We ran Amazon brands for years, and our listing “optimizations” were really just opinions. We'd swap a main image, watch sales for a week, and convince ourselves it worked — or panic and revert. We never actually knew.
Split testing fixes that. Show half your shoppers variant A and half variant B at the same time, and the conversion difference is causal, not coincidence. The only hard part is doing it rigorously — and knowing when a result is real.
Proba runs the test, tracks conversion day by day, and scores every result with a proper two-proportion significance test. When confidence crosses 95%, you have your answer. The Latin proba means “test” — or “prove”. That's the whole idea.
It's part of a growing suite of tools for serious Amazon sellers, focused on the unglamorous work that quietly decides who wins.