It's surprisingly simple. Split testing tools are really just feature toggles with RNG and a little tracking.
The stats and maths are mostly on the reporting side. We just rely on a bunch of Chi-squared hypothesis tests - that's about the hardest part to an A/B testing system.
The stats and maths are mostly on the reporting side. We just rely on a bunch of Chi-squared hypothesis tests - that's about the hardest part to an A/B testing system.