A/B Testing Landing Pages: A Practical Guide for Small Businesses

On this page

1. What A/B testing is (and what it isn't)
2. Do you have enough traffic?
3. What to test first
4. Tools for running a test
5. Setting up a fair test
6. Avoiding false conclusions
7. Keep a simple test log

You've built a landing page for your Google or Facebook ads and it's bringing in some enquiries. Could a different headline, a shorter form or a WhatsApp button bring in more? A/B testing is how you find out with evidence instead of opinions. It's also easy to do badly, especially on the modest traffic most small businesses have. Here's how to test sensibly, and when not to bother.

What A/B testing is (and what it isn't)

In an A/B test, visitors are split at random between two versions of a page: the current version (A, the "control") and a changed version (B, the "variant"). Both run at the same time, and you compare how many visitors on each version take the action you care about, such as submitting a form or tapping WhatsApp.

Running both versions at the same time is the important part. Changing your page this month and comparing it with last month is not an A/B test. Festivals, school holidays, changes to your ad budget or a competitor's offer can move your numbers far more than your new headline did.

Do you have enough traffic?

This is the question most guides skip. What matters is not visits but conversions: the number of enquiries, calls or orders each version produces. With only a handful of conversions per version, random chance can easily make a worse page look better.

A simple illustration: if version A gets 6 enquiries and version B gets 9 from similar traffic, B looks 50% better. With numbers that small, though, the difference could vanish next week. Free A/B test sample size calculators show how many visitors you need, based on your current conversion rate and the size of improvement you hope to detect. For many small business pages, the honest answer is "more than you'll get in a few months".

If your traffic is low, you still have good options:

• Test big changes, not small ones. A different offer or page structure is far more likely to produce a difference you can see than a new button colour.
• Make clearly better changes directly. Fixing a slow page, a broken form or a hidden phone number doesn't need a test.
• Use qualitative evidence. Ask customers what nearly stopped them from enquiring, and listen to how leads describe their problem on calls.

The CRO basics guide explains how to find and prioritise these improvements.

What to test first

Start with the things that shape a visitor's decision, not cosmetic details. A good test idea comes from something you've noticed, written as a hypothesis: "If we show a starting price, we'll get fewer unsuitable enquiries, because price is the first question on most calls."

Element	Example test	Why it can matter
Headline and offer	Service-focused headline vs outcome-focused headline	It's the first thing visitors read and decides whether they stay
Call to action	Form only vs form plus WhatsApp button	Some visitors, especially on mobile, would rather chat than fill in a form
Form length	Five fields vs name and phone only	Fewer fields can mean more leads, but check their quality
Proof	Testimonials beside the form vs lower down the page	Trust matters most at the moment of decision
Pricing	"Starting from" price shown vs "call for a quote"	Can change both the number and the quality of enquiries

If your page has obvious problems, fix those before testing anything; see landing page mistakes that waste ad budget.

Tools for running a test

Google Optimize, once the go-to free option, shut down in 2023. These are the main routes now; check current features and pricing before committing.

• Ad platform experiments: Google Ads and Meta Ads Manager both have built-in experiment features that split traffic between versions of a campaign or ad, and these can often be set up to compare two landing page URLs. For ad landing pages, this is usually the simplest option.
• WordPress testing plugins: plugins such as Nelio A/B Testing run tests from inside WordPress.
• Dedicated testing platforms: services such as VWO and Optimizely are powerful, but usually priced for larger businesses.

Avoid running two identical ads with different URLs and comparing them: ad platforms show whichever ad they predict will perform better more often, so the split isn't random.

Many testing tools swap content with JavaScript after the page loads, which can cause a visible flicker and slow things down, so check both versions on a mid-range phone. If test pages can be indexed, Google's guidance is to point the variant at the original with a canonical tag, use temporary redirects rather than permanent ones, and end the test once you have an answer.

Setting up a fair test

1. Pick one primary goal. Usually a form submission, call or WhatsApp click, tracked as a key event; see GA4 events explained.
2. Write down the hypothesis and what result would make you switch.
3. Decide the duration in advance. Run for full weeks, because weekday and weekend visitors behave differently, and avoid overlapping with a festival sale unless that's what you're testing.
4. Test one idea at a time. A variant can include several edits serving one idea, such as "make the offer clearer", but not five unrelated ideas.
5. Check both versions properly. Test forms, buttons, tracking and the mobile layout on each version before sending traffic.
6. Leave everything else alone. Changing ad copy, targeting or budgets mid-test muddies the result.

Avoiding false conclusions

Most bad decisions from A/B tests come from reading the results too eagerly.

• Stopping early: results swing a lot in the first few days. Checking daily and stopping the moment B pulls ahead is the most common mistake.
• Ignoring lead quality: a shorter form might bring more enquiries but fewer genuine customers. Track which leads become paying clients, not just form counts.
• Slicing the data until something wins: if B lost overall, discovering that it "won on Android on Tuesdays" is usually noise.
• Misreading "significance": when a tool says a result is statistically significant, it means the difference is unlikely to be pure chance, not that the improvement is large or permanent.

An inconclusive result is still useful. It tells you that change didn't matter much, so your next test should be bolder.

Keep a simple test log

Record each test in a spreadsheet: dates, page, hypothesis, what changed, visitors and conversions per version, the decision and what you learned. After a few tests, it becomes a record of what your customers respond to, useful for your ads and sales calls too.

Need a landing page that's built to be tested and improved? See landing page design.