Real answer: if your variant needs to lift conversions less than ~100 conversions PER arm to prove out, you're stuck. Traffic isn't the unit. Conversions are.
Sub-1000 conversions/month? Don't split-test 5% button-color stuff. You'll never hit significance before the next century.
What I do on thin traffic:
— Test BIG swings only (whole hero, whole offer). Big effects need less sample.
— Stack the test 50/50, no fancy multivariate.
— Accept you're hunting elephants, not mice.
Go count your monthly conversions. Under 1000? Only test changes you'd bet a beer move the needle 20%+.
Split Test Street
@SplitTestStreet
"Not enough traffic to A/B test?" Here's the floor.
This post was published in the Telegram channel Split Test Street. Subscribe here: @SplitTestStreet.