You have two versions of a landing page. One has a blue button. The other has an orange button. Which one converts better?
The only honest answer is: you don't know until you test it.
A/B testing (also called split testing) is how you answer that question with data instead of opinion. And while most people think it requires a complex testing platform and a developer, you can run meaningful split tests using nothing more than tracked short links and a bit of patience.
What A/B testing actually is
An A/B test means you show two different versions of something to two different groups of people, then measure which version produces better results.
"Version A" might be your current landing page. "Version B" is the variation you think might perform better. You split your incoming traffic between the two — half goes to page A, half goes to page B — and after enough visitors have seen both, you compare the conversion rates.
The version with the higher conversion rate wins. You keep it, and use the losing version's weaknesses to inform your next test.
The simple split-test setup using links
Most A/B testing tools require you to install code on your website and handle the traffic splitting automatically. That's the sophisticated way. But there's a simpler approach that works well for early-stage testing:
Create two pages, and split your traffic by controlling which link you share where.
Here's how:
Step 1. Build your two landing page variants. They can be on your actual website (different URL paths like /landing-a and /landing-b) or on a simple landing page tool like Carrd, Webflow, or Notion.
Step 2. Create two tracked short links in gometrify — one pointing to each variant.
links.yourbrand.com/test-a→ points to landing page version Alinks.yourbrand.com/test-b→ points to landing page version B
Create your free account → to set these up.
Step 3. Distribute traffic evenly between the two links. For example:
- Send your Monday email to half your list with the
test-alink - Send the same email to the other half with the
test-blink
Or run two ad sets in Facebook Ads Manager — identical targeting, identical budget, but one sends to test-a and the other to test-b.
Step 4. After enough traffic (at minimum 100–200 visitors per variant, ideally 500+), compare the results.
What to measure
Click-to-conversion rate. Of everyone who clicked the link and landed on your page, what percentage completed your goal (purchase, signup, form fill)?
This is the core metric. You're not comparing click counts — you're comparing what happened after the click.
To track conversions, you can:
- Check your checkout or form submission data filtered by date
- Use a UTM parameter on each link so Google Analytics separates the two (
utm_campaign=test-avsutm_campaign=test-b) - Use a tool like Google Optimize, VWO, or Optimizely if you want automatic traffic splitting and built-in stats
gometrify's dashboard shows you click counts and timing for each link. Pair that with your conversion data to calculate conversion rates for each variant.
What to test (in order of impact)
Not all elements are worth testing equally. Start with the things that have the most leverage:
1. The headline. This is the most impactful thing on any landing page. A headline change can move conversion rate by 20–50%. Test this first.
2. The CTA (call-to-action) button. Text, color, and placement all matter. "Start free trial" vs. "Try it free" vs. "Get started" can produce meaningfully different results.
3. Social proof. Adding testimonials vs. not. One testimonial vs. three. A five-star badge vs. a customer photo and quote.
4. Page length. Long-form page vs. short page. This often depends on whether the audience is warm (already knows you) or cold (discovering you for the first time).
5. The offer. Free trial vs. demo request. Monthly pricing vs. annual pricing shown first. This is a bigger change — it's testing your business offer, not just the page — but it often has the largest impact.
How long to run the test
This is where most beginners make a mistake: ending the test too early.
If you've only sent 50 visitors to each variant, any difference in conversion rate could easily be random noise. Statistical significance requires enough data for the result to be reliable.
A rough rule: run until each variant has at least 100 conversions (not just visitors — conversions). If your conversion rate is around 5%, that means you need roughly 2,000 visitors per variant.
If you're not getting enough traffic for a proper A/B test, a simpler approach is sequential testing: run page A for two weeks, switch to page B for two weeks, and compare. This doesn't control for external variables (different weeks can have different buyer intent), but it gives you directional data when traffic is limited.
After the test
Once you have a winner, implement it everywhere. Then document what you learned — specifically, why you think the winner worked. Was the winning headline more specific? Did the winning CTA have less friction? Understanding the why makes your next test smarter.
Then run another test. Conversion rate optimization is a process, not a one-time event.