How to A/B test a SaaS landing page with real visitor votes
Compare SaaS homepage variants with LandingCrush swipe votes: how to submit each version, how crush and pass work, and how that differs from the AI Panel.
To A/B test a SaaS landing page with real visitor votes on LandingCrush, submit each variant as its own homepage, wait until both are approved, and let registered participants swipe. Right is a crush. Left is a pass. Compare crush rate, vote count, and comments. Use a Creator Duel when you want a 48-hour head-to-head and a link you can share. The votes are cast by people using the LandingCrush feed. They are real pass-or-crush choices on the page you submitted. They are not a random split of traffic already on your domain, and LandingCrush does not record whether those voters started a trial on your site.
Use this when you need a preference read on a hero, a headline, or a first screen. Hold the decision until the counts are large enough to read. The product’s own accuracy process waits for 50 or more real community votes before an AI prediction is compared with the community outcome. A 15-vote crush rate can tell you that something looks confusing. It should not be the only reason you rewrite the live site.
Prepare two variants that differ on purpose
A useful comparison changes one idea. Typical pairs are a headline that names the job versus a headline that names a metric, a product screenshot versus an illustration, a button that says what happens next versus a button that says “Learn more,” or pricing visible in the first screen versus pricing pushed down. Keep the company, the offer, and the audience the same so a crush refers to the page change.
Capture both variants the way voters will see them. Paste a live URL if you want desktop and mobile screenshots captured for you, or upload images when the alternative is not on a public URL yet. Give both submissions the same industry and a description that says they are variants of one homepage, so the dashboard stays readable. Write the single difference down before anyone votes. “Variant B moves the customer logos under the headline and shortens the subhead to one line” is a hypothesis you can check against comments. “Variant B is the new one” is not.
Submit each variant so people can swipe
Open /submit for each variant. Add the company name, the URL or the upload, the industry, and a short description. The page enters a moderation queue and stays pending until it is approved or rejected. The swipe feed is for approved homepages. One submission produces one crush rate. Putting two screenshots in a single description does not create a winner. The vote attaches to the homepage record, not to a sentence inside the description.
The free plan includes one active homepage submission. Creator Pro, at $19 per month or $190 per year, includes unlimited homepage submissions. That is the plan that lets two variants sit in the feed at the same time. Version history can attach a later iteration to an earlier submission with a parent id. The creator dashboard lists each submission you own with status, total votes, and crush rate, and the analytics view adds weekly votes and velocity.
After approval, participants vote at /swipe. Ask people who already know the brand to turn Blind Mode on when you want the layout judged without the logo. Voters can leave comments on the detail page. You do not assign the audience. Persona role and company size are self-reported and optional, so a role breakdown is a description of who showed up, not a panel you recruited.
Run a Creator Duel for a head-to-head
A duel is the timed comparison already in the product. From /duel, challenge another homepage to a 48-hour vote. You need at least one approved homepage to start. The page returns a share link both sides can send. Voters arrive at that matchup instead of only browsing the general feed. The product describes a Battle Tested badge for the winner. The badge records the duel. It does not record your conversion rate.
Choose a duel when you want one clock and one link. Choose two independent submissions when you want each page to accumulate a crush rate in the ordinary feed, including weekly votes and velocity. If the second design exists only as an image, upload it as its own homepage once your plan allows another submission, wait for approval, then start the duel. Do not expect the live URL’s crush rate to score an image that was never submitted.
Keep the AI Panel in a separate column
The $29 test covers one variant. The $49 test covers 2 to 4 variants head-to-head. Both are the AI Panel: fictional personas, a written verdict, and an AI Panel Preview label. The methodology describes 50 niche-matched personas and three deliberation rounds. That preview is useful when you want a structured read before a page is public. It is not the swipe test. If you buy one, put the panel winner in one column and the crush rate in another. LandingCrush states that community votes are the ground truth when they conflict with the panel.
Approval also queues a design critique with scores from 0 to 10 for hero clarity, call-to-action effectiveness, visual hierarchy, and mobile responsiveness, plus suggestions. Creator Pro uses the priority queue. Leave those scores out of the crush-rate average. They are a different instrument. Averaging a 7.4 visual score with a 61 percent crush rate produces a number neither system defined.
Read a winner only when the counts support it
For each variant, write down total votes, crush count, pass count, crush rate, weekly votes, and velocity. Velocity tells you whether one page was simply active in the last 48 hours. Then read comments that mention the change you actually made. If one variant has 12 votes and the other has 80, there is no winner yet. If both are past the 50-vote mark the product uses when it audits panel predictions, and one crush rate is clearly higher while comments point at the same change, you have a preference you can act on. If the rates are close, the test found no preference. Keep the version that is easier to maintain, or change a larger idea and submit again.
After you pick a presentation, change the live homepage and watch the conversion event in your own analytics. LandingCrush answered which submitted page these voters crushed. Your analytics answer what visitors on your domain did next.
A sequence you can follow
- Write the one difference between variant A and variant B.
- Submit each as its own homepage. Use Creator Pro when you need more than the free plan’s one active submission.
- Wait for approval. Use Blind Mode when brand recognition would flatter a weak layout.
- Let both pages collect votes, or run a 48-hour Creator Duel if you want a shared link.
- Compare crush rate only when the vote counts are in a similar range, and read comments before you rewrite the page.
- Leave the $29 or $49 AI Panel, and the post-approval critique, in their own columns.
- Update the live homepage, then measure the action that matters on your own traffic.
Common questions
How do I A/B test a SaaS landing page with swipe votes?
Submit each variant as its own LandingCrush homepage, wait until both are approved, and let people swipe right to crush or left to pass. Compare crush rate, vote count, and comments. A Creator Duel runs a 48-hour head-to-head with a share link.
Are these votes visitors from my website?
They are real people voting on LandingCrush. The product records a crush or a pass on the submitted homepage. It does not split traffic on your domain and it does not record purchases or trial starts on your site.
Can I test two variants on the free plan?
The free plan includes one active homepage submission. Unlimited submissions, which let you keep two separate variants in the feed, are part of Creator Pro at $19 per month or $190 per year.
Does the $49 AI Panel replace the swipe test?
No. The $49 test is an AI simulation of 2 to 4 variants. The swipe is the human vote. Keep the results separate. The product treats community votes as ground truth when they disagree with the panel.
How many votes should I wait for?
LandingCrush waits for 50 or more real community votes before it compares an AI prediction with the community result. Treat a smaller sample as directional, especially when the two variants have very different vote counts.