Use the Studio workflow
Navigate the lifecycle overview, manage experiments in Pipeline or Table view, and create a governed experiment with the staged builder.
Studio is the organization-scoped workspace for planning, launching, monitoring, and closing experiments. Its visual workflow keeps the governed domain states intact while presenting them as six easy-to-scan stages.
The production workspace is available at fabric-experiments-prod. Databricks Apps authentication is required before the Fabric Experiments login screen appears.
Find your way around
| Surface | Route | Use it to |
|---|---|---|
| Overview | /app/{orgSlug} | See portfolio health, work needing attention, lifecycle counts, and recent activity. |
| Experiments | /app/{orgSlug}/experiments | Move between the visual Pipeline and dense Table views. |
| Create | /app/{orgSlug}/experiments/new | Build a draft in five guided sections. |
| Experiment detail | /app/{orgSlug}/experiments/{id} | Inspect lifecycle position, variants, results, actions, and audit history. |
| Install | /app/{orgSlug}/install | Generate the tag-loader installation snippet. |
| Audit | /app/{orgSlug}/audit | Review organization-wide activity. |
| Settings | /app/{orgSlug}/settings | Manage API keys and organization settings. |
The active organization appears once in the application header or sidebar. Returning users with an existing membership are taken directly to their organization instead of being asked to create it again.
Start from Overview
Overview answers four operational questions at a glance:
- All experiments — the complete organization portfolio.
- Live or paused — experiments currently serving or temporarily held.
- Needs attention — review and paused experiments that require a human decision.
- Evidence captured — concluded or archived experiments with a durable outcome.
The dotted Portfolio by stage map shows the same six columns used by Pipeline. The action queue pulls forward review and paused work, while recent activity links directly to the latest experiment records.
Understand the six stages
The UI groups closely related domain states without changing the API state machine:
| Visual stage | Domain states | Meaning |
|---|---|---|
| Design | draft | Define the hypothesis, audience, variants, and success metrics. |
| Review | review | Validate the specification and request an independent approval. |
| Ready | approved | Approved and ready to start when the launch window opens. |
| Live | running, paused | Serve variants, monitor exposure quality, or hold traffic temporarily. |
| Decision | concluded, killed | Record a winner or stop the experiment with an auditable reason. |
| Archive | archived | Preserve the final record as durable organizational evidence. |
On an experiment detail page, each stage is labeled Complete, with the current state name, or Upcoming. Status is always conveyed with text as well as color.
Work in Pipeline or Table view
Pipeline is the default view. Each experiment card shows its state, last update date, name, and stable ID under one of the six lifecycle columns.
Use Table when you need a compact operational inventory. Switching views does not modify experiments or their lifecycle states.
Both layouts are responsive. Pipeline becomes a single-column sequence on small screens so the lifecycle stays readable without horizontal scrolling.
Create an experiment in five sections
Select Create experiment from Overview or Experiments. The builder keeps every field on one page while the dotted stage navigation provides quick anchors:
- Basics — name, stable ID, description, sample rate, holdback, salt, diversion, and manual exposure.
- Activation — choose an automatic, URL, selector, or event-based trigger; add shared code and metadata when needed.
- Variants — define keys, display names, weights, recipe sampling, payloads, DOM operations, and advanced JS/CSS.
- Metrics — add conversion or guardrail events and choose the primary outcome.
- Review — inspect the generated specification before creating the draft.
All sections remain mounted while you navigate, so browser validation can lead you back to any required field. The generated specification is the same shape used by YAML and GitOps workflows.
Promote and monitor the experiment
After creating the draft, use the actions on its detail page to move through the governed state machine:
The detail page keeps lifecycle position, variants, audience, result summaries, available actions, and audit history together. Preview buttons create short-lived signed preview links and do not expose a persistent secret.
Continue with YAML or automation
The visual workflow and code-driven workflow converge on the same domain actions. You can:
- import one or more files from
/app/{orgSlug}/experiments/import, - edit a Studio-created experiment as YAML,
- create an API key under Settings and use
fx push, or - call the hosted API from CI.
See Create your first experiment for an end-to-end example and Studio UI for implementation and security boundaries.