A goal you can check
Choose a visitor action, then review the measurement before starting the test.
Use case · Optimization
Choose a conversion, check how it is measured and let Autopilot test a focused change. Read the counts and the outcome before calling it an improvement.
Eligible plan, published site, connected domain and a measured goal.
What Frontpage does
A button label is easy to change. Knowing whether it helped takes a working measurement, a fair comparison and enough visitors. Frontpage keeps those parts with the site: the page, the challenger and the experiment report.
Autopilot runs one test at a time on an eligible published Frontpage site with a connected custom domain. You describe the conversion you want, check how it will be measured and choose whether to enable automatic testing and publication.
Choose a visitor action, then review the measurement before starting the test.
Test a headline, CTA label or another allowed change while keeping the offer and prices intact.
Compare visitor and conversion counts. A leading challenger is not automatically a proven winner.
Autopilot can keep the control or finish without a winner. It does not have to call every test an improvement.
Side by side
The problem
An A/B test needs more than two versions of a headline. The conversion has to mean the same thing for both versions, and the result has to reflect enough activity. A small difference early in the test can disappear as more people visit.
“More engagement” is hard to judge. Choose the action that matters, such as a particular signup or CTA click.
Keep the destination and tracking intact when testing a button label. Otherwise you may be comparing different actions.
Read visitors and conversions alongside the rate. One extra conversion can look large in a small sample.
A live test may need more data or finish inconclusive. Starting tests regularly does not guarantee a useful result on the same schedule.
How it works
Use an eligible plan, publish the site and connect the custom domain. In Autopilot Setup, describe the conversion and review its measurement.
Autopilot runs one test at a time. Read the exact change and compare visitor and conversion counts in Activity.
A supported winner can be kept and published automatically. Other tests retain the control or finish inconclusive. Review history before deciding what to do next.
Before you enable it
Keep the setup focused on one real customer action.
Scale, Enterprise or Internal, with a published build and connected custom domain.
Describe the conversion and verify the page or action Autopilot will count.
Keep the underlying offer, destination and tracking consistent while testing the allowed change.
Tell Autopilot the constraints that matter to the site. Review the allowed change types and keep public facts accurate.
Enable automatic testing only when automatic publication fits how you manage the site.
Use Activity and history to see the change, counts, status and available owner actions.
The testing process
Keep the action consistent, the change small and the publishing behavior clear.
Describe the visitor action in Setup and review the measurement Autopilot identifies before enabling it.
Use one allowed change, such as “Request a quote” versus “Get a quote.” Do not change pricing, legal wording or the meaning of the offer.
Check the control and challenger, visitor counts, conversions and confidence. Use the result’s status rather than the largest displayed percentage.
Autopilot can publish the working copy when it keeps a winning change. Applying a challenger or rolling back also publishes; review other draft work before using those actions.
Score the current stack
A website audit can help find usability problems before a test. Fix a broken customer path before trying to optimize the wording around it.
Two working processes
Use Growth Advisor when you want suggestions to review. Use Autopilot when you want the automatic testing workflow and understand its publishing behavior. Neither establishes a fixed traffic or conversion improvement.
How Growth Advisor works →
What the result means
An experiment measures the conversion you chose. It does not establish that the page will rank higher or be indexed. Keep useful content, titles and links intact while measuring the customer action.
See it in action
An illustrative setup conversation, not a measured customer experiment.
Review the actual measurement in Autopilot Setup. The example does not promise a winner or a conversion increase.
The gaps we fill
The purpose is to learn which version to keep, not to produce a success claim every time.
Winner
A challenger needs enough visitors and evidence before the automatic decision can keep it. A displayed lead by itself is not that decision.
Control
Keeping the control is a useful outcome. It avoids turning every experiment into a permanent site change.
Inconclusive
Low traffic or a small difference can leave a test without a clear result. Say that plainly rather than turning a tentative rate into a proven improvement.
Questions
This workflow tests a published Frontpage site on its connected custom domain. It is not a testing script for arbitrary external sites.
Autopilot is available on Scale, Enterprise and Internal plans. Growth Advisor is separate. Check current plans for current availability and limits.
No. Start cadence limits when another test can begin. A running test still needs sufficient evidence and may finish without a winner.
Its current change policy excludes pricing values, legal text and changes to the meaning of the offer. Use allowed copy or layout changes that preserve the underlying facts.
Pause stops new experiments. A running experiment can still finish. Check the current experiment before treating pause as a stop to all activity.
Yes. Its winning-change workflow publishes the working copy. Applying a challenger or rolling back also publishes. Review other unpublished work before enabling or using those actions.
Build a site around a clear customer action. When the site is ready, decide whether automatic testing and publication fit your workflow.
Autopilot can publish automatically. Review other draft work before enabling it.