A/B testing a landing page when you do not have much traffic

Work out the sample size first

A/B testing a landing page starts with a bit of maths. Run the numbers before you change anything. At a 3% conversion rate, spotting a 20% relative lift with 95% confidence and 80% power takes about 14,000 visitors per variant. That is 28,000 visitors for one test. If your landing page gets 2,000 visitors a month, the test would run for over a year. Knowing this on day one saves you from a test that was never going to finish.

Big changes need less traffic

The same maths works for you when the change is large. On that 3% baseline, spotting a jump to 4.5% takes about 2,500 visitors per variant. Push it to 6% and you only need about 750. A new headline, a new offer and a rebuilt page can move the number that much. A button colour cannot. Low traffic pushes you towards testing whole pages against each other, and this is usually the more useful question anyway.

Run one landing page A/B test at a time

When traffic is thin, splitting it across three variants leaves every arm too weak to read. Run one test, then the next one. Let each test cover full weeks, since Tuesday traffic and Sunday traffic rarely behave the same. Two to four whole weeks smooths out payday, campaign bursts and the odd viral post. Four tests run one after another over a quarter beat four run side by side that all end unclear.

Pick the end date before you start

Checking results every day and stopping the moment one variant hits 95% is the fastest way to fool yourself. On small samples, random noise crosses that line often, and stopping there locks the noise in. Decide the sample size and the end date up front. Write both down, and do not look at the significance number until you get there. If you cannot resist, hide the dashboard.

Track something that happens more often

Purchases are rare, so they take forever to reach significance. Events higher up the funnel fire far more often: primary CTA clicks, form starts, pricing section views, demo video plays. Those reach a readable sample much sooner. Treat them as a signal rather than a verdict, because a headline can lift clicks and still send the wrong people through. Confirm with revenue over the following month.

Five user tests beat a tied result

When the traffic will not settle the question, ask people. Five moderated sessions with buyers who match your target will surface the same two or three blockers again and again. Session recordings show you where the scroll stops. A one-question exit survey catches the objection you never thought of. Sales call notes are full of the exact words your page should use. None of this needs statistical power.

Ship on judgement when the test cannot decide

Most low traffic tests end with no clear winner, and this is a normal outcome. Do not stretch the run until the number tips your way. Write down what you learned, pick the version your qualitative evidence supports, ship it, and move to the next test. Over a year, six well chosen page rewrites will beat six months spent waiting on a result that a 1,200 visitor sample was never going to give you.