Landing Page A/B Testing | Improve by Measuring

Landing page A/B testing splits the same traffic between two page versions and measures which converts better. Built correctly it ends the guesswork argument; built incorrectly it lends legitimacy to a wrong decision.

Fix first, then test

Before testing, obvious problems on the page have to be fixed. Running a test on a page that loads slowly, breaks on mobile or fails to deliver the ad’s promise means choosing between two bad options.

Once the basic fixes from landing page optimisation are done, testing becomes meaningful. At that point “which is better” is a real question.

What is worth testing

Small changes do not produce measurable differences. Changing a button colour and hunting for an effect wastes time on most sites.

The tests that matter are structural: a different main headline and promise, form length and position, whether price is shown, the type and placement of social proof, page length. These genuinely affect the decision.

When traffic is not enough

A/B testing needs enough data to say the result was not chance. On low-traffic sites that takes weeks or months, and market conditions change in the meantime.

In that case sequential comparison is more practical: make the change and compare against the previous period. It is less certain but the only workable route at low traffic. Alternatively, run the test on your single highest-traffic page and apply what you learn to the others.

What breaks a test

The most common mistake is ending it early. The version ahead in the first days can fall behind as data accumulates.

Second, making other changes during the test: raising budget, launching new campaigns, changing the bidding strategy. These alter the traffic mix and invalidate the test. Third, showing the two versions in different periods — seasonality distorts the result.

What the measure should be

The measure should be conversion rate, not time on page or bounce rate. Those intermediate metrics are hard to interpret: a long visit can mean interest or confusion.

In e-commerce the measure should be conversion rate together with basket value; a change that raises conversion rate while lowering average order value does not produce profit.

How do we start? First I identify structural alternatives worth testing on your page and calculate whether your traffic supports a test. If it does not, I propose the alternative method. You can reach me from the contact page.

Frequently Asked Questions

Do I need a testing tool?

You need something to split traffic in a controlled way. A simple comparison can also be built inside Google Ads by pointing different ads at different pages.

Do two similar pages cause SEO problems?

Test pages should not be left open to search engines; that is managed with canonical tags and indexing settings. For short tests this rarely causes issues.

When should I make the winner permanent?

Once the result reaches sufficient data and the difference is measurable. After that, starting a new test is how you keep improving the same page.