Experiment Lifecycle

Every experiment in Split Test Pro moves through a small set of states. Knowing what each state does — especially to traffic and data collection — saves you from accidentally invalidating a test with the wrong click.

The Six States

State	What it means	Variants shown?	Data collected?
Draft	You’re still setting it up.	No	No
Scheduled	Set to launch at a date and time you picked.	No	No
Running	Live. Visitors are assigned and tracked.	Yes	Yes
Paused	Stopped for now. You can resume from the same point.	No	No
Completed	Ended. Results are kept as a final snapshot.	Only while a winner is rolled out	No
Archived	Hidden from your active list.	No	No

A completed test serves its variant again only when you roll a winner out to all visitors — see Winner Rollout.

State Transitions

Draft → Running

On the last step of the wizard, Variants, click Launch Test. The button stays disabled until the launch checks pass. When it succeeds you see a Launched toast and the experiment opens on its Results tab. Visitors who match your targeting start entering the test on their next page view.

Draft → Scheduled → Running

To launch later, answer When should it run? on the Variants step with Schedule, pick a launch time, and click Launch Test. The experiment becomes Scheduled. You can also click Schedule next to a draft on the Dashboard. At the launch time, Split Test Pro runs the same launch checks and starts the test. See Scheduling Experiments.

Running ⇄ Paused

Click Pause. The confirmation reads: “Sessions stop being recorded immediately. You can resume from this same point at any time.” Click Resume to go live again. If something required is missing, Resume opens Finish setup to resume instead and lists what to fix.

You can edit a test’s setup only while it is a Draft or Paused. On a running test, the Setup tab tells you to “Pause the test to edit.”

Running or Paused → Completed

Click End test in the ⋯ menu, or Declare winner on the Results tab when a variant is winning. Both open the same confirmation: “Visitors stop entering the test and everyone sees the original page. The results so far are kept.” The Results tab then shows a Final snapshot note: the test has ended, so its results are frozen. See Declaring a Winner.

A running test can also end on its own:

• A scheduled end date. When the end time you set passes, the test is completed with a final snapshot. Pausing a test removes its end date, so set a new one after you resume.
• Autopilot. When Autopilot manages your test plan, it can complete tests for you — see Autopilot.

Any State → Archived

Click Archive in the ⋯ menu to hide an experiment from the active list. Its results are kept. Archiving a scheduled test cancels its schedule.

Unarchive brings it back: a test that had ended returns as Completed, a test that ran but hadn’t ended returns as Paused, and a test that never launched returns as a Draft. Unarchiving never makes a test live again — you choose when to resume.

Completed vs Archived

The two states answer different questions:

• Completed is about the experiment — it ran, it ended, the results are final.
• Archived is about your list — hide tests you no longer need to see. Most teams archive after completing, but you can archive a draft you decided not to ship, too.

Duplicating an Experiment

To start a new test from an existing one — running, completed or archived — open it and choose Duplicate in the ⋯ menu. The dialog asks for a name for the copy: “Variants, targeting, and goals will be copied; results will not.” The copy is a new Draft. On the Experiments list, a row’s Duplicate acts at once and creates the draft without asking for a name.

This is the right pattern for iterating on a hypothesis: don’t restart a completed test — duplicate it and change the copy.

Common Mistakes

• Pausing instead of ending a finished test. A paused test stays in your active list and looks unfinished. When you’ve decided, click End test or Declare winner.
• Resuming a paused test with changed variants. That’s effectively a new experiment with a different population. Duplicate it instead.
• Archiving a running test. Archiving stops serving it immediately, but the test never gets a proper ending. End it first so its results are a final snapshot, then archive.

Next Steps

• Plan launch and end dates ahead of time: Scheduling Experiments.
• Control how visitors are split between variants: Traffic Allocation.
• Run several tests at the same time without stepping on each other: Running Multiple Experiments.
• Learn the rules for ending a test confidently: Declaring a Winner.