Most conversion rate optimisation advice assumes a volume of visitors that the majority of small and mid-sized businesses simply do not have. Run an A/B test with a few hundred visitors a month split across two variants, and the result you get back is closer to noise than signal — you could run the same test twice and get two different "winners" purely from chance. Treating that result as evidence is how a business ends up optimising for coincidence.
Why the usual advice doesn't apply
Standard A/B testing needs enough conversions per variant, over a reasonable time window, to distinguish a real effect from random variation. A site converting at two or three percent, with a few hundred monthly visitors, would need months to reach a result it could trust — and by the time it got there, the business, the offer, or the season would likely have changed underneath the test. This isn't a reason to give up on improving the page. It's a reason to stop relying on the one method that needs volume you don't have.
What actually works instead
Qualitative research fills the gap that statistics can't. Watching a handful of real people try to complete the page's main action — book, enquire, buy — surfaces the kind of problem no split test would ever catch: the button that's there but doesn't look clickable, the form field nobody understands, the price that's technically visible but easy to miss. Five sessions of someone visibly hesitating at the same spot is a stronger signal, at low volume, than a fractional lift in a test too small to trust.
Sequential changes are the second tool. Instead of splitting traffic between two versions and waiting for a result neither will reach fast enough, change the page, give it a real stretch of time — weeks, not days — and watch what happens to the conversion rate against its own recent history. It's a weaker method than a true A/B test in isolation, but it's the only one that produces a usable read at low volume, and it compounds: each change either sticks or gets reversed, and the page ratchets forward instead of staying frozen while it waits for enough traffic to test properly.
Bigger swings beat small tweaks
At high traffic, a two percent lift is worth detecting, because the volume makes it real money. At low traffic, a two percent lift is invisible — it's inside the noise. That changes what's worth testing. Button colour and micro-copy tweaks are the kind of change built for high-traffic testing; at low volume they're a waste of a testing cycle you don't have room to run. Bigger, more obviously different changes — a different headline making a different promise, a restructured page order, removing an entire step from the process — produce effects large enough to actually notice against a small sample, even without formal statistical confidence.
Reading the signal you do have
Session recordings, scroll depth, where people drop out of a multi-step form, and what they search for on the page if there's a search box — these all work regardless of volume, because you're reading individual behaviour, not averaging across a sample. A single recording of someone abandoning at the same point three times in a row tells you more than a thousand-visitor test with an inconclusive result. Low traffic doesn't mean no signal. It means the signal comes from watching, not from splitting.
Our website team builds this kind of low-volume optimisation into client work directly, because most of the businesses we work with never reach the traffic level formal split testing assumes — and waiting for it means leaving a weak page live for months on the theory that testing it properly requires patience the business doesn't have to spare. This sits inside a broader SEO and content strategy, because driving more visitors to a page that quietly loses most of them is the more expensive way to solve the same problem.
