
You launched a landing page and it converts at 2 percent. Someone says, "Try a green button, I read it increases sales." You do. Nothing happens.
That is the story of many A/B tests. People test tiny, random changes because they are easy, and learn nothing. Let's do it better.
What A/B testing means
You show two versions of a page to different visitors at the same time. Version A is the current page. Version B has one change. After enough visitors, you compare which got more sign-ups or sales.
The key phrases are "at the same time" and "one change". Run them one after another and you cannot tell if traffic or season changed the result.
First, check you have enough traffic
A/B tests need a decent number of visitors and conversions to be reliable. If your page gets 100 visitors a month and 3 leads, a test will take forever, and the results will be noise.
As a rough guide, you want at least a few hundred visitors per version and a few dozen conversions before drawing conclusions. If you cannot reach that, make bold changes based on common sense and customer feedback instead of testing small tweaks.
What to test, in order of impact
1. The offer. What are you actually giving? A free consultation vs a free audit vs a discount can change results more than any design tweak.
2. The headline. It is the first thing people read. Test benefit-led vs problem-led, specific vs general. Our post on website copy that gets enquiries has examples.
3. The main image or video. A real photo vs an illustration, or a short video vs none.
4. The call to action. The words on the button ("Get my free quote" vs "Submit"), and where it sits.
5. The form. Fewer fields vs more. Asking for phone number or not.
6. Social proof. Reviews, logos, numbers, and where you place them.
7. Page length and layout. A short page vs a long one with more explanation.
Buttons colours and tiny font changes sit at the bottom of this list.
How to run a clean test
- Pick one goal. For example, form submissions.
- Write a hypothesis. "If we show the price range up front, fewer unqualified people will submit, but the quality will go up."
- Change one thing. Keep everything else identical.
- Split traffic evenly. Tools like Google Optimize have closed, but alternatives such as VWO, Convert, and built-in tests in some landing page builders exist.
- Run it long enough. At least one or two full weeks to cover weekday and weekend patterns.
- Look at the result honestly. Wait for enough data before declaring a winner.
Do not stop at conversion rate
A higher conversion rate is not always better. If the form with fewer fields brings 40 percent more leads but half are junk, you may be worse off. Look at lead quality and, ideally, sales. See how to qualify leads.
Use qualitative data too
Numbers tell you what happened. People tell you why.
- Watch session recordings with a tool like Microsoft Clarity, which is free.
- Look at heatmaps to see where people click and stop scrolling.
- Ask customers how they found you and what almost stopped them.
- Check chat and call logs for repeated questions your page does not answer.
These often give better test ideas than guessing.
Common mistakes
- Testing without enough traffic
- Stopping the test the moment one version looks ahead
- Changing several elements at once
- Testing during unusual periods, like a big festival sale
- Not recording what you learned
Keep a record
Write down every test: what, why, dates, results and what you will do next. Over a year, you build a clear picture of what your audience likes.
If you want fresh page designs built for testing, our landing page team builds fast pages with tracking and multiple variants.




