keep shared definitions consistent. When you correct an answer, decide whether the correction belongs in reusable context or just that question.
A missing refund rule belongs in shared context. A request to focus on Germany last quarter belongs in the question. An incorrect join or stale source data needs attention in the analysis or data itself.
Put improved context to work
When a shared definition changes, reload context and rerun the affected analysis. Existing results aren't recalculated just because a context file changed.
Context guides analysis; workspace and datasource permissions control access.
For file locations, formats and loading rules, see Context files.
Last updated on October 27, 2026