Orion vs Sigma

Every report, on schedule.Click a number to see why it moved.

Sigma is the spreadsheet-style workbook that a team builds on the warehouse. Orion runs the report, checks the number, and explains what changed.

An Orion dashboard for a marketing team. It shows sourced pipeline against plan by month. The gap widens to 81% of plan in October.

Orion detected a change

Pipeline fell to 81% of plan in October, the widest gap since the LinkedIn forms launch.

Ask Orion why

Trusted by teams at

DaveIbottaExcess TelecomDMI Partners

What is the difference between Orion and Sigma?

Sigma is where analysts build with live data. Orion runs the recurring report, detects significant changes, and keeps the notebook behind each analysis.

Orion

Orion is the analytics platform for a team's tools and its warehouse: one set of definitions, every number traceable, and the report, deck, or alert delivered where people already work. It reads the semantic layer that the team maintains, such as LookML or dbt, or works from the warehouse itself. Saved metrics run again on a schedule, and Workflows deliver the same report every week with no workbook to rebuild.

Sigma

Sigma is a BI tool on live warehouse data. Its spreadsheet interface lets analysts explore, build dashboards and AI apps, and write data to the warehouse. Sigma Assistant builds dashboards from prompts, and agents that a team builds run tasks on a schedule.

How do Orion and Sigma compare feature by feature?

CriteriaOrionSigma
Proactive investigationOrion runs each tracked metric on its schedule and detects significant changes and anomalies. A Workflow decides whether the numbers are worth flagging and delivers the report to Slack or email. Detection is automatic.Sigma sends scheduled alerts on set conditions. Agents that a team builds per workbook can detect and explain anomalies in their data.↗↗
One Room for the whole teamEveryone posts in one Room and sees the same answer, and every message shows its author. An @Orion mention turns a Slack thread into a Room. Replies land in the Room, and Orion's answers post back to the thread.Live Edit lets analysts share one workbook draft. When chat history is on, each user sees their own recent Sigma Assistant and agent chats.↗↗↗
Semantic layer reuseOrion reads LookML and dbt directly from their repositories. It reuses the dimensions, measures, and models that a team already defined to write better SQL. Without LookML or dbt, Orion works from the warehouse schema and its query history.A team builds a Sigma data model, or an AI assistant with a migration skill rebuilds LookML as a Sigma data model. Sigma's dbt Semantic Layer connection works with dbt Cloud only.↗↗
Who builds the analysisOrion builds the analysis. It investigates why a number moved, writes the report, and a Workflow delivers the report to Slack or email.The team builds the analysis in spreadsheet-style workbooks and AI apps, or asks Sigma Assistant.↗↗
Spreadsheet-style exploration and app authoringOrion delivers reports, interactive dashboards, and slide decks, built in an investigation and kept current on a schedule. A plain-language request changes a dashboard, and one click on a number shows why it moved.Sigma gives users a spreadsheet-style workbook on live warehouse data. Analysts build dashboards, AI apps, and write-back by hand, or Sigma Assistant builds them from a prompt.↗↗↗

Orion vs Sigma: common questions

Orion adds the investigation. Sigma is where a team builds spreadsheet-style workbooks, dashboards, and AI apps on live warehouse data, with Sigma Assistant for natural-language questions. One click on an Orion dashboard number shows why it moved. A Workflow runs the analysis on a schedule and delivers the report to Slack or email without anyone opening Orion. The recurring 'why did this move' questions no longer become new Sigma workbooks.

The difference is how a team controls the analysis behind the answer.

Orion holds one set of definitions in its Metrics Catalog and Context pages, and every MCP client reads the same values. Each analysis keeps its notebook with the instructions, the logic, and the code. Warp Drive can limit answers to approved content. A Workflow delivers a report, a dashboard, or a slide deck on a schedule. Orion's MCP server brings those definitions, analyses, and deliverables into ChatGPT, Claude, Cursor, and Gemini Enterprise.

No. Sigma Assistant, formerly Ask Sigma, answers questions about data and builds dashboards and apps from prompts. Sigma agents do specific tasks that a team configures in a workbook. Sigma Assistant works on request. Agents work on request or on a schedule that the team configures.

Orion changes what the team builds. The team builds the context once: the semantic layer that Orion reads and the cited Context pages that any project can use. From that context, Orion builds the analyses and writes the report on what it found and why. A Workflow delivers the report without anyone opening Orion. Sigma Assistant helps a person in Sigma get an answer faster. Orion detects the change that nobody asked about.

Partly. Sigma has conditional exports and alerts that run threshold and formula checks on a schedule and send the result to email, Slack, Teams, and webhooks. A team builds each Sigma agent in a workbook, gives it instructions and data sources, configures a notification, and schedules it. A well-built agent detects anomalies in that data and explains them. Each Sigma agent covers the task and the data that its builder chose.

Orion detects significant changes across the project's metrics, including the changes that nobody configured a check for. The team writes no agent. A Workflow decides whether the numbers are worth flagging and delivers the report to Slack or email.

Partly. Sigma converts LookML and reads dbt metadata. Sigma builds its own semantic layer in Sigma data models. A Looker migration skill uses an AI assistant to rebuild LookML as a Sigma data model. Its dbt integration shows dbt metadata from dbt Cloud or dbt Core, but the dbt Semantic Layer connection works with dbt Cloud only.

Orion reads the LookML and dbt that a team already maintains, from dbt Cloud or dbt Core. It reuses those dimensions, measures, and joins to write better SQL. Cited Context pages add the rest. Without LookML or dbt, Orion works from the warehouse schema and its query history.

Sigma Assistant answers from any data model or warehouse table that a user can access. An admin can limit it to specific data models or tables. Sigma applies column-level and row-level security on its data models.

Orion answers from the semantic layer that a team already owns, such as LookML or dbt, plus cited Context pages. Orion keeps each analysis in a notebook with the instructions, the logic, and the code. A user can ask Orion to change the query and run it again. On a dashboard, Lineage shows where each card gets its data. Orion issues only read queries. It connects through a dedicated service account that the team scopes to the data Orion can analyze, with optional per-user OAuth on BigQuery. Roles on two levels, tenant and group, control who sees which projects and data.

Orion separates the questions that a team asks every week from the questions that it asks once. Saved metrics run again on a schedule and keep their history. The Metrics Catalog holds one definition that any project can reuse. Views save a dataset as a base, so repeat analyses start from the same rows. Warp Drive is fast because Orion answers only from the content that the team approved. Warp Drive shows a rejected draft as a draft, not as an answer.

For a new question, the full notebook records the instructions, the logic, and the code. A user can ask Orion to change the query and run it again.

Yes. A Room lets several people work on one question with Orion in a single chat. Each message shows its author. A business user can join the Room as a Viewer and follow the investigation.

Yes. An embed serves Orion dashboards, reports, and slide decks inside the company's own site, with no Orion login. A protected portal serves the same content at its own link, behind the company's SSO or a shared password. A company can white-label its deployment with its own logos.

Each customer gets its own group. A group keeps its projects and data isolated from other customers in the same tenant, with its own data sources, Context pages, and Integrations. The Embedded Orion page has the details.

Yes. The Gravity team migrates a team's dashboards from Sigma or any other BI tool into Orion. Dashboards become projects, calculations become metrics, and verified metrics go into the Metrics Catalog.

See what your team looks like with an AI Analyst.