Experiments

Experiments let you run up to five versions of a form against each other and measure which one actually converts better — so you change your forms based on evidence, not opinion. Formtoro splits your storefront traffic between the versions, tracks conversion and revenue for each, and tells you when the result is statistically trustworthy.

How a test worksCopy Anchor Link

Every experiment has a control and one to four challengers — up to five arms in total:

• Control — the form you already have. Its identity stays stable: embeds, integrations, and analytics keep pointing at it no matter how the test ends.
• Variants — the challengers. Each one is another form you pick — published, or a draft you haven't published yet. A test with one challenger calls it Variant; with more, they're lettered Variant B, Variant C, and so on.

Experiments compare forms you've already built. Create each version you want to test as its own form first — duplicate the control from the Forms list if you want a starting point — then pick them when you set up the test. A variant can stay a draft: starting the test publishes it, so it never shows on its own beforehand.

While the test runs, each visitor is assigned to one arm and always sees the same arm on every visit — assignment is deterministic per visitor, so nobody flip-flops between versions mid-test. You choose the traffic split when you create the test (an even split by default). Every arm is always served to the control form's audience, so the versions are compared against the same visitors.

A form can only be in one active experiment at a time, in any role.

How many arms? Each extra arm splits your traffic further, so every arm gets fewer visitors and the test takes longer to reach a verdict. Start with one challenger unless you have plenty of traffic and genuinely different ideas to compare.

Run your first experimentCopy Anchor Link

Open Experiments in the navigation and click New experiment. One modal walks you through it:

1. Pick a form to test. The picker lists your published forms, most recently edited first — a form needs at least one saved version to qualify. Only the first few are listed; search by name to find the rest. Forms already in an active experiment appear disabled, labeled with the experiment they belong to.

2. Add challengers. The test starts with one; click Add challenger for up to four. For each one, pick another form — published, or a draft (marked "Draft — published when the test starts"). If its audience differs from the control's, you'll see a heads-up: starting the test aligns its audience to the control's. If the form is on, creating the draft turns it off until the test starts, so it never shows as a separate popup alongside your control. You need a published form to test and at least one more form to challenge it.

3. Configure the test. Give it a name (pre-filled as "{form} — A/B Test") and set the traffic split: a percentage for each arm, from 1% to 99%, adding up to 100%. Drag each arm's slider or type its percentage; the other arms adjust automatically, in proportion, to keep the total at 100%. Split evenly resets it to an even split (adding or removing a challenger does too).

Click Create draft. This creates the experiment but does not start it — no traffic is split yet.

You land on the experiment's detail page. From here:

• Open any arm's form (each arm card links straight into the builder) to fine-tune it before launch. Test one meaningful change at a time — a different headline, offer, image, or flow — or you won't know what caused the difference.
• When you're ready, click Start test. Starting turns on every form in the test, publishes the latest saved version of each, and begins splitting traffic. You don't need to turn the forms on yourself: while the test is a draft its variants stay off, and turning one on from the Forms list or the builder is refused with "This form is a variant in the draft test … Start the test to turn it on." If a form was archived or is in test mode, a banner lists what to fix before you can start.

While the test runsCopy Anchor Link

• The Forms list shows who's in a test. Forms in an experiment carry an A/B Test pill plus a pill naming their arm — Control, Variant, or Variant B, Variant C… — click them to jump to the experiment.
• Audiences stay in sync. Every variant's audience is locked in the builder while the test is active; audience changes you make on the control mirror to every variant automatically. This keeps the comparison fair.
• Leave the forms alone. You can still open any of the forms in the builder, but changing a form mid-test muddies what the results mean. Make your edits before starting, or end the test and run a new one.
• Name and split are locked after start. You can edit the test's name and traffic split only while it's a draft — changing the split mid-flight would change the meaning of data you've already collected.
• Formtoro watches the test's health. If someone turns off one of the forms while the test runs, the experiment page shows a warning — an arm is getting no traffic and the results are being skewed — with a one-click fix to turn it back on.

You can Pause a running test at any time (traffic stops splitting; visitors see your forms as normal) and Resume it later.

Reading the resultsCopy Anchor Link

The experiment page shows every arm side by side, control first. By default it covers the whole test — from the moment you started it through to now, or through to the day it finished — so the significance verdict weighs every visitor the test has seen rather than an arbitrary recent slice. The date picker narrows that window if you want to look at a particular stretch; a range wider than the test is trimmed to the days the test actually ran, and the header always names the window you are looking at.

Under the page title you'll see when the test started and, once it's over, when it ended — the same dates the Experiments list shows.

Within that window, each arm card leads with two rates side by side:

• Opt-in rate is the headline number for each arm — submissions as a share of unique visitors who saw the form. This is the rate the test is judged on: the arm with the highest opt-in rate wears a Leading badge once every arm has traffic, and the significance banner is computed on it.
• Sub → conversion answers the next question — of the people this arm signed up, how many went on to buy? It's the share of signups (submissions that gave you an email or phone) with at least one attributed order — the same subscriber-grain definition as the Sub → conversion tile on the Analytics overview. A repeat buyer counts once. An arm can win on opt-in rate and lose here, which is exactly the trade-off you want to see before promoting a winner.
• Traffic — viewers, submissions, completions, and dismissals. Each counts people, not events: a shopper who reloads the form, or returns on another day or device, is one viewer. Submissions here means everyone who sent a step, which is not the same as the Forms tab's Subs — that counts only the people who gave you contact details. Where the two differ, read the arm cards as a measure of engagement.
• Revenue — revenue per visitor, attributed orders, and net revenue (after refunds), so you can see whether the "winner" actually makes more money.
• Pooled totals sum every arm — the experiment's overall footprint.

Each variant's card also says how it compares with the control: its opt-in lift (for example "+20.0% vs control") and a verdict — Beats control, Trails control, Not significant yet, or Not enough data yet.

Above the cards, a banner tells you how much to trust what you're seeing:

• "Not enough data yet to call a winner" — the sample is still too small for the math to mean anything. A 100% opt-in rate on three visitors is noise, not signal. Keep the test running.
• "Not yet statistically significant" — there's data, but the gap between the arms could still be luck. Let it run longer.
• "Significant at 95% confidence" — the difference is real with 95% confidence, and the banner names which arm is winning. With several variants, it says which variant beats the control (or that the control beats every variant): each variant is compared with the control.

More variants, higher bar. Every extra comparison is another chance for luck to look like a win. When a test has more than one variant, Formtoro corrects each comparison for the number of variants (the Holm method), so the p-values it shows are already adjusted and adding arms doesn't raise the odds of a false win. The card verdicts use the same adjusted numbers as the banner, and no card claims a win while any variant is still short on data.

Be patient. Most tests need hundreds of visitors per arm — often a week or two of traffic — before a trustworthy verdict emerges. Ending a test the moment one arm pulls ahead is the most common way to ship the wrong form. Significance is a nudge, not a gate: you can always declare a winner manually, but the banner tells you when the data has your back.

Ending a testCopy Anchor Link

From the experiment page's top bar:

• Declare winner — the decisive move. A modal shows every arm with its opt-in rate; the leading arm is pre-selected, but you pick. On confirm, traffic splitting stops, the winning version keeps serving your storefront, and every variant form is turned off at the same moment — they stay in your forms list, so you can reuse them later or archive them to free up form slots. If a variant won, its design is carried over to your original form — so every embed, integration, and report that pointed at your original form keeps working, now serving the winning version. If the data is still thin, the modal warns you before you commit. Declaring a winner ends the test permanently.
• End test — stop collecting data without picking a winner yet. Your control form keeps serving your storefront and every variant form is turned off at the same moment, so they don't show up as separate popups. The results stay available on the page, and you can still declare a winner later.
• Cancel test — abandon the experiment. On a running or paused test, your control form keeps serving and every variant form is turned off. Cancelling a draft leaves its variants off. Either way, no form is changed or deleted — turn a variant on from the Forms list if you want it live on its own.

An experiment's status moves through Draft → Running (with optional Paused detours) and ends as Completed, Promoted, or Discarded — always visible as a badge in the Experiments list. Drafts and running or paused tests sit under Active tests; finished ones move beneath them to Previous tests, each with the dates it started and ended, and stay there so you can revisit any test's results later.

Next stepsCopy Anchor Link

• Analytics & submissions — the metrics behind your experiments, for every form.
• Building forms — craft the variants you want to test.
• Audience filtering — who is eligible to see your forms.
• Luigi, your AI assistant — ask "How are my A/B experiments doing?" for a plain-language readout.