A/B testing compares two versions of a page or element to learn which performs better against a defined metric. It sounds straightforward, but small businesses often have too little traffic to detect modest differences reliably. If you stop a test as soon as one version looks ahead, random variation can be mistaken for a real improvement. A practical testing plan starts with a clear hypothesis and honest expectations about data limitations.
TL;DR
- Match the offer, audience and destination page.
- Check mobile usability, tracking and lead follow-up.
- Compare qualified enquiries and business outcomes, not clicks alone.
- Test changes against a baseline; do not treat a single result as proof.
Write the hypothesis before changing the page
A useful hypothesis describes the observed problem, the proposed change and the outcome it should affect. For example: “The form completion rate may be low because the first screen does not explain the service area; adding that information above the form should reduce unsuitable starts and improve qualified submissions.” This is more useful than “try a new design and see.”
Choose one primary metric
Select the measure closest to the business goal. For an enquiry page, this may be successful submissions, but qualified leads may be more meaningful if the sales process can record them. Track secondary diagnostics such as scroll depth or button clicks without treating them as equivalent to a qualified enquiry. Define the conversion event before the test begins.
Understand the traffic limit
The sample needed depends on the baseline conversion rate, the smallest improvement worth detecting and the volume of traffic. Low-converting pages and small expected improvements can require many visitors. If the page gets only a few dozen visits a week, a formal split test may take too long to be useful. In that case, prioritise obvious usability faults, collect qualitative feedback and make carefully documented changes without claiming a definitive test result.
Need help with your ad-to-enquiry journey?
The Super Shoppe’s Ad Funnel service can help identify a testable bottleneck in your campaign journey. Before running experiments, make sure the page and conversion tracking work properly and that there is enough traffic to learn from the result.
Avoid changing many factors at once
If you change the offer, headline, image, form length and audience simultaneously, you will struggle to learn which change mattered. Some redesigns require bundled changes because the parts work together, but describe the experiment honestly. Keep campaigns and traffic sources as comparable as possible, since a shift in audience can create apparent performance changes unrelated to the page variation.
Use proper testing discipline
Decide in advance how long or how much traffic the test needs, what result would justify a change and how you will handle incomplete or duplicate conversions. Do not stop the test simply because a dashboard temporarily shows a winner. Check device and audience mix, note external changes and consider whether the observed difference is large enough to matter commercially.
What to do when you cannot A/B test
Use a conversion audit: watch a person unfamiliar with the page attempt the task, test forms on mobile, check message match and review actual lead-quality data. Make high-confidence fixes such as broken forms, unclear price conditions or poor mobile usability. Then monitor the baseline and the post-change period while acknowledging that it is not a controlled experiment.
How The Super Shoppe can help
The Super Shoppe’s Ad Funnel service can help identify a testable bottleneck in your campaign journey. Before running experiments, make sure the page and conversion tracking work properly and that there is enough traffic to learn from the result.
Sources
- Google Ads Help: Evaluate landing page performance
- Google Ads Help: Optimise ads and landing pages
- Google Search Central: Helpful, reliable, people-first content
*Editorial note: Keyword choices are intent-based hypotheses informed by search-result research, not verified search-volume or ranking-difficulty scores. Check current UK search results and existing site URLs before publishing. Search rankings, lead volume and conversions cannot be guaranteed.