# Stop AI from guessing what “revenue” means. - URL: https://betteranalyst.com/product/data-context Formula Bot is now Better Analyst—same account, more powerful analysis.Formula Bot is now Better Analyst—same account and workflows, with more powerful analysis, dashboards, and automation. [Learn what changed](https://betteranalyst.com/formula-bot-is-now-better-analyst) Data Context for SQL For connected SQL databases, review and publish business meaning for tables, fields, relationships, and metrics, so analysis starts from your team’s definitions instead of a guess. [Set up Data Context](https://app.betteranalyst.com/dashboard/library?tab=connectors)[Start free](https://app.betteranalyst.com/signup) ![Data Context Definitions editor with table grain and field aliases, descriptions, roles, and review status](https://betteranalyst.com/screenshots/product/data-context/definitions-v2-light.webp?dpl=c3a99e2b7961f51c9b379add33be693e641879a6)![Data Context Definitions editor with table grain and field aliases, descriptions, roles, and review status](https://betteranalyst.com/screenshots/product/data-context/definitions-v2-dark.webp?dpl=c3a99e2b7961f51c9b379add33be693e641879a6) Swipe to inspect the interface **Definitions**A coffee sales table with its grain and description, and each field’s alias, description, role and review status. Sounds right, quietly wrong ## Why general AI gets metrics wrong. A general assistant sees column names, not your business rules. It fills the gaps with reasonable guesses, and the result reads like a real answer. Data Context replaces three common guesses with definitions your team publishes. 1. Wrong definition ### “Revenue” is gross in one answer and net in the next. Without your rule, AI picks a reasonable meaning. Gross or net of refunds? Is an “active user” someone who logged in or someone who paid? **Published metrics.** Define Net Revenue or Active Users once, with a description and format, and analysis of that data starts from it. 2. Wrong grain ### Order lines counted as orders. When a table holds one row per line item, counting rows overstates orders and average order value comes out too low. **Dataset grain.** State what one row represents, such as “one row per order line,” so counts and averages start from the right unit. 3. Wrong join ### A join that quietly double-counts revenue. Join orders to a table with several rows per order, and each order’s revenue is repeated once for every match. **Approved relationships.** Set the join fields and cardinality, such as many-to-one, so Better Analyst knows how your tables connect. Data Context covers connected SQL databases today. For uploaded files, Google Analytics, and Search Console, state your definition in the question or in an AI Agent’s instructions, and check the code shown with every answer. Definitions people can read ## Turn physical fields into business language. Add a description and grain for each table. For every field, review its Alias, Description, Role, and Review status alongside the physical database name. - Keep physical names visible for traceability. - Use approved aliases and descriptions in analysis. - Mark identifiers, dimensions, measures, and timestamps by role. Joins and calculations ## Approve how data connects—and how the business calculates. Relationships tell Better Analyst which tables can be joined. Metrics give recurring calculations a shared name and definition. 1. ![Relationship editor connecting the coffee sales State field to the earthquake state field with one-to-one cardinality](https://betteranalyst.com/screenshots/product/data-context/relationships-v2-light.webp?dpl=c3a99e2b7961f51c9b379add33be693e641879a6)![Relationship editor connecting the coffee sales State field to the earthquake state field with one-to-one cardinality](https://betteranalyst.com/screenshots/product/data-context/relationships-v2-dark.webp?dpl=c3a99e2b7961f51c9b379add33be693e641879a6) Swipe to inspect the interface **Relationships**Define the source, destination, join fields, and cardinality so Better Analyst knows how records connect. 2. ![Metric editor defining Stores as a distinct count of the coffee sales Area Code field](https://betteranalyst.com/screenshots/product/data-context/metrics-v2-light.webp?dpl=c3a99e2b7961f51c9b379add33be693e641879a6)![Metric editor defining Stores as a distinct count of the coffee sales Area Code field](https://betteranalyst.com/screenshots/product/data-context/metrics-v2-dark.webp?dpl=c3a99e2b7961f51c9b379add33be693e641879a6) Swipe to inspect the interface **Metrics**Publish simple aggregations and derived calculations such as Net Revenue with descriptions and formats. Documents and SQL feedback ## Add the meaning schemas miss. Teach from corrections. Upload business vocabulary and schema notes, then save SQL feedback from chat to steer future queries. 1. ![Uploaded Docs screen for adding schema notes, data dictionaries, and business vocabulary](https://betteranalyst.com/screenshots/product/data-context/uploaded-docs-v2-light.webp?dpl=c3a99e2b7961f51c9b379add33be693e641879a6)![Uploaded Docs screen for adding schema notes, data dictionaries, and business vocabulary](https://betteranalyst.com/screenshots/product/data-context/uploaded-docs-v2-dark.webp?dpl=c3a99e2b7961f51c9b379add33be693e641879a6) Swipe to inspect the interface **Uploaded Docs**Add schema notes, data dictionaries, and vocabulary that inform field drafting and answers. 2. ![SQL Feedback screen explaining how reviewed query annotations steer future SQL](https://betteranalyst.com/screenshots/product/data-context/sql-feedback-v2-light.webp?dpl=c3a99e2b7961f51c9b379add33be693e641879a6)![SQL Feedback screen explaining how reviewed query annotations steer future SQL](https://betteranalyst.com/screenshots/product/data-context/sql-feedback-v2-dark.webp?dpl=c3a99e2b7961f51c9b379add33be693e641879a6) Swipe to inspect the interface **SQL Feedback**Save correct or incorrect query annotations from chat so reviewed feedback can steer future SQL. A controlled meaning layer ## Describe connected data without changing the data itself. Data Context works as a lightweight semantic layer for teams without a data team. Data Controls keep excluded data out of Data Context and AI. Drafts, published revisions, and history make it clear which definitions are active. On a team source with per-member access, a member who has tables or columns hidden from them doesn't get that source's Data Context, uploaded documents, or SQL feedback, since those can name what's hidden. Data Context describes connected data; it is not the dataset and it does not replace AI Agents. How analysis should run ## Need to configure the analyst too? AI Agents define analysis instructions, Business Context, output preferences, and presentation defaults. [Explore AI Agents](https://betteranalyst.com/product/ai-agents) FAQ ## Data Context questions What a semantic layer is, which data it covers, and who publishes definitions. Is Data Context a semantic layer? Yes, a lightweight one for teams without a data team. A semantic layer maps raw tables and columns to business terms, metrics, and approved joins so questions start from the same definitions. Data Context does that for connected SQL databases: describe tables and fields, state what one row represents, approve relationships, and publish metrics. There is no modeling code to write, and your data is never changed. Which data sources does Data Context support? Connected SQL databases: BigQuery, Snowflake, Supabase, MySQL, PostgreSQL, Microsoft SQL Server, Amazon Redshift, and Databricks. Google Analytics, Search Console, and uploaded files are analyzed in chat but do not have Data Context definitions yet. For those, state your definition in the question or in an AI Agent’s instructions, and check the code shown with the answer. Does Data Context make every answer correct? No tool can promise that. Data Context removes the most common guesses, such as what a metric means, what one row represents, and how tables join, so analysis starts from your team’s definitions. Every answer still shows the SQL it ran, so you can check the numbers that matter. Do we have to write every definition by hand? No. Better Analyst can draft table and field descriptions from your schema and from documents you upload, such as a data dictionary or schema notes. You review the drafts, edit what is wrong, and publish what you trust. Who can change published definitions? Edits start as a draft. Workspace owners and admins publish them, and each published revision is kept in history so you can see what changed and restore an earlier version. How does Data Context work with Data Controls? Data Controls decide which tables, columns, and rows Better Analyst can query on a database connector. Excluded data stays out of Data Context and AI. If access settings or the schema change after you publish, the published definitions are withheld until they are published again. Can Better Analyst reuse queries we trust? Yes. With SQL Feedback, endorse a query from chat and it is reused as an example when similar questions come up on that database. Which plans include Data Context? Data Context definitions and metrics are included on every paid plan. Compare your options ## Decide with evidence, not a sales pitch. Open a neutral research prompt in the AI tool you already use. Compare capabilities, limitations, and fit before you choose. 1. [01ChatGPTOpen a balanced comparison prompt in ChatGPT.](https://chatgpt.com/?prompt=Compare%20Better%20Analyst%20with%20using%20a%20general-purpose%20AI%20assistant%20for%20data%20analysis.%20Evaluate%20spreadsheet%20and%20document%20analysis%2C%20dashboards%2C%20repeatable%20reports%2C%20code%20transparency%2C%20data%20connections%2C%20pricing%2C%20and%20limitations.%20Use%20current%20public%20sources%2C%20distinguish%20verified%20facts%20from%20assumptions%2C%20and%20explain%20which%20types%20of%20users%20each%20option%20fits%20best.) 2. [02ClaudeOpen the same research prompt in Claude.](https://claude.ai/new?q=Compare%20Better%20Analyst%20with%20using%20a%20general-purpose%20AI%20assistant%20for%20data%20analysis.%20Evaluate%20spreadsheet%20and%20document%20analysis%2C%20dashboards%2C%20repeatable%20reports%2C%20code%20transparency%2C%20data%20connections%2C%20pricing%2C%20and%20limitations.%20Use%20current%20public%20sources%2C%20distinguish%20verified%20facts%20from%20assumptions%2C%20and%20explain%20which%20types%20of%20users%20each%20option%20fits%20best.) 3. [03PerplexityResearch the comparison with linked web sources.](https://www.perplexity.ai/search?q=Compare%20Better%20Analyst%20with%20using%20a%20general-purpose%20AI%20assistant%20for%20data%20analysis.%20Evaluate%20spreadsheet%20and%20document%20analysis%2C%20dashboards%2C%20repeatable%20reports%2C%20code%20transparency%2C%20data%20connections%2C%20pricing%2C%20and%20limitations.%20Use%20current%20public%20sources%2C%20distinguish%20verified%20facts%20from%20assumptions%2C%20and%20explain%20which%20types%20of%20users%20each%20option%20fits%20best.) AI answers can be incomplete. Check source links, current pricing, and product documentation before deciding. A shared SQL vocabulary ## Teach AI what your SQL data means. Connect a supported SQL database, review its definitions, and publish the context you trust. [Set up Data Context](https://app.betteranalyst.com/dashboard/library?tab=connectors)[Start free](https://app.betteranalyst.com/signup)