Here's the lean version. Pick ONE variable, headline OR button OR hero, never three at once or you learn nothing. Define the success metric before you start, usually a single conversion event. Split traffic with a native feature or a lightweight script, 50/50, same traffic source. Run it until you hit a real sample, not until it 'looks good' on day two. Calculate significance with a free online calculator, don't eyeball a 3% lift on 80 visitors. Ship the winner, document it, then test the next thing. Most 'tests' are just two pages and a guess.
Am I wrong, or did you call a winner on 50 sessions?
Drag Drop Done
@DragDropDone
You can run a clean A/B test on a no-code site without a paid tool, if you follow the steps.
This post was published in the Telegram channel Drag Drop Done. Subscribe here: @DragDropDone.