How to A/B Test Your Landing Page (Without Going Broke)

A/B testing sounds expensive. Optimizely wants $50k/year. VWO wants $30k. You just want to know if changing your headline will work.

Here's how to A/B test without selling a kidney.

What Actually Matters

Most A/B tests are bullshit. Testing button colors is a waste of time when your value proposition sucks.

Test these (in order):

1. Value proposition - Your headline and subheadline
2. Call to action - What you're asking people to do
3. Social proof - Which testimonials convert best
4. Page length - Long-form vs short-form
5. Pricing presentation - How you show your price

Button color? Font choice? Emoji vs no emoji? Test these last (or never).

Free A/B Testing Tools

You don't need enterprise software. Use what you have.

Google Optimize (Free)

• Integrates with Analytics
• Visual editor (no code needed)
• Good for basic tests

Vercel/Netlify Edge Functions (Free tier)

• Split traffic at the edge
• Zero performance impact
• Requires coding

Plausible/Simple Analytics ($9-19/mo)

• Event tracking
• Manual A/B testing
• Privacy-focused

Just Use URL Parameters (Free)

• Version A: yoursite.com
• Version B: yoursite.com?v=b
• Track conversions manually
• Ghetto but works

Start with the free option. Upgrade when you're making money.

How to Actually Run a Test

Step 1: Pick ONE thing to test

Don't change the headline AND the CTA AND the image. You won't know what worked.

Change one thing. Test it. Learn from it.

Step 2: Define success

What are you measuring?

• Email signups?
• Demo requests?
• Purchases?

Pick one metric. Ignore everything else during the test.

Step 3: Calculate sample size

You need enough traffic for statistical significance.

Quick math:

• 1% conversion rate → need ~40,000 visitors
• 5% conversion rate → need ~8,000 visitors
• 10% conversion rate → need ~4,000 visitors

Don't have that traffic? Test bigger changes. Small tweaks need big sample sizes.

Step 4: Run the test for at least 2 weeks

Traffic varies by day of week. Run Monday through Sunday, twice.

Ending a test on Friday gives you incomplete data.

Step 5: Implement the winner (or don't)

Winner has to be statistically significant (95% confidence minimum).

If results are close? Keep the original. Changing stuff for a 2% lift isn't worth the risk.

Tests Worth Running

Test 1: Benefit-driven vs Feature-driven Headline

• Version A: "Project Management Software with Kanban Boards"
• Version B: "Ship Projects 2x Faster"

The benefit version usually wins. But test it.

Test 2: Social Proof Placement

• Version A: Testimonials at bottom
• Version B: Testimonials right after headline

Putting proof near the CTA often boosts conversions 15-30%.

Test 3: CTA Copy

• Version A: "Get Started"
• Version B: "Get My Free Roast"

Specific CTAs outperform generic ones. But test your audience.

Test 4: Long vs Short Page

• Version A: 500-word page
• Version B: 2000-word page

B2B and expensive products often need long-form. E-commerce might not. Test it.

Common A/B Testing Mistakes

Mistake 1: Stopping tests too early

You got 50 conversions and version B is winning by 10%. Stop the test?

No. Run it for 2 full weeks minimum. Early wins often disappear.

Mistake 2: Testing too many things

5 tests running simultaneously. You're splitting traffic so much that none reach significance.

Run 1-2 tests at a time. Get clear results.

Mistake 3: Not having a hypothesis

"Let's try a blue button" isn't a hypothesis.

"Blue buttons will convert better because they contrast with our green brand" is a hypothesis.

Know WHY you're testing something.

Mistake 4: Ignoring mobile

70% of your traffic is mobile. You're testing the desktop version.

Test mobile separately or use mobile traffic only.

The Ghetto A/B Test Method

No tools. No budget. Still want to test?

Week 1: Run version A. Track conversions. Week 2: Run version B. Track conversions.

Compare results. Account for traffic differences.

It's not perfect. But it's better than guessing.

When to Stop Testing

You've run 10 tests. 8 showed no significant difference. 2 made things worse.

Maybe your page is already good enough. Maybe testing isn't your problem.

Focus on traffic instead. A good landing page with no traffic = $0.

The Reality

Most landing pages don't need A/B testing. They need basic fixes.

If your page:

• Has no clear value proposition
• Loads in 8 seconds
• Looks like shit on mobile
• Has zero social proof

Fix those first. Then test.

Want to know what's broken before you start testing? Get roasted by PageRekt. Our 6 personas will tell you exactly what to fix first.

Try it free, or get the full Deep Dive roast for $26.99 (one-time) or $29.99/month unlimited.