Skip to content
Shuttlebase Docs

Experiment Lifecycle

Every experiment on your account has a status. The status tells you where the experiment stands: being built by Shuttlebase, waiting for your review, live on your site, or finished. This page lists every status and walks through the typical path from draft to deployed winner.

The experiments list grouped by stage: running experiments, then paused, then deployed.

Status reference

Section titled “Status reference”
StatusStageWhat it means
DraftBuildShuttlebase is building and testing the experiment. It has not been sent to you yet, and you do not need to act.
Pending ApprovalReviewShuttlebase sent you a launch request. You can approve it, request changes, or reject it.
Pending ModificationsReviewYou requested changes. Shuttlebase is making them. The experiment goes through QA again, then you receive a new approval request.
Approved: Pending LaunchReviewYou approved the experiment. It sits in the launch queue and starts automatically.
Blocked by Live ExperimentReviewA live experiment touches the same pages. This one waits until that experiment finishes.
Awaiting CapacityReviewYour account is already running its maximum number of concurrent experiments. This one launches when a slot opens.
On HoldReviewThe experiment is held at your request. It stays out of the launch queue until you release it.
RunningLiveThe experiment is live and splitting traffic between its variants.
PausedLiveThe experiment is temporarily stopped. It can resume at any time.
EndedOutcomeThe experiment stopped without rolling out a winner.
DeployedOutcomeThe winning variant now serves to 100% of your traffic.
Deployed Paused / Deployed EndedOutcomeThe winning variant was deployed, then later paused or stopped.
Implemented by ClientOutcomeYour team hard-coded the winning change into your own site, so it no longer needs to run through the snippet.
CanceledOutcomeThe experiment was stopped before it ever launched.

The typical path

Section titled “The typical path”

Most experiments follow the same route.

  1. Shuttlebase builds it. The AI creates the concept and implements the variant code. The experiment then passes an automated QA gate before you ever see it. During this phase it shows as Draft. See Variants and QA for how the gate works.
  2. You review it. Once QA passes, you get a launch request and the experiment moves to Pending Approval. You approve, request changes, or reject each variant. If you request changes, the experiment moves to Pending Modifications, gets fixed, passes QA again, and comes back to you as a fresh request. See Reviewing and Approving.
  3. It waits its turn. Approved experiments enter the launch queue as Approved: Pending Launch. If something is in the way, the status shows why: Blocked by Live Experiment or Awaiting Capacity.
  4. It runs. The experiment goes live as Running and splits traffic between variants. You can pause and resume it at any time.
  5. It concludes. A winning experiment gets deployed to all traffic, a losing one gets ended, and if your team codes the change into your site directly, you can mark it Implemented by Client. See Experiment Terminal States and Reading Results.

How launch timing works

Section titled “How launch timing works”

You never pick a launch date. Once you approve an experiment, a scheduler launches it automatically based on three things.

  • Priority. Every experiment has a priority of high, medium, or low. Higher priority experiments launch first.
  • Capacity. Your account has a cap on how many experiments run at once. Experiments over the cap wait as Awaiting Capacity.
  • Collisions. Two experiments that change the same pages never run at the same time. The later one waits as Blocked by Live Experiment.

Shuttlebase has no calendar scheduler, so you cannot book a launch for a specific date. If you need to delay a launch, ask Shuttlebase to put the experiment on hold. It will stay On Hold until you release it.