You can A/B test a small site — just not the way most guides describe. Under about 1,000 visitors a month, run sequential tests: change one thing, then measure a full 4–6 week period against the one before it. Spend your few tests a year on headline clarity, CTA wording, and form length, where effects are large enough to detect.
Why doesn't standard A/B testing work on small sites?
Go search "how to A/B test your website" and count how many articles casually assume you have ten thousand visitors a month. Statistical significance calculators. Split-testing platforms that want thousands of conversions before they'll call a winner. Cheerful advice to "run it for two weeks and see."
If your site gets 400 visitors a month, that advice is worse than useless. Follow it and you'll either wait forever for a result or, more dangerously, act on a "winner" that was pure noise.
An A/B test splits traffic between two versions of a page and measures which converts better. For the difference to mean anything, you need enough volume that random chance washes out — and for most realistic conversion-rate differences, "enough" is thousands of visitors per variant. Not hundreds.
Run a split test with 200 visitors per side and you will absolutely get a result. Version B wins by three points! Great. Except you have no way of knowing whether that's a real effect or the kind of swing that flips the other direction next week, because at that sample size, both look identical.
Nobody selling testing software leads with this, but it's the truth: most small business websites simply do not have the traffic to run a valid split test on a single page element in any reasonable window of time.
Sequential testing: the version that works at your scale
Instead of running two versions simultaneously, you test one change at a time and compare before against after over a longer window.
The process:
- 1.Establish a baseline conversion rate over at least 4–6 weeks. More if traffic is very low — you want weekly noise averaged out.
- 2.Make one specific change.
- 3.Run the new version for a comparable period. Same length, similar mix of weekdays, no seasonal weirdness (comparing December against a random October tells you about December, not your headline).
- 4.Compare the two periods.
Is this as rigorous as a proper split test? No. Outside factors can shift between periods — a seasonal dip, a change in ad spend, something happening in your industry — and you can't fully control for them. But it beats pretending a 200-visitor split test proves something, and unlike the textbook method, it works at the traffic level you actually have.
The cost is time. Each cycle takes a couple of months instead of a couple of weeks. That's the trade for trustworthy signal on limited traffic, and it's worth making.
What should you test when you only get a few tests a year?
If you can only run a handful of tests annually, spend them where the effect sizes are big enough to detect. Button color isn't on that list. These are:
Headline clarity. Does your hero headline say what you do, for whom, in words a stranger parses in five seconds? This is usually the highest-leverage change on any page — and often you don't need a live test at all. A five-second test with a few outside people (we describe it in our DIY conversion audit framework) tells you nearly as much, for free, in an afternoon.
CTA specificity. "Get Started" versus "Book a Free 15-Minute Call." Specific, low-commitment CTAs beat vague ones consistently, and the effect tends to be large enough to see even with modest traffic. This one deserves a real sequential test.
Form length. A HubSpot study found that cutting a form from four fields to three lifted conversions by nearly 50%. Effects that size don't need thousands of visitors to show up.
Social proof placement. A testimonial above the fold versus buried down the page. Trust signals produce real lifts, often visible within weeks.
Page speed. Not technically an A/B test, but treat any speed fix like an experiment anyway: note your conversion rate before, fix, measure after. Speed work produces some of the most reliable lifts of anything here.
What to skip
Micro-copy. "Submit" versus "Send Message" versus "Get In Touch" — whatever difference exists is too small to detect at your traffic level. Pick the clearest wording and move on with your life.
Color variations, unless something is genuinely failing contrast or accessibility. The internet is littered with case studies where a button color change "increased conversions 40%," nearly always from companies whose traffic and confounding variables go conveniently unmentioned.
Anything that only runs a few days. Weekdays and weekends behave differently. So do the days around payday (your analytics can show you exactly when your customers get paid, which is equal parts useful and unsettling). One week of data is noise with a start and end date.
And never more than one variable at once. Change the headline, the CTA, and the hero image together and you'll have a result with no idea what caused it. It's tempting to bundle changes to save time. Resist.
What A/B testing tools do you actually need?
Not Optimizely. Not a $500-a-month platform. This:
- Google Analytics 4, free (and probably already half-configured on your site since 2023), to track conversion rate over each testing period. Our analytics guide covers which numbers to watch.
- A spreadsheet logging what changed, when, and the before/after numbers. Primitive? Sure. It's also exactly right, and it forces the one-change-at-a-time discipline that makes the whole method valid (also, no vendor has ever sunset a spreadsheet).
- Microsoft Clarity, also free, for heatmaps and session recordings — so you can see why a change might be working, which matters more at low traffic where the quantitative signal alone is thin.
Paid split-testing platforms earn their keep once the specific page you're testing sees a few thousand visitors a month. Until then, they're an expensive way to generate inconclusive dashboards.
A calendar that respects your traffic
Testing constantly on low traffic just manufactures noise. A saner year looks like this:
Spend the first quarter fixing the high-confidence stuff without any formal testing — headline clarity, CTA specificity, form length. The research behind these is strong enough that you can just implement them.
In the second quarter, run one sequential test on your highest-traffic page. Usually the homepage or your best landing page.
Third quarter, run a second test, informed by what the first one taught you.
Fourth quarter, review. Which changes correlated with sustained lifts? Keep those, revert anything ambiguous, and plan next year's tests.
Four deliberate tests, run properly, will teach you more than twelve rushed ones that never reached a meaningful sample.
Some things don't need a test
Broken links. Confusing navigation. A contact form that fails on mobile. A page that takes eight seconds to load. Fix these today — you don't need an experiment to confirm that broken is bad. (If you do run that test, please send us the data.)
Testing capacity is scarce on a small site, so reserve it for genuine judgment calls: cases where reasonable people could disagree and where the answer depends on your audience rather than general best practice.
One more note: if you're building a new landing page from scratch, build it on established conversion patterns first and test refinements afterward. We've written up what makes a landing page convert — starting from that baseline beats testing your way out of a weak foundation.
Where this leaves you
The A/B testing most marketing blogs describe was never built for a site your size. Sequential testing — focused on high-leverage changes, run over honest timeframes — gets you real answers without pretending your traffic is something it isn't.
And if you'd rather hand the testing calendar to someone who's run this loop across dozens of sites, our landing page design service includes ongoing conversion testing as part of the engagement rather than a separate line item.
Not sure your site has the traffic to test properly, or the patience to run the process yourself? Talk to us — we'll tell you what's realistic for your numbers, and what to just fix outright.