Split visitors between versions of a page and see which one brings more sign-ups or forms. The AI can also write an improved version for the next round.
Not sure whether "Book your first class free" works better than "Reformer classes from 60 shekels"? An A/B test shows both versions to real visitors and counts which one leads to more conversions, such as a filled form. You need the app published and two or more versions of the page. Ask the AI in the chat to create the second version first.
Ask the AI for a second version: "Create a copy of the home page called HomeB with a different headline: "Your first reformer class is on us". Keep everything else the same."
1. Open More, then A/B Testing: The screen is called Split Testing. Click + New Experiment.
2. Name it and choose the base page: Under Experiment Name write something like "Home headline". The base page is the address visitors open; they are sent to one of the variants.
3. Add the variants and the split: Choose a page for each variant and the share of traffic it gets. The shares must add up to exactly 100%. Use + Add Variant for a third version.
4. Choose the conversion goal: Under Conversion Goal pick what counts as success: any data saved (any form or record created), a specific data type such as Leads, or no goal (views only).
5. Click Save Experiment: The test starts counting. Each variant shows its visitors, conversions and CVR (conversion rate).
When one variant is clearly ahead, click Set as the Winner on it. If the experiment stays active, all new traffic goes to that page until you stop the experiment or create a new one. You can also restart a run: existing data is kept, and a new run counts only new leads from now.
Turn on the option that lets the AI generate improved variants based on performance. After every set number of views it checks the results, picks a winner and writes a new variant to replace the loser. Max Rounds limits how many times it does this (0 means unlimited). Each new variant is AI work and uses credits.
Wait for enough visitors: Ten visitors are not a result. Let the test run until each variant has at least a few dozen visitors and several conversions before you decide.
Q: Where do I see test results over time?
A: On the A/B Testing screen itself, per variant. Overall traffic is in Analytics.
Q: Can I mark my own test form as not real?
A: Yes. Conversion records can be marked as test-only so they do not count.
Q: Why is A/B Testing locked?
A: It is locked during the very first build of the app. Wait for it to finish.
Q: Do visitors notice the test?
A: No. Each visitor simply sees one version of the page.
Related: See who visits your app (Analytics), Publish your app, The database: tables and records, Track your credits and earn free ones