A/B testing
Stop arguing about button copy
Run two versions of your primary call to action against live traffic, split the audience, and read the result. It replaces the meeting where everyone has an opinion and nobody has data.
Pro and above.
Test the thing that carries the click
The primary button is where nearly all the leverage is. Wording it as an instruction rather than a noun typically moves the rate more than a redesign does — but "typically" is doing a lot of work in that sentence, which is exactly why you test it on your own audience instead of trusting a blog post.
Set a goal
Without one there is nothing to compare against.
Write variant B
Change one thing. Two changes tell you nothing about either.
Choose the split
50/50 unless you have a reason.
Wait for volume
A day of traffic is noise. Give it a week.
The most common mistake is calling it early. Small samples produce large, confident, wrong differences.
A/B testing starts at Pro. See pricing.
Common questions
How much traffic do I need?
Enough that a few dozen conversions land in each variant. Below that you are reading randomness.
Can I test more than the button?
Today the test covers the primary button. Everything else you change by editing and comparing periods, which is weaker but still useful.