A/B testing without the jargon
A/B testing means showing two different versions of the same page (or button, or headline) to similar groups of visitors at the same time, then measuring which version gets more people to take the desired action it is a way of letting real visitor behaviour, rather than personal opinion or guesswork, decide which version is genuinely better.
A simple example
Imagine you are unsure whether your CTA button should say "Get My Free Audit" or "Book a Free Call." Instead of guessing, or asking your team's opinion, you show half of your visitors one version and half the other version, at the same time, then see which version generates more actual clicks or form completions. Whichever version performs better, based on real visitor behaviour, is the one you keep using going forward.
Why "at the same time" matters
A common mistake is testing one version this month and a different version next month, then comparing results but this comparison is unreliable, because many other things change between two different time periods (seasonal factors, different ad campaigns running, general market shifts) that have nothing to do with which page version is actually better. Testing both versions simultaneously, to similar audiences, removes this confounding factor.
Why waiting for enough data matters
If you test two headlines and after only 20 total visitors one version has 3 conversions and the other has 1, this difference could easily be due to random chance rather than a genuine, real difference between the two versions small sample sizes produce unreliable, often misleading results. A reasonable rule of thumb for most small business tests: wait for at least 100 total conversions across both versions combined (not just visitors, conversions) before drawing a confident conclusion, and ideally run the test for at least one to two full weeks to account for any day-of-week variation in visitor behaviour.
What to test, and in what order
Start with high-impact, easy-to-change elements: headlines, CTA button text, and form length are typically the highest-leverage, lowest-effort tests available, since changing them requires no design or development work, just a text edit.
Test one thing at a time. If you change the headline, the CTA colour, and the form length all simultaneously, and conversion improves, you will not know which specific change actually drove the improvement isolating one variable per test, even though it takes longer overall, produces clearer, more reliable learnings.
Move to more involved tests once the basics are settled page layout, image choices, the overall structural order of elements these require more design effort and are reasonably tackled after the simpler, higher-leverage tests have already been run.
The tools available for small businesses
Google Optimize (Google's free A/B testing tool) was discontinued, but several alternatives exist at accessible price points or free tiers for small businesses including built-in testing features in many landing page builders (Unbounce, Leadpages) or dedicated tools like VWO or AB Tasty, which offer testing capability without requiring custom development for each test.
When A/B testing is not yet the right tool
For a business with very low traffic (a few dozen visitors per week), reaching statistically meaningful sample sizes through formal split testing can take many months in this situation, applying known best practices directly (the structural and copy principles covered throughout this pillar) rather than running formal tests is often the more practical, immediate path to improvement, with formal testing becoming more valuable once traffic volume grows.
Frequently asked questions
For simple tests, some landing page tools include built-in A/B testing features directly; for more basic situations, manually alternating which version is live during different equivalent time periods (while being aware this is less statistically rigorous than true simultaneous testing) can provide directional, if less precise, insight for very small businesses without testing software.
Generally a minimum of one to two full weeks, and ideally until reaching at least 100 total conversions across both versions combined, whichever takes longer ending a test too early risks concluding based on random noise rather than a genuine, reliable difference.
This is a valid and useful result, not a failure. It tells you that particular element does not significantly affect this specific outcome for your specific audience, allowing you to move on and test something else rather than continuing to debate a question the data has already answered as low-impact.