Documentation
Welcome to Tattva.See how every number is made.
Tattva answers business questions from your data, computes the numbers in code from definitions your data analysts approved, and shows its working.
Browse by section
- 01Getting startedQuickstart · Create a project · Connect data sources · 1 more
- 02For data analystsYour role and the lifecycle · Metric rules: the concepts · One metric, end to end · 16 more
- 03Core conceptsProjects · Workspace library · Sources · 9 more
- 04WorkflowsAsk a question · Review an answer · Track a decision · 2 more
- 05Use casesUse cases
- 06Trust & administrationWorkspaces & members · Permissions · Account & profile · 9 more
- 07ReferenceFAQ · Supported data sources · Keyboard shortcuts · 3 more
Tattva answers questions about your business data. You ask in plain English, and you get an answer with the figures, the reasoning behind them, and a trail you can check line by line.

What makes the answers trustworthy
Your data analysts write down how each metric is calculated: net revenue, RTO rate, repeat rate, AOV. Tattva stores those definitions, tests them against your warehouse, and uses them in code whenever a question needs that metric. Code computes the figures and lays out the answer's tables, notes and chart; the language model reads your question and writes the prose about the figures, but it does not make them up. Before you see an answer, a second model lists every claim the prose makes and code checks each one against the figures: a sentence that fails is revised once, then dropped. The Trail tells you which definition each figure came from.
Who does what
Business users ask and act:
- Ask questions in Reasoning and check the answer on its Trail.
- Answer a clarification when a question is ambiguous.
- Keep answers as trackers, reports and boards, and set goals.
- Look up how a metric is defined before quoting it.
Data analysts make the answers right:
- Connect BigQuery tables and describe what columns mean.
- Write metric documents, review the definitions Tattva proposes from them, and accept the ones that pass their checks.
- Group metric variants into families, and say how Tattva should handle each family.
- Keep the ontology healthy as sources change.
If that second list is you, start at For data analysts.
What you can do with it
- Ask questions of your data: sales, marketing, customer, product or operations.
- Dig into why: root cause, forecast, attribution and allocation analyses on demand.
- Watch what matters: turn a question into a tracker that re-runs on a schedule and flags changes.
- Look ahead: research risks and opportunities with Foresight.
- Keep a record: decisions, their assumptions and their outcomes, linked to the data.
Where to start
Quickstart
Sign in, create a project, add data and ask your first question.
For data analysts
Connect data, write metric documents, review definitions and keep the ontology healthy.
Ask a question
The full workflow, from phrasing a question to keeping the answer.
Review an answer
How to check the figures on the Trail tab.
Core concepts
Projects, sources, the ontology, trackers, goals, boards and reports.
Use cases
Planned: worked examples that need NRoAS, CAC, marketing % and target definitions.