Conversion Optimization

How to Run a Landing-Page A/B Test You Can Trust

Updated: 9/1/26

Short answer: Start with one important hypothesis, define the primary metric and stopping rule before launch, split comparable traffic, validate tracking, and let the test run through normal business cycles. Judge the result by meaningful business outcomes and uncertainty, not by whichever variation happens to lead early.

Write the hypothesis first

A useful hypothesis connects a change to a customer reason and an expected outcome: “Adding specific delivery timing near the purchase button will increase completed purchases because uncertainty about arrival is a common objection.”

Define the test before seeing results

Decision	Define before launch
Primary metric	Qualified lead, purchase, revenue, or another business outcome
Guardrails	Lead quality, refunds, order value, page speed, or errors
Audience	Eligible traffic and exclusions
Duration	Minimum sample and complete business cycles
Decision rule	What evidence supports launch, rejection, or more testing

Avoid common invalid tests

Stopping when a preferred version takes an early lead
Changing traffic sources or budgets during the test
Testing several unrelated changes and claiming one caused the result
Using clicks when the business needs qualified leads or purchases
Ignoring mobile performance or form errors in one variant
Running overlapping experiments that affect the same users

Important GA4 note

GA4 can receive and analyze experiment data, but A/B tests require a third-party or in-house experimentation system. GA4 is not the tool that assigns visitors to variants.

Document what you learned

Save the hypothesis, screenshots, dates, traffic rules, sample sizes, outcomes, guardrails, and decision. An inconclusive test can still teach you that the proposed change was too weak, the audience was too small, or measurement needs improvement.

Frequently asked questions

How long should an A/B test run?

Long enough to reach the planned sample and cover normal weekly variation. Avoid arbitrary one-week rules when volume is low.

Should I test one change at a time?

Yes when you need causal clarity. Test a complete redesign when the strategic question is whether the whole experience performs better.

Can GA4 run an A/B test?

No. Google directs users to integrate a third-party experiment tool or build an in-house framework, then use GA4 for interpretation.

Sources

Published by Marketing That Clicks
Last reviewed September 2026.