How to Use Optibase to A/B Test a Webflow Landing Page

The short answer: if your landing page sees fewer than 10,000 new visitors a month, Optibase's Free plan covers this whole test, one test at a time (pricing page, checked on 26 October 2026), so start there: install the Webflow app and the script, create a Submit conversion on the form, create an A/B test, connect the original element as Main and a duplicate as the challenger, attach the conversion, preview without touching the data, publish, let an assistant read progress through the MCP server, and stop at a P2BB of 95%.

Disclosure: I may earn a commission through some links on this site. It does not change what this page says.

The job here is narrow on purpose: one landing page, one form, one change. Optibase's Webflow app lets a marketer do all of it inside the Designer, and the docs are clear about the few places where a setup quietly goes wrong. For plans and rivals, see the Optibase glossary entry.

Before you start:

• A Webflow project you can publish, with access to Site Settings and Custom Code.
• One change worth testing. A headline, an offer line or a shorter form. Optibase's conversions overview advises focusing tests on macro goals such as signups, not hovers.
• A realistic clock. A docs FAQ gives 300 users per variant as a minimum. For example, a page with 4,000 new visitors a month gets 4,000 / 30 = 133 a day, so two variants of 300 take 600 / 133 = 4.5 days; the P2BB article warns the numbers swing early, so plan for weeks, not days.

1. Install the app and the script

Add the Optibase Webflow app, then copy your workspace script. In Webflow, click the logo, open Site Settings, choose Custom Code and paste the script into Head code; save, publish, then click Check in the Optibase dashboard, as the script guide describes. Place it at the very top of the head: the flicker FAQ traces flicker to a script that loads too late.

2. Create the conversion and connect it to the form

In the app's Conversions tab, click New Conversion, name it, pick Submit and click Create (create a conversion). Then select the Form element on the page and click Connect Submit Event (connect a conversion). Track the submission or the button click, not both; the conversions overview flags that as double counting.

3. Create the test

Open the Tests tab, click New Test, enter a name, choose A/B Test (Variant Group) and click Create, per create a test.

4. Connect the Main variant and the challenger

Select the element you want to test and click Connect; the first variant becomes the Main. Duplicate the exact element, select the duplicate, make your change and click Connect again, as create an A/B variant explains. Leave the Main visible when you publish: whichever variant is visible at that moment is the one shown by default while a test is restricted or in draft.

5. Attach the conversion to the test

Open the test, go to its Conversions tab and pick the conversion from the dropdown; it links at once. You can attach several, and one serves as the primary metric in the results, per connect conversions to a test.

6. Narrow the audience if the change demands it

In the test's Restrictions section you can limit it by country, device type, operating system and browser (restrictions). A challenger designed for desktop belongs on desktop only; with no restriction set, the test runs for everyone.

7. Preview without touching the data

Click the Preview/Force icon next to a variant, enter your base URL and copy the link; it adds an optibaseVariants parameter that forces that variant. Previews record no page views or conversions, per preview variants, so send them to colleagues freely. Add ?optibaseEnableDebugMode to a URL to see in the console whether the conversion listener fires.

8. Publish

Set the test to active and publish the Webflow site. From here, every new visitor who loads the page counts as a tested user and keeps the same variant on later visits, per what is a tested user.

9. Let an assistant watch it, read-only

Optibase runs an MCP server. In Claude Desktop, open the MCP servers settings, add https://my.optibase.io/api/mcp, then log in and pick the workspace when the browser opens. Ask: "Which of my tests have enough data to declare a winner?" and "Fetch the page for my landing page test and tell me what each variant shows."

The human check: the server is read-only and cannot create, edit or stop tests, per the MCP announcement, so the assistant reports and a person decides. Before anything ships, that person opens the test's results and confirms the P2BB and the users per variant.

10. Stop, then make the winner the page

Optibase's docs give two bars: a stop FAQ accepts above 90% with 300 users per variant, the P2BB article says wait for 95%. Take 95%. On Business and above, Auto Stop can pause the test at a P2BB you set and show the leading variant to everyone. Then make the winner the Main in Webflow and publish; search engines see the change only from that point, per the SEO FAQ.

What the docs warn about

• Two tests on one element. Several tests can share a page, but overlapping elements give unreliable results.
• Your own visits in the data. The IP blacklist keeps a team's actions out of the numbers, but the pricing table offers it from Business; on Free and Basic, use preview links instead of the live page.
• Split tests you want to pause. A split variant cannot be paused or disconnected without deleting it, per create a split variant.

The Free plan's single active test is the limit that does the selling: the second idea on your list waits until you pay $79 a month or end the first test. Run them in order of what each one could move.

FAQ

Do preview links change my test results?

No. Preview URLs force a variant for your browser without recording page views or conversions, per the preview variants guide.

Can I run two tests on the same page?

Yes, as long as they do not change the same elements; overlapping tests make both results unreliable.

What happens when Auto Stop fires?

The test status changes to Paused, traffic is no longer split, and the variant performing better on current results is shown to all users. The history stays, and you can restart the test, per the Auto Stop guide.