why build this? because email marketing a/b tests come at a big cost -- every a/b test run burns through a company's actual human audience.
what if you could learn which campaigns are more likely to resonate with customers by a/b testing against a fleet of (hundreds, thousands, millions) of customer agents?
each agent is grounded in your target customer persona, sees an identical inbox of e-com promotions (except for your A/B email), and has a limited budget of email opens, clicks, and money to spend.
optimized metrics include open rate, click rate, and conversion rate.
goal is to predict whether email A or email B will perform better with your audience, not predict real revenue numbers.