Skip to content
Docs

Getting started

Four steps take you from an empty workspace to your first answer. Each one builds on the last, connect the data, teach Adverity Atlas your language, add the context behind it, then start asking.

  1. Connect: Data sources

  2. Configure: The glossary

  3. Add: Knowledge

  4. Ask: Questions about your data

Note

Answer quality depends on two things, the data you connect and the business context you configure. Steps 2 and 3 are what make Adverity Atlas understand your data rather than generic SQL.

Step 1: Connect data sources

Adverity Atlas queries your existing warehouse with read-only access. It does not write to your tables or replicate them.

Seven connectors are supported:

  • Google BigQuery: Service account, workload identity federation, or Google sign-in

  • Snowflake: Password, key pair, or Snowflake OAuth

  • PostgreSQL: Username and password

  • Amazon Athena: IAM role, access keys, or assume role

  • Databricks: Access token, OAuth M2M, or Azure service principal

  • Amazon Redshift: Password, or IAM in three forms

  • Trino / Starburst: Password, JWT, or OAuth 2.0

Files are data sources too. Upload CSVs and spreadsheets for anything that lives outside the warehouse, budgets, targets, campaign plans.

  1. Choose a connector, pick the one your warehouse uses.

  2. Set up read-only credentials, each connector offers several authentication methods. Adverity Atlas only ever needs read access.

  3. Bind to workspaces, connect as many sources as you need and decide which workspaces can query which. Schema discovery and field profiling then run automatically in the background.

Tip

Each connector has its own page with the exact setup steps, required cloud roles, and every connection field. See Data sources.

Step 2: Configure the glossary

The glossary aligns the language used across your platforms and your organization. It is what turns “what was our ROAS last month” into the right columns and the right arithmetic.

It operates at three levels, and the most specific one wins:

  • Standard: A standard set of marketing terms maintained by Adverity. You cannot delete these, but you can override them

  • Organizational: Your company’s shared vocabulary, applied across every workspace. Use it for company-wide metrics

  • Workspace: Local overrides, for when the same word means different things to different teams

Define canonical terms, attributes such as currency, calculated KPIs, and synonyms. Add synonyms generously: if your team says both “CVR” and “conversion rate”, add both.

  1. Review what is already there, open Glossary from the workspace navigation. Connecting a data source proposes concepts for you, so the list is rarely empty.

  2. Work through the drafts, set the Status filter to Draft, open each one, and click Activate on the ones that are right. A draft is invisible to queries until it is activated, so this is the step that makes them count.

  3. Add what is missing, click Add Concept, enter a Display Name and Definition, and set the Category.

  4. Add the aliases, list every word your team uses for the term under Aliases. A question using an alias you skipped will not resolve.

  5. For a calculated KPI, add the Expression, the formula referencing other canonical names. Adverity Atlas previews it as you type.

Note

When you connect a data source, Adverity Atlas automatically classifies its columns against the glossary and proposes concepts as Draft. Reviewing and activating those drafts is the fastest way through this step.

Step 3: Add knowledge

Some context does not fit a term definition. Knowledge is the one central place for it, the rules, guidelines, and background that would otherwise live in Slack threads and people’s heads.

  • Brand guidelines: Voice, visual identity, and messaging rules, so generated outputs match your standards

  • Business rules: Attribution models, calculation methodologies, reporting conventions

  • Conversations and insights: Context from Slack, meeting notes, and past analyses

  • Data interpretation: Known quirks, seasonal patterns, and what “normal” looks like

  1. Open Knowledge from the workspace navigation.

  2. Start a note, click New and choose Note.

  3. Write it, give it a Title, a one-line Summary, and the detail in Content. Markdown works in the body.

  4. Decide how it reaches Chat, leave Agent context on Auto so the item is retrieved when it is relevant. Reserve Always for the handful of facts that shape every answer.

  5. Save note. The item is usable straight away, it does not need confirming the way an extracted item does.

  6. Confirm what Adverity Atlas found, profiling and past conversations propose items of their own. Work through the Detected ones and Confirm the ones that hold.

Tip

Glossary covers individual terms; Knowledge covers everything broader. If a definition alone cannot answer “why does this number look like this”, it belongs here. See Knowledge.

Step 4: Ask questions about your data

With data connected and context configured, open Chat. It is the hub of the product: ask questions, and also create automations, generate artifacts, and reach every other feature from the same conversation.

  • Performance: “What was our total ad spend last month?”

  • Trends: “Show me spend trends for the last 6 months by channel”

  • Exploration: “What data do I have available?”

  • Comparison: “Compare Meta vs Google performance this month”

Conversations are multi-turn. Follow-ups inherit context, so “now just for last week” or “make that a bar chart” work without restating the question.

  1. Open Chat and type a question in plain language, then send it. Adverity Atlas resolves the terms, generates and runs the SQL, and returns a chart or table.

  2. Narrow the scope if you need to, type @ to point the question at a specific data asset or uploaded file.

  3. Drill in with follow-ups, start broad and refine, each message inherits the context of the last.

  4. Check the working, click a step pill beneath the answer to open Data Provenance and see the query that produced it.

  5. Keep what is worth keeping, ask for an artifact to share the result, or click Create Automation to turn the conversation into something that runs on a schedule.

Tip

If Adverity Atlas misreads a term, the fix usually belongs in the glossary rather than the question, correcting it there fixes every future conversation. See Chat for @mentions and output options.

What’s next?

  • Chat, tips for better answers, @mentions, and output formats

  • Data sources, per-connector setup, schema discovery, and profiling

  • Glossary, concept types, overrides, and how definitions reach Chat

  • Automations, turn a recurring question into a schedule

Was this page helpful?