A/B tests
A/B tests let you compare your live funnel (Original) against a Variant B with different step content — without duplicating the whole funnel.
Use them to test headlines, paywall copy, quiz questions, pricing presentation, and other step-level changes while real visitors are split between the two versions.
Before you start
- Publish the funnel first — you cannot create a test on a draft funnel.
- Decide one primary metric you care about (for example confirmed paid or email captured).
- Plan which steps you will change in Variant B.
One active test per funnel
Only one test can be running or paused on a funnel at a time. Finish or stop the current test before starting another.
Open A/B tests
- Go to Funnels.
- Open the ⋯ menu on the funnel row and choose A/B test.
You can also open A/B test from the toolbar while editing the funnel.
Create a test
- On the A/B tests list, click Create A/B test.
- You land on the test detail page in draft status.
- In Settings, set:
- Name — a label for your team.
- Primary metric — what Rate and Lift are based on (see table below).
- Variant B traffic — percentage of visitors who see Variant B (Original gets the rest).
- Click Save settings while still in draft.
| Primary metric | Counts when a visitor… |
|---|---|
| Confirmed paid | Completes a captured payment |
| Email captured | Submits email on an email capture step |
| Checkout viewed | Reaches the checkout screen |
| Success viewed | Reaches the success screen |
Edit Variant B
- In the Steps list, pick a step.
- Switch the editor to Variant B.
- Change copy, images, plans, or other fields the same way you would in the funnel editor.
- Changed steps show a changed badge in the list.
Original is read-only on the test page — it always reflects what visitors on the Original arm see from the live funnel.
To undo a Variant B change before launch, select the step and click Revert step.
At least one change required
You must change at least one Variant B step before you can Start test.
Start, pause, and stop
Start
- Click Start test.
- Review the traffic split, primary metric, and list of changed steps.
- Click Confirm start.
The test moves to running. Live visitors are assigned to Original or Variant B and stay on that arm for the test.
Pause and resume
While running, use Pause to stop splitting traffic. All visitors see Original until you Resume the test.
Stop
Click Stop when you have enough data. The test becomes completed and final results are saved.
Read results
The Results table on the A/B test page compares Original and Variant B at a glance:
| Column | Meaning |
|---|---|
| Viewed | Visitors who opened the funnel in this arm |
| Lead | Email captures |
| Checkout | Checkout screen views |
| Success | Success screen views |
| Paid | Captured payments |
| Rate | Primary metric ÷ Viewed |
| Lift | How Variant B compares to Original on the primary metric (Variant B row only) |
Use Rate and Lift to pick a direction before you Keep Original or Make B original.
Step drop-off and variant comparison
Step drop-off per variant lives in Analytics, not on the test settings page:
- Open Analytics (or click A/B running / A/B paused in Funnel Breakdown — it opens Analytics filtered to that test).
- In Analytics scope, confirm the Experiment and set Variant to Original or Variant B.
- Expand the funnel row in Funnel Breakdown to read step drop-off for that arm.
Switch Variant to compare where each arm loses visitors — alongside the Lift summary on the test page.
Active test badge
While a test is running or paused, Analytics → Funnel Breakdown shows A/B running or A/B paused. The badge opens Analytics with that experiment pre-selected so you can filter by variant immediately.
After the test ends
When status is completed, choose what happens to the live funnel:
| Action | Effect |
|---|---|
| Keep Original | Live funnel stays as it was before the test |
| Make B original | Variant B content is copied into the live funnel for all changed steps |
Pick the winner based on your primary metric and business judgment — the product does not auto-declare a winner.
While a test is running
A banner appears in the funnel editor when a test is running or paused.
| What | Behavior |
|---|---|
| Funnel structure | Frozen — you cannot add, delete, or reconnect steps |
| Changed steps | Frozen in the editor — edit them from the A/B test page |
| Other steps | Still editable — updates apply to both Original and Variant B |
| Paywall price IDs | Use Operational fix on the test page to correct a plan's payment price without unfreezing paywall content |
Variant B after launch
Once the test is no longer in draft, Variant B content is read-only on the test page. Stop the test to finish, or use Operational fix for paywall price corrections only.
Tips
- Test one meaningful change at a time when possible — easier to interpret results.
- Run the test long enough for your traffic volume; low Viewed counts make Lift unreliable.
- Use the same public link in ads — assignment happens automatically for each visitor.
- Preview from the editor does not participate in traffic splitting; use a normal browser session on the live link to verify both arms.
Related: Funnels · Funnel editor · Analytics