Experiment Statuses

Every experiment in A vs B has a status that tells you where it is in its lifecycle. There are five: Draft, Scheduled, Running, Paused, and Completed. Knowing what each one means, and what you can and cannot do in each, helps you manage your experiments with confidence.

Archiving is separate from all five. It files an experiment out of your main list without changing its status, and restoring brings it back exactly as it was filed. See Archiving and restoring below.

The builder header shows only the actions that make sense for the current status: a Draft shows Publish and Schedule (both need the pre-launch checklist to pass first), a Running experiment shows Pause and Stop, and a Paused experiment shows Resume and Stop. Scheduled and Completed experiments show no launch buttons at all. Every one of these actions needs the Publish to production permission.

Statuses at a glance on the Experiments list: the tabs filter by lifecycle state, and each row wears its status badge.

The list keeps itself current. When you come back to it with the browser's Back button, or return to its tab after a while, it reloads the rows, so a test you just paused shows as Paused without a refresh.

Draft

A new experiment starts in Draft status. A draft experiment is:

  • Not running: no visitors are being sorted into a variation (one specific version being tested, like "Control" or "Blue button").
  • Editable: you can change any part of the experiment freely.
  • Not visible to visitors: the snippet (the small script A vs B installs on your site) ignores draft experiments entirely.

An experiment stays in Draft until you publish it, and there's no time limit: it can sit in Draft indefinitely while you build and refine it.

Moves to: Running (when you publish it), or Scheduled (when you set a future launch time).

Scheduled

A Scheduled experiment is set to launch on its own at a future moment. No visitors are bucketed yet: it behaves like a draft until the scheduled time arrives.

  • Hidden from visitors, the same way a draft is.
  • The schedule banner on the experiment page shows the launch moment in your local time, the time zone of the person who scheduled it, and UTC.
  • You can edit or cancel the schedule from that banner at any point before it fires.

When the scheduled moment arrives, the experiment moves to Running and the person who scheduled it gets a notification. See Scheduling Experiments for the full mechanics, including what happens if the platform misses the exact moment.

Moves to: Running (when the scheduled time arrives), or Draft (when you cancel the launch schedule and nothing else is scheduled).

Running

A Running experiment is live. Every eligible visitor who lands on a targeted page gets evaluated, bucketed into a variation, and shown it. Events are tracked and results build up on the dashboard.

  • Visitors are being bucketed and shown variations.
  • Results update within a few minutes, not instantly.
  • Some edits are possible but should be made carefully; see Managing Running Experiments.

Moves to: Paused, or Completed (stopped).

Paused

A Paused experiment is temporarily halted. No new visitors are bucketed, but every visitor already collected stays in the data, and you can resume at any time.

  • Visitors see the original page again: the snippet skips a paused experiment.
  • Everything collected before pausing is kept.
  • You can freely edit a paused experiment, since it isn't live and there's no running data to disturb.
  • Resume is one click away from the row menu on the experiments list and on the dashboard, as well as from the experiment itself. If the experiment holds changes you saved but never published, A vs B asks what should happen to them before anything goes live.

Pause an experiment when you find a bug in your variation code, you're making a major site change, or you want time to study results before deciding whether to keep going.

Moves to: Running (resumed), or Completed (stopped).

Completed

A Completed experiment has stopped running. No visitors are bucketed, and its full results history stays in place permanently.

  • Variation code stops running for every visitor.
  • All data stays saved and viewable on the results page.
  • The experiment can't be resumed directly. To run the same test again, use Duplicate, which gives you a fresh draft of it.

Completed is what every screen calls this status: the list, the experiment page and the builder header all say Completed.

An experiment reaches Completed in two ways:

  • Stop, from the builder, the results page, or the row menu of a running or paused experiment on the experiments list. Stop always asks you to confirm first, because it can't be undone.
  • Frequentist experiments (see Frequentist Engine) get an extra check on it: if you stop before the experiment reached its planned sample size, A vs B asks you to confirm and log a reason first, because stopping early skews the statistics.
Completed is a one-way trip out of the running list

An experiment can't go back to Running from Completed. To run the same test again, use Duplicate: it gives you a fresh draft with the same variations, code, audiences, metrics and settings, and leaves the finished experiment and its results exactly as they are. If you might want to pick a test back up later and keep collecting, consider Pausing instead of Stopping.

Archiving and restoring

Archiving is filing, not a change of status. It's how you get a finished, paused or draft experiment out of your main list once you're done looking at it.

An archived experiment:

  • Leaves the main list and appears under the Archived filter tab, and leaves the dashboard's recent experiments.
  • Keeps its real status. A completed test still says Completed, a paused one still says Paused, a draft still says Draft, shown in grey with an Archived tag beside it.
  • Keeps everything else too: its results, its dates and its history.

Restore puts it back exactly where it was. A completed test comes back Completed with its results and its completion date. A paused test comes back Paused, with Start waiting for you if you want it running again. Nothing ever comes back Running on its own: restoring never starts sending traffic to visitors.

A running or scheduled experiment has to stop first

Archiving must never be the thing that changes what your visitors see, so a Running or Scheduled experiment can't be archived as it is. Choosing Archive on one tells you so and offers the ways forward in the same place: Pause it to keep its place, Stop it to finish it, or cancel the schedule. Then archive it.

Archive and Restore both live in the three-dot row menu on the experiments list, and on the experiment page itself. Opening an archived experiment's own setup pages says the same thing: the builder header shows the Archived tag beside its status, Publish, Schedule and Resume are switched off with the reason beside them, and Restore is right there in the header. Finished experiments you have not archived stay in your main list as Completed: nothing is filed away until you file it.

Status summary

StatusVisitors bucketed?Data collected?Editable?How it moves onCan it be archived?
DraftNoNoYes, freelyPublish → Running, or set a schedule → ScheduledYes, and it comes back a Draft
ScheduledNoNoSchedule only (edit or cancel)Fires → Running, or cancel → DraftNo, cancel the schedule first
RunningYesYesYes, with carePause → Paused, or Stop → CompletedNo, pause or stop it first
PausedNoNo (kept)Yes, freelyResume → Running, or Stop → CompletedYes, and it comes back Paused
CompletedNoNo (kept)NoDuplicate → a new DraftYes, and Restore keeps it Completed
Was this helpful?